Method for achieving CONSISTENT PROFITABILTY

acrary   Registered: Apr 2002 Posts: 700

06-03-04 10:07 AM This thread is about the process of developing systems/methods and the detail that surrounds them. I've been semi-retired for about three years now. I'm now 48 and the older I get, the less desire I seem to have doing trading related work. At this point I have about six looseleaf notebooks filled with material I've not found elsewhere. I think some of it would be useful to just about anyone. There is some material I don't want to divulge while I'm still actively trading, so what I post about will vary and not be all inclusive. I have ADD, so this is probably the best way for me to post some material from some of my notebooks over time. In the past I've posted a idea or two when I had time and then moved on to other things. This has no doubt made others mad, so I hope to only hang out here and post followups. I'm not around all the time, so it may be days or weeks before I get back to this thread. I hope Magna won't close it without me requesting it. As far as the material, I'll post stuff that I believe is somewhat unique and has some value to people wanting to improve their overall trading or models and methods. I respect all opinions as they reflect the persons knowledge, values, and life experiences to that point. I don't respond to opinions as there's no way to argue objectively such matters. I try to post objective material so others can do the same calculations and come up with the same conclusions. If we differ on interpretations of the output, then so be it. I have no interest in participating in flame wars. Hopefully the differing opinions will not be edited out of this journal so others that share the viewpoint of the poster can PM them and do their own research. Please do not PM me with questions. I've let my mailbox fill up several times because I get overwhelmed when I have 15-20 PM's since the last time I logged in. Anything worth discussing is worth posting to this journal. If you have some secret formula and want my opinion, please keep it to yourself or ask someone else's opinion.

acrary   Registered: Apr 2002 Posts: 700	06-10-04 03:41 AM I hope you guys aren't offended if I ignore your questions for now. I'd like to spend my limited time posting some new info. from my notes to get things started. For the first topic, I tought I'd post on some work in the area of consistency. Any serious trader knows they are responble not only for profits, but also the minimization of drawdowns. The first thing I think of when I think of consistency is timeframe. I want to be consistently profitable over x timeframe. The x could be each day, a week, a month, a year, 10 trades, 50 trades, etc. Each person needs to decide the timeframe that they need to be consistently profitable. It's pretty common to see posts like I want to be a daytrader and make y dollars per-day every day. That trader has decided that their timeframe is one day for consistent profitability. Make a decison on a timeframe over which you want to be consistently profitable. In the coming posts I'll show you some tools to help reach your goal and show what it takes to reach it. Before I go any further I wanted to make sure everyone knows these two concepts. Expectation of a trade = (PW * AW) - (PL * AL) Expectation of profit factor = (PW * AW) / (PL * AL) where PW = probability of a winning trade AW = average size of winning trade PL = probability of a loss AL = average size of a loss Just about everyone is familiar with the expectation of a trade or just plain "expectation". I've not seen anyone ever post about the expectation of the profit factor so I'm mentioning it here. This will be useful in the following posts.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-10-04 04:04 AM Ok, now you have a timeframe for being consistently profitable. Now what does it take to achieve it? One more decision has to be made before continuing. What level of confidence do I need for consistent profitability. Some will be happy with 90%, others 95%, and some like me require 99%. So, to put it together in my case I require consistent profitability on a monthly basis with a 99% level of confidence. So in other words it's ok for me to have a losing month once every 8+ years. One series of studies I did was to find out what were the important factors in consistency. I did tests on size of expectation, % wins, profit factor, number of trades within a timeframe, and the effect of dispersion of trades (std. deviation) has on the results. For instance, here's a run for a daytrader that does 10 trades a day with 70% wins and makes $500 on each win and loses $500 on each loss. As you can see this is quite profitable at the 50% level (the average over time). However you can also see that after the 80% level the trader actually loses money. So for a daytrader with these numbers they only have a 80% chance of achieving their goal of consistent profitability on a daily basis. If that level of confidence is ok, then they have a example to use going forward. Attachment: daytrade.txt This has been downloaded 2485 time(s).

acrary   Registered: Apr 2002 Posts: 700	06-10-04 04:11 AM Now what happens if that daytrader decides he wants 95% level of confidence so that he'll accpect one losing day per month. What changes would he make? If he doubles his expectation does that help? As you can see from this run, he still only has about 80% confidence level with double the expectancy on each trade. Attachment: daytrade2.txt This has been downloaded 1322 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-10-04 04:26 AM How about if he increases the winning percent to 80% and keeps the expectancy at $400, so that each win now becomes $600 and each loss now becomes $400. Attachment: daytrade3.txt This has been downloaded 1026 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-10-04 04:51 AM Obviously after the last test this daytrader would know that with him winning 80% of the time and having a higher win size than loss size he'd be rich in record time. He'd now know that he has more than a 99% chance of making a profit each and every day he trades. What do think his psychology might be like. How about ..."time to trade can't wait to see how much I make today". Notice the expectancy didn't change from this test to the last one. I did this to show that expectancy really isn't a critical component of being consistently profitable. Remember a couple of posts ago I posted the expected profit factor. In this case it worked out as: Epf = (PW * AW) / (PL * AL) Epf = (.8 * 500) / (.2 * 400) Epf = (400) / (80) Epf = 6 or the 50% level described in the test Now that we've seen the expectancy didn't improve the likelihood of achieving consistency, what effect did changing the expected profit factor. We can use the equation to keep the same win % and win size to solve for the old PF of 2.33 at the 50% level as in the previous test. 2.33 = (.8 * 500) / (.2 * AL) 2.33 = 400/.2AL 2.33*.2AL = 400 .2AL = 400/2.33 or 171.67 AL = 171.67 / .2 or 858.35 When we plug in the new test of 80% winners, $500 win for each winner and 858.35 for each loser (to keep the profit factor at 2.33) we get these results. Attachment: daytrade4.txt This has been downloaded 978 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-10-04 05:00 AM From this past test we can see that if we kept the profit factor the same but changed the win % and expectancy, we'd have the same confidence level as we started with 80%. From this we can tell the win % and expectancy are not critical to consistency. One of the keys that is important is the expected profit factor. The higher the profit factor, the more liklely we are to achieve consistent profitabilty. What would have happened if instead of changing the win % we just changed from 10 trades per day to 20 trades per day. Then our daytrader would have a win% of 70%, win of $500, loss of $500 and twice as many opportunities per-day. Here's the results of that test. Attachment: daytrade5.txt This has been downloaded 868 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-10-04 05:15 AM As you can see from this last test by increasing the number of trades from 10 to 20 and keeping the profit factor, win %, and expectancy the same we've improved the confidence level to above 95%. So if our daytrader wanted to be 95% confident that he'd make money every day he could have also increased the number of trades per-day to achieve his goal. Now we have two variables that have an impact on consistent profitability (profit factor and frequency of trades). If our daytrader wasn't using strict targets and let the profits fluctuate and used a trailing stop in addition to the initial stop we'd increase the dispersion of the trades (std. deviation). For this example I'm using our 20 trades, 70% winners, $500 win, $500 loss, and letting the std. deviation of winners grow to $500 (100% of average) and std. deviation of losers grow to $250 (50% of average). This will let you see the effect of dispersion of outcomes on consistency. As you can see, by letting the winners ride a little and dragging a stop loss behind, the trade effect of the disperison of the trades on the overall results was there but not very important in impacting the consistency we're looking for. Attachment: daytrade6.txt This has been downloaded 711 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-10-04 05:46 AM The main idea of this first series of posts was to show what is important in moving toward becoming a consistent winning trader. By choosing a timeframe and then working on both the profit factor and number of trades you can move toward the goal of consistent profitability. Here's some rough estimates of the relationship between trade frequency and needed profit factor to achieve a 95% level of confidence that you'll be profitable within a timeframe for a single method. # trades......profit factor needed 10............4.00 20............2.50 30............2.00 40............1.75 60............1.50 So, if you wanted to be profitable every week at the 95% confidence level (you'd still have about 3 losing weeks per-year), and all your method could produce was a 1.50 profit factor, then you'd need 60 trades out of it to achieve your goal. From this you can tell that to achieve consistent profitability on a daily basis, either you've got some miracle system or you trade like a madman. In the next series of posts I'll go over using multiple systems to improve consistency. That's all I'm going to post today. I'll be in and out at least through the first hour of trading for the SP today, so if anyone has any questions on this topic I'll try to answer them. Have a good day!

acrary   Registered: Apr 2002 Posts: 700	06-10-04 07:32 AM Attachment: dsc389.txt This has been downloaded 1232 time(s). Quote from nov_trader: Acrary, Amazing stuff. " Model name daytrade # of trades in series 10 % of trades that are winners 70 Mean of winning trades 500 Std. Dev. of winning trades 0 Mean of losing trades 500 Std. Dev. of losing trades 0 Outcome Profit Factor Max DD 1% level 5,000.00 10.00 0 5% level 4,000.00 9.00 -500 10% level 4,000.00 9.00 -500" Could you please tell us, how did you get the probability of $5000 with 1% level.? Did you use some kind of Montecarlo analysis. Thx It's called a Monte Carlo Var Analysis. If you do search on Google you should find lots of info. For these tests I used a normal distribution curve for the results. I have another version that I use to simulate fat tails (which is slightly more representative of actual trading). For the big picture types of tests this one works well and is easy for others to replicate. To get an idea of how accurate this is, here's a test for one of my reject models called dsc389. I ran the test for monthly numbers (only 10 trades per-month), so I can show the dsc389 numbers from tradestation for comparison.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-10-04 07:43 AM From the test you can see it should win 70-80% of the time each month, have a monthly max drawdown of about 16k and a max profit of about 33k. Here's the graph of the monthly results for the past 5 years. The red indicates losing months and the green are winning months. As you can see there were 12 losing months out of 60 for 80% win rate, the actual max profit month was about 34k and the largest losing month was about 24k. I attribute the large losing month outside of the norm to the fat tail effect in the markets and is one of the reasons why I came up with another version to simulate the fat tails. Attachment: dsc389.gif This has been downloaded 1473 time(s).
		Edit/Delete • Quote • Complain

sprstpd   Registered: Apr 2003 Posts: 1745	06-10-04 11:02 AM Quote from acrary: Epf = (PW * AW) / (PL * AL) Epf = (.8 * 500) / (.2 * 400) Epf = (400) / (80) Epf = 6 or the 50% level described in the test Just trying to follow the math and got stumped on these equations til I noticed you used AW = 500 rather than AW = 600. If you use AW = 600, Epf = 6. But if AW = 500, Epf = 5.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-18-04 01:10 PM Quote from sprstpd: Just trying to follow the math and got stumped on these equations til I noticed you used AW = 500 rather than AW = 600. If you use AW = 600, Epf = 6. But if AW = 500, Epf = 5. Yes, you're right. On the post with the daytrade4.txt the Epf should be: Epf = (PW * AW)/(PL * AL) Epf = (.8 * 600)/(.2 * 400) Epf = (480)/(80) Epf = 6 or the 50% level described in the test Maybe Magna will be nice enough to edit that post so that it won't confuse others.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-18-04 01:31 PM The next series of posts will be about some of the methods I use to make sure I achieve my goal of consistency. In my case the goal is 99% chance of profitability per-month. If your goal is daily, weekly, or yearly profitability then just think "daily or yearly" for everything I say about the month. From the first set of posts you can see it's nearly impossible to reach a goal of 99% profitability every month based on trades or profit factor for a single method. Most of my models trade no more than 15 times per-month and the best Profit Factor I've developed that's consistent is around 2.50. With that in mind I knew I was going to have to use more than one method to achieve my goal. When I started out I had lots of questions like "How do I know if I should use more than one model?", "Since the numbers are different for each model how much should I trade of model x for every contract of model y?". I'll show you how I came up with some ideas about using more than one model to improve performance. I spent some hours going through my notebooks to pull out the couple of items to share. I think it'll be easier to understand if it's done through with examples so I've pulled the monthly performance numbers for 3 models and put them into a text file. If you want to work through the examples in a spreadsheet you'll need to save this file. To import it to Excel just use Tab delimited, and then for the first field change the format to Text. It should import correctly. Attachment: models.txt This has been downloaded 1765 time(s).

acrary   Registered: Apr 2002 Posts: 700	06-18-04 02:01 PM The 3 models all trade the same market (SP). I don't want to share the exact details so I'll post some of the overview on each. The file contains the monthly numbers for each model from 1/1999 - 5/2004. I chose the 3 models because they give a wide spread of numbers like you might find in different methods. Things like win rate, profit factor, number of trades, and total profit are spread out so it'd be tough to just guess at whether to trade 1,2, or all 3 of the models and how much to trade of each one (before money management). Model 1 Trades about 120 times per-year or about 10 trades per-month The profit factor is 1.99 and 41.72% of the trades have been profitable. The total profit over the time period is $1,551,005. The Max drawdown during the period was $54,095. 10 of the 65 months were losers. Model 2 Trades about 145 times per-year or about 12 trades per-month. The profit factor is 1.70 and 59.37% of the trades have been profitable. The total profit over the time period is $1,365,415. The Max drawdown during the period was $90,170. 18 of the 65 months were losers. Model 3 Trades about 90 times per-year or about 7.5 trades per-month. The profit factor is 1.57 and 48.78% of the trades have been profitable. The total profit over the time period is $495,365. The Max drawdown during the period was $97,000. 21 of the 65 months were losers. This is what the speadsheet should look like after importing the text file. Attachment: models.gif This has been downloaded 1536 time(s).
		Edit/Delete • Quote • Complain

steve46   Registered: Mar 2003 Posts: 3575	06-18-04 02:41 PM Hello: I wonder how important consecutive losers and winners is to your evaluation of a system. I assume you that the drawdowns we see are acceptable, but how far from these figures would you allow it to go before withdrawing the system from trading? For my own systems, I look at standard deviation of profits on a monthly basis as a warning signal. I look forward to your reply. Steve46
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-18-04 02:52 PM The first of the important concepts is to avoid trading correlated methods. You've probably read in some system development books that you should use negatively correlated methods. That's nice to hear but how do you achieve it? In most cases you don't, however you can achieve non-correlation by avoiding using the same stop methods, or same entry methods from one model to the next. I'll go over creating negatively correlated methods in another series but right now I just want to show the benefits of non-correlation. Using the 3 models as a example, I've added the monthly numbers for each model into column E. Then at the bottom I computed the Monthly average, Monthly Min, Monthly Max, and Total net profit for all the methods. Below the numbers is the correlation coefficient for each of the models versus the other. For example, C74 is the correlation between the monthly numbers for model 1 and model 2. D74 is the correlation for model 1 and model 3, and D75 is the correlation for model 2 and model 3. From my experience a number between -.2 and +.2 usually means a random correlation between the numbers. As you can see each of these has a slightly negative correlation which to me means they aren't correlated. In columns E68-E71 you can see the benefit of using the systems together. In this case the total monthly profit jumps up and the largest monthly loss is actually lower than 2 of the 3 models individually. Imagine if you could trade model 1 and then add the other 2. When the methods aren't correlated, the profits go up without increasing drawdowns. When combined, the number of profitable months grows to 58 out of 65 for the entire period, or about 89% of the months. Compare this with model 1 (55/65 or about 85% profitable), model 2 (47/65 or 72% profitable), and model 3 (44/65 or 68% profitable). Just from this technique you can see there is some benefit to trading multiple models. Here's what I'm looking at: Attachment: sum.gif This has been downloaded 1382 time(s).

acrary   Registered: Apr 2002 Posts: 700	06-18-04 03:15 PM Quote from steve46: Hello: I wonder how important consecutive losers and winners is to your evaluation of a system. I assume you that the drawdowns we see are acceptable, but how far from these figures would you allow it to go before withdrawing the system from trading? For my own systems, I look at standard deviation of profits on a monthly basis as a warning signal. I look forward to your reply. Steve46 Steve, the number of winners or losers in a row has no importance to me as long as the trades are independent. If I see numbers outside of normal bounds then I search for dependency (loss begets loss, etc.). To know if the number of wins or losses is outside of the normal bounds I created a formula to estimate the number of winners or losers in a row I should expect to see. The basic formula for figuring the expected maximum losing streak is: S = ln(1/T)/ln(L) where: L = % losers S = Streak T = # trades Ex. T = 500 trades L = .6 or 60% losers S = ln(1/500)/ln(.6) S = -6.21461/-.51083 S = 12.16581 or a expected max. losing streak of 13 trades If you increase the number of trades to 1,000 then: S = ln(1/1000)/ln(.6) S = -6.90776/-.51083 S = 13.52273 or a expected max. losing streak of 14 trades confidence level is 1 - (1 / number of trades) ex. 500 trades = 1 - ( 1/ 500) = .998 confidence level In this case if I were to see 20 losers in a row from a test sample of say 400 trades, then I'd check other streak levels such as the number of 3 in a row or 4 in a row to see if they are also outside the bounds. If so, I'd look for dependency. As you'll see from this series on consistent results it would very rare for me to see a 10% drawdown. As long as the model is operating with a consistent edge I would not pull it. The range of drawdowns is somewhat predicatable using the Monte Carlo tests. Those, combined with these other ideas should ensure that I'll have few and far between losing months (at least they have up until now).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-18-04 03:23 PM Quote from ig0r: are those systems curve-fit or are those forward-tested numbers? not that it really matters in the context of the lesson but they have some pretty spectacular results and I was just curious No, they aren't curve fit. I've done some work on model 1 last year because it has an edge and I wanted to use that part of it for another model. The other two have been around for ages but I'm not doing anything with them because they have no edge. The results are based on 1 unit size (not 1 contract). A unit is determined by largest market volatility divided by the current market volatility so the number of contracts varies to keep the results consistent with market volatility. No money management was applied to any of this (if I did, it'd be obvious).

acrary   Registered: Apr 2002 Posts: 700	06-18-04 04:41 PM One other thing about correlation I wanted to post was that even if we found a negative correlation, do a couple checks. If we had say -.4 between two models you might think you found a great negative method to complement your primary model. Another way I found to check it is to create a table of correlations with rolling periods (in the case of monthly 12 periods). If the correlation isn't consistent and smooth then you're probably looking at fools gold (expect the correlation to revert to random over time in the future). Using our 3 models I've put in columns for the correlations between each of the models. You can see by inspection that the 12 month rolling correlations are moving around wildly. This is a pretty clear example of random correlations between the models. Attachment: rolling.gif This has been downloaded 1053 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-18-04 04:49 PM Correlation cont'd. Here's a screenshot of crude versus unleaded gas using 30 day average correlations between the two. Over the past 5 years the correlation was greater than 95% so you'd expect them to be highly correlated. This is what it looks like when they are. Attachment: gas.gif This has been downloaded 1307 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-18-04 05:42 PM Correlation cont'd. Ok, so far most of this is probably old news to most of you. I needed to post it as background for those just getting started. The whole idea of combing models is to improve the consistency of the results. We've seen in a macro way how they can provide some benefit. Now, how do we know which models to combine? Also how much of model 1 should we trade with how much model 2 and how much of model 3? Also if I had a model 4 how could I tell if I should add it in as well? Those are questions I wanted to answer when I started down this path. Hopefully it won't be too difficult to follow. If you've taken a elementary stats class you'll know that std. deviation measures dispersion from the norm. Norm being defined as the median or average. You'll also note that the level of std. deviation measures the distance away from the median. We also know that if we can estimate the number of std. deviations away something is, we can look it up as a Z score in a normal distribution table to see how far away from the norm we are. One of the nice tools in trading are the sharpe ratios which are designed to measure consistency. One of them is the modified sharpe ratio. It's defined as average return / std. deviation. For example, if the average return is $100 and the std. deviation is $50 then the sharpe ratio is 2.0. So in other words for me to breakeven or start to lose money I'd have to have a return that was 2 std. deviations worse than the average return. If you look up 2 std. deviations in a stats book under z score you'll note that it equates to 95.44%. So if my returns are normally distributed then I have a 95.44% chance of breaking even or making a profit. The modified sharpe ratio can be thought of as a z score. The higher the number, the closer we are to achieving consistent profitability. In my case I want to be 99% sure of making a profit each month. So if I were to look it up in a normal distribution table I'd know the Z score I need is approx. 2.58 or a modified sharpe ratio of 2.58. We also know that the returns are not going to be normally distributed. There will be fat tails so whatever sharpe ratio we come up with, it will be higher than we should expect in normal trading. This is where using the modified sharpe ratio and non-correlated methods pays off. I don't have a way to measure directly the benefit of using non-correlated methods but I know it improves the smoothness of the equity curve. If I combine that with a high modified sharpe ratio I can have a high degree of confidence that my models will be consistently profitable.

acrary   Registered: Apr 2002 Posts: 700

06-18-04 06:07 PM Correlation cont'd. In this example I've taken our 3 models and applied the modified sharpe ratio in column F. I believe the normal method of computing the modified sharpe ratio is to use 36 periods. In this example I've only chosen to use 12 periods. If I used 30+ periods of the 65 total I'd have very little information to evaluate. I included the 36 period numbers in column G. As you can see the numbers in column F are all over the place. I think is because the sample size of 12 periods is too small to get reliable numbers. You can see it's much smoother in column G. In a sec. I'll post how I try to use most of the data and adjust it so we can get a feel for the overall modified sharpe ratio. Attachment: modified.gif This has been downloaded 967 time(s).

acrary   Registered: Apr 2002 Posts: 700

06-18-04 06:23 PM Correlation cont'd. To come up with a single modified sharpe ratio what I've done is simply average all of the 12 period modified sharpe ratio's. In this way I've smoothed it and used most of the data so the results are more likely to represent what is going on. On this screen I've highlighted it as 1.44 in column F. In column G I've posted the average of the 36 period number. As you can see they are resonably close. What does 1.44 represent? It means we'd have to have a month that was more than 1.44 std. deviations away from the average before we'd expect to lose money. If we look it up as a z score we'd see 85.02%. So 85% of the time we'd expect to have a winning month. 85% of 65 months is 55 1/4 months, so we'd expect to see about 10 losing months. In our case we have only 7 losing months. I have no way to measure it but my guess is the other 3 months are improved by the 3 methods being non-correlated. Attachment: ave.gif This has been downloaded 748 time(s).

acrary   Registered: Apr 2002 Posts: 700

06-18-04 07:05 PM Correlation cont'd. With the one number I now have a way of doing comparisons. If I combine say model 1 and model 2 and run the same tests I'll get one result. If I combine model 2 and 3 I'll get another. And if I combine 1 and 3 still another. Then if I do many runs in which I weight each pass say 1 unit of model 1 and 2 units of model 2 I'll get more results. In the end what I found I had to do was develop a program to do all these passes. It weighs each model from 1 - 100 units and determines the optimal modified sharpe ratio. Then it determines the ratio between each of the methods to determine how much of each should be traded. If I have 4 models it'll do tests on all four of them, five, six, 10, 50, etc. All it takes is compute time. For 3 models it takes about 15 sec. for 30 it takes about 1 1/2 hours. If I choose to do all my models, I leave for a couple of days. In the end it gives me the optimal balance for the best modified sharpe ratio. For my trading I had to use 8 very good models to get the number up to 2.58 so that's why I trade against 8 models. In our 3 model example I ran my program and here were the results: combine model 1 & 2 Best modified sharpe ratio 1.395 using 1 unit of model 2 to 1.5 units of model 1 combine model 1 & 3 Best modified sharpe ratio 1.666 using 1 unit of model 1 and 1.333 of model 3 combine model 2 & 3 Best modified sharpe ratio 1.057 using 1 unit of model 2 and 2.04 units of model 3 combine model 1 & 2 & 3 Best modified sharpe ratio 1.461 using 1 unit of model 2, 1.26 units of model 1, and 1.08 units of model 3 From this test I can see I should be trading only models 1 and 3 in the ratio shown to achieve the maximum consistency. Adding model 2 actually reduces the consistency. At 1.66 I should see 90.3% of the months being winners. Not up to the 99% level but a definite improvement. If I had a fourth model I could do the same test and see if it should be included, and if so, what the optimal ratio should be. It might seem silly to use a lower profit factor, lower win %, and lower total profit model but I'll show the comparison and why I'd do it.

acrary   Registered: Apr 2002 Posts: 700	06-18-04 07:44 PM Correlation cont'd. Well it didn't turn out as I had expected. I applied the weightings as spit out by the program and did the same totals as was done earlier. In column E are all 3 models combined with the optimal weighting. In column F is the combination of models 1 & 2 and in column G is the combination of models 1 & 3. It's pretty easy to see that using 1 & 3 is superior to 1 & 2 mainly because the largest losing month for 1 & 2 is more than double the largest loss for 1 & 3. So if you were to trade at twice the size using 1 & 3 you'd have a lower max losing month and a higher average profit per-month than 1 & 2. However this shows that if you used 1,2, and 3 in the ratios provided it would kick some serious butt over using 1 & 3 by itself. Also 1 & 3 have 11 losing months as compared to just 7 for all 3 models combined. In digging through the data I noted that model 3 didn't have very consistent results so the 1.57 profit factor probably isn't representative of it's overall results. In fact model 3 lost money for all of 1999 (a year in which the combined results for 1 & 3 show 6 of the 11 losing months). This just shows once again that the concept is only as good as the underlying models it's built on. Attachment: tot.gif This has been downloaded 757 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-18-04 08:02 PM Correlation cont'd. Well, here's the snapshot of model 3's annual results. As you can see the win % ranges from 35 - 60 and the profit factor ranges from .79 - 2.89. Obviously this was a bad choice of a model to include in the tests. If the models are inconsistent, then everything else that flows from them will be suspect. One thing I meant to point out was look at the largest losing month for all 3 models combined with the weightings. Notice it's now lower than any of the models individually. I love this stuff. I hope I was able to convey the general concept without getting too deep. I spent hours on this trying to figure out how to boil it down to these few posts. In reality this took more than one full notebook of work over a long period of time. If you have any questions I'll try and answer them tomorrow morning. I'm going out to dinner now and I'll be around then. Enjoy the weekend!

EricP   Registered: Dec 2001 Posts: 702

06-19-04 11:24 AM Quote from acrary: Correlation cont'd. Well, here's the snapshot of model 3's annual results. As you can see the win % ranges from 35 - 60 and the profit factor ranges from .79 - 2.89. Obviously this was a bad choice of a model to include in the tests. If the models are inconsistent, then everything else that flows from them will be suspect. One thing I meant to point out was look at the largest losing month for all 3 models combined with the weightings. Notice it's now lower than any of the models individually. I love this stuff. I hope I was able to convey the general concept without getting too deep. I spent hours on this trying to figure out how to boil it down to these few posts. In reality this took more than one full notebook of work over a long period of time. If you have any questions I'll try and answer them tomorrow morning. I'm going out to dinner now and I'll be around then. Enjoy the weekend! Thank you very much for the ongoing sharing of your thoughts on systems development. It has been very helpful and is very much appreciated. I'm hoping you might have thoughts on the following question. I use automation to trade individual stocks, and use systems on many different stocks. At any given time, I might be paper trading 300+ stocks, and have 100+ stocks active for live trading. I use a 93% confidence level criteria over the last 120 paper trades to determine whether to activate a security for live trading. In this way, I activate strictly based upon the risk/return of a specific security's past profit performance. However, I really like the way you activate based upon the actual diversification that the new security (or system) adds to your overall profitability. The issue I face is liquidity. For individual stocks, I run into liquidity issues for the less active stocks ( Thanks for your thoughts. -Eric

acrary   Registered: Apr 2002 Posts: 700	06-20-04 08:12 AM Sorry I was away yesterday.Weather forecast for yesterday was great and today was supposed to be bad so I went flying. Today I'll be around for most of the day. In my haste to get finished with last topic I made a mistake. I went back this morning to try and understand how models 1 & 3 couldn't do better than 1,2, and 3 combined. With the difference in the two numbers being about .2 I expected different results. It's also been my experience that a .2 improvement shows in the overall performance numbers. When I re-ran the tests, I found the correct number for 1 & 3 was 1.166 not 1.66. Sorry for the mistake. I'm including a edited copy of that post for anyone cutting and pasting this stuff. Correlation cont'd. With the one number I now have a way of doing comparisons. If I combine say model 1 and model 2 and run the same tests I'll get one result. If I combine model 2 and 3 I'll get another. And if I combine 1 and 3 still another. Then if I do many runs in which I weight each pass say 1 unit of model 1 and 2 units of model 2 I'll get more results. In the end what I found I had to do was develop a program to do all these passes. It weighs each model from 1 - 100 units and determines the optimal modified sharpe ratio. Then it determines the ratio between each of the methods to determine how much of each should be traded. If I have 4 models it'll do tests on all four of them, five, six, 10, 50, etc. All it takes is compute time. For 3 models it takes about 15 sec. for 30 it takes about 1 1/2 hours. If I choose to do all my models, I leave for a couple of days. In the end it gives me the optimal balance for the best modified sharpe ratio. For my trading I had to use 8 very good models to get the number up to 2.58 so that's why I trade against 8 models. In our 3 model example I ran my program and here were the results: combine model 1 & 2 Best modified sharpe ratio 1.395 using 1 unit of model 2 to 1.5 units of model 1 combine model 1 & 3 Best modified sharpe ratio 1.166 using 1 unit of model 1 and 1.333 of model 3 combine model 2 & 3 Best modified sharpe ratio 1.057 using 1 unit of model 2 and 2.04 units of model 3 combine model 1 & 2 & 3 Best modified sharpe ratio 1.461 using 1 unit of model 2, 1.26 units of model 1, and 1.08 units of model 3 From this test I can see I should be trading all the models in the ratio shown to achieve the maximum consistency. At 1.461 the zscore translates into 85.56% winning months (about 1 3/4 losing months per-year). Not up to the 99% level but a definite improvement over any combination of two systems. If I had a fourth model I could do the same test and see if it should be included, and if so, what the optimal ratio should be.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-20-04 09:14 AM Quote from bdixon619: How does one go about finding markets for these systems? Is each system you develop market specific? Or, was a system design inspired by experience with one market or another? What individual market tendencies do you look for before selecting a system for trade in that market? How do you characterize a market? I don't look for a market for a system. Each system I've developed is targeting some behavior. If that behavior is present in multiple markets, then I could test it to see if my system captures the behavior better than random. If so, I'd just trade it on that market and check to make sure the behavior was persistent. For instance I have a volatility breakout model that I've used successfully in the SP market. I tested it against the DAX market and found the edge (ability to capture profits at better than random), was better in the DAX than the SP. I've been trading the DAX with it since then and it's done very well. Only thing I don't like is getting up in the middle of the night to trade. Every model I've worked on has gone through the same process. Look at the behavior's present in a market, characterize them by creating a rule and checking the fit until all behavior's are noted. Then start looking to see if there is a component to the behavior that is non-random. If so, develop a system to mine it and create a way to monitor the behavior to ensure it's persistent over time. For example, one of the behavior's widely known is the trend day in the SP market. It can be identified just by visually inspecting a chart. I characterized it as a low/high within 10% of the low/high of the day and the close within 20% of the high/low of the day. With the definition I can see how many of these days have persisted over the years (averages about 25 days per-year). Then I can see if there is a way to identify these days in advance (realizing I'm going to also be capturing some false days as well).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

06-20-04 09:35 AM Quote from EricP: Thank you very much for the ongoing sharing of your thoughts on systems development. It has been very helpful and is very much appreciated. I'm hoping you might have thoughts on the following question. I use automation to trade individual stocks, and use systems on many different stocks. At any given time, I might be paper trading 300+ stocks, and have 100+ stocks active for live trading. I use a 93% confidence level criteria over the last 120 paper trades to determine whether to activate a security for live trading. In this way, I activate strictly based upon the risk/return of a specific security's past profit performance. However, I really like the way you activate based upon the actual diversification that the new security (or system) adds to your overall profitability. The issue I face is liquidity. For individual stocks, I run into liquidity issues for the less active stocks ( Thanks for your thoughts. -Eric You already know, as liquidity goes down time in trade must go up. Applied to your situation, you might have to create a smaller pool of securities that have sufficient liquidity for your current strategy. Then create another method that uses a longer holding time for the less liquid securities. This way you'd be diversifying by method, time, and securities. You could treat each method and the secuities as a single pool and test the correlation between the two strategies. If the srategies were found to be non-correlated you could take on larger size and reap the benfits of diversification.

acrary   Registered: Apr 2002 Posts: 700	06-20-04 09:50 AM Quote from ig0r: acrary, can you post the algorithm you use to adjust the number of contracts traded to volatility? Is it similar to the one used in the turtle trader method? I don't know what turtle trader uses. I'm sure my method is pretty simplistic. For daytrading I just calculate the range (high - low), then average it for the past ten days. I use ten because I want my model to cut back on size pretty quickly if the volatility jumps. Then I divide the highest historical 10 day volatility (approx. 48 pts.) by the current volatility (ex. 8 pts) to come up with a multiplier (ex. 6). The model would then apply 6 contracts for the next trade. This is not the final size used to trade. It's just used to adjust the model for volatilty levels so I measure one period against another without volatilty being a consideration. By doing so, I can see if the same level of opportunites persist from period to period. I can also use these normalized trades to feed into money management models as well as Monte Carlo tests to estimate future performance and drawdowns. If you were to use trades from say 2000 and 2004 for the SP market in a Monte Carlo test without normalizing volatility you'd get a very distorted estimate of future performance.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	06-20-04 10:20 AM Quote from onelot: acrary, regarding the correlation of multiple instruments versus model correlations: after finding no correlation between instruments should we look at the instruments as separate models if we are using multiple instruments to one model in order to carry on the sharpe work? i'm not sure if that's what you were implying here in response to virgins post: so for instance, instead of the spread sheet comparing model1>model2>model3 to one instrument it could compare instrumentX>instrumentY>instrumentZ to one model, no?... as opposed to only measuring say the correlation of the 30 day price average between instruments. Yes, that's the idea. You'd just have to make sure the periods were identical (in this case monthly). If you were doing long term trendfollowing you might have to switch to quarterly or longer timeframes for comparison due to the fewer trades. just to make sure i understand the big picture, this is all being done to increase frequency in the desired profitable timeperiod, correct? so if i'm understanding, sub-par models tested individually with low frequency can be morphed into an above-par model when combined with other non-correlated sub-par models (assuming they're not too sub-par), thus increasing frequency, consistency, and lowering the need for a higher profit factor? thank you for presenting the information the way you did, i would not have made that connection otherwise (assuming i'm on the right track). fascinating. Yes, you understand it perfectly. If you ever trade money for others the first thing they want to know is that you've done everything possible to avoid losing periods. If you tell them "my performance period is one month and I've planned my trading so that there is a 95% chance of a profit within any single period", they'll be happy to hear what else you have to say. You'll also see by working hard on consistency that you have better money management options available to you than "risk no more than 2% of capital on any trade".
		Edit/Delete • Quote • Complain

Profit Factor	Number of Required Trades    - Chart Provided to ETers by "acrary"
	5	10	15	20	25	30	35	40	45	50	55	60	65	70	75	80	85	90	95	100	105	110	115	120
1.40	61.8	67.0	70.8	73.7	76.3	78.3	80.1	81.8	83.2	84.6	85.4	86.7	87.5	88.6	89.4	90.1	90.7	91.5	91.9	92.7	93.1	93.5	93.9	94.4
1.42	62.0	67.7	71.8	74.5	77.2	79.1	80.9	82.6	84.0	85.6	86.8	87.6	88.8	89.4	90.3	91.2	91.7	92.3	92.8	93.4	93.7	94.2	94.7	95.1
1.44	62.7	68.3	72.3	75.2	78.0	80.3	82.1	83.6	85.0	86.3	87.5	88.8	89.4	90.3	91.1	92.0	92.5	93.1	93.7	94.1	94.4	94.9	95.3	95.7
1.46	63.1	69.1	72.8	76.1	78.9	81.0	83.0	84.4	86.0	87.1	88.4	89.4	90.3	91.2	91.8	92.6	93.2	93.8	94.3	94.7	95.1	95.6	95.9	96.3
1.48	63.3	69.5	73.9	76.9	79.5	81.8	83.6	85.3	86.9	88.2	89.1	90.1	91.1	92.0	92.6	93.3	93.9	94.3	94.9	95.4	95.7	96.0	96.4	96.7
1.50	63.9	69.9	74.5	77.6	80.3	82.5	84.5	86.3	87.5	88.8	90.1	91.1	91.9	92.6	93.3	93.9	94.5	94.9	95.6	95.8	96.2	96.5	96.8	97.1
1.52	64.4	70.7	74.9	78.7	81.2	83.5	85.3	86.9	88.3	89.7	90.6	91.6	92.4	93.3	93.9	94.4	95.1	95.5	95.9	96.3	96.6	96.9	97.4	97.5
1.54	64.9	71.1	75.6	79.2	82.0	84.3	86.0	87.8	89.0	90.3	91.4	92.3	93.1	93.8	94.4	95.0	95.4	96.0	96.2	96.7	97.1	97.2	97.6	97.8
1.56	65.2	71.8	76.0	79.9	82.4	85.0	86.9	88.2	89.7	91.0	91.9	92.9	93.6	94.3	95.1	95.5	96.0	96.4	96.7	97.1	97.4	97.7	97.9	98.1
1.58	65.3	72.2	77.1	80.5	83.1	85.5	87.4	89.1	90.3	91.6	92.4	93.5	94.1	94.9	95.5	95.9	96.4	96.8	97.1	97.4	97.7	98.0	98.2	98.3
1.60	65.9	73.2	77.4	81.0	83.9	86.3	88.0	89.7	90.9	92.0	92.9	93.8	94.5	95.2	95.8	96.3	96.6	97.1	97.4	97.7	97.9	98.2	98.4	98.6
1.62	66.5	73.4	78.2	81.5	84.5	86.8	88.7	90.1	91.4	92.6	93.6	94.4	95.1	95.7	96.2	96.7	97.0	97.4	97.8	98.0	98.2	98.4	98.6	98.8
1.64	66.8	74.0	78.5	82.3	85.1	87.5	89.2	90.7	92.0	93.1	94.0	94.8	95.4	96.1	96.5	96.8	97.3	97.7	98.0	98.3	98.5	98.6	98.8	98.9
1.66	67.0	74.4	79.3	82.9	85.5	87.9	89.8	91.0	92.5	93.7	94.5	95.3	95.9	96.4	97.0	97.2	97.6	97.9	98.2	98.4	98.7	98.8	99.0	99.1
1.68	67.3	74.8	79.7	83.5	86.3	88.4	90.3	91.7	92.8	93.9	94.8	95.6	96.2	96.7	97.1	97.6	97.9	98.1	98.4	98.6	98.8	98.9	99.0	99.2
1.70	68.1	75.5	80.6	83.9	86.7	88.9	90.8	92.1	93.3	94.4	95.3	95.8	96.6	97.0	97.4	97.8	98.1	98.4	98.6	98.7	98.9	99.1	99.2	99.3
1.72	68.3	75.9	80.8	84.3	87.2	89.4	91.0	92.5	93.8	94.7	95.6	96.2	96.8	97.3	97.7	98.1	98.3	98.5	98.7	98.9	99.1	99.2	99.3	99.4
1.74	68.8	76.4	81.3	84.9	87.7	89.9	91.6	93.0	94.1	95.0	95.8	96.6	97.1	97.5	97.9	98.2	98.4	98.7	98.8	99.1	99.2	99.3	99.4	99.5
1.76	68.9	76.9	81.6	85.6	88.0	90.1	92.0	93.4	94.4	95.5	96.2	96.8	97.2	97.6	98.1	98.4	98.6	98.8	99.0	99.2	99.3	99.4	99.5	99.5
1.78	69.0	77.2	82.4	85.9	88.6	90.8	92.3	93.7	94.8	95.7	96.5	97.0	97.4	97.9	98.2	98.5	98.8	99.0	99.1	99.3	99.4	99.5	99.6	99.7
1.80	69.5	77.6	82.4	86.3	88.9	91.2	92.7	94.1	95.2	96.0	96.7	97.1	97.8	98.1	98.4	98.7	98.9	99.1	99.3	99.3	99.5	99.6	99.6	99.7
1.82	70.0	78.0	83.0	86.7	89.3	91.3	93.1	94.3	95.4	96.2	97.0	97.4	98.0	98.3	98.5	98.8	99.0	99.2	99.3	99.4	99.5	99.6	99.6	99.7
1.84	70.0	78.2	83.4	87.2	89.8	91.9	93.4	94.7	95.6	96.5	97.2	97.6	98.1	98.4	98.7	98.9	99.1	99.3	99.4	99.5	99.6	99.6	99.7	99.7
1.86	70.5	78.9	83.9	87.5	90.2	92.2	93.9	95.1	95.9	96.7	97.3	97.9	98.3	98.6	98.8	99.0	99.3	99.3	99.5	99.6	99.7	99.7	99.8	99.8
1.88	70.7	79.2	84.3	87.9	90.5	92.4	94.1	95.3	96.1	97.0	97.5	98.1	98.4	98.6	98.9	99.1	99.3	99.4	99.5	99.6	99.7	99.8	99.8	99.8
1.90	70.9	79.4	84.6	88.2	90.9	92.9	94.3	95.5	96.4	97.2	97.7	98.2	98.5	98.8	99.1	99.2	99.3	99.5	99.5	99.7	99.8	99.8	99.8	99.8
1.92	71.4	79.8	85.2	88.5	91.4	93.1	94.7	95.8	96.6	97.2	97.9	98.3	98.7	98.9	99.1	99.3	99.5	99.5	99.6	99.7	99.8	99.8	99.8	99.9
1.94	71.7	80.3	85.5	88.9	91.6	93.4	94.9	95.9	96.9	97.5	98.0	98.5	98.8	99.0	99.2	99.4	99.4	99.6	99.7	99.7	99.8	99.8	99.9	99.9
1.96	72.0	80.6	85.8	89.4	91.9	93.7	95.2	96.3	97.1	97.8	98.2	98.5	98.9	99.1	99.3	99.4	99.6	99.6	99.7	99.8	99.8	99.9	99.9	99.9
1.98	72.5	80.8	86.1	89.4	92.0	94.1	95.4	96.4	97.1	97.9	98.3	98.6	99.0	99.2	99.3	99.5	99.6	99.7	99.8	99.8	99.8	99.9	99.9	99.9
2.00	72.3	81.1	86.4	89.9	92.5	94.2	95.8	96.6	97.4	97.9	98.4	98.8	99.1	99.2	99.4	99.5	99.6	99.7	99.8	99.8	99.9	99.9	99.9	99.9
2.02	72.7	81.5	86.4	90.2	92.8	94.5	95.8	96.7	97.5	98.0	98.6	98.9	99.1	99.3	99.4	99.6	99.7	99.8	99.8	99.8	99.9	99.9	99.9	99.9
2.04	73.2	81.8	87.1	90.4	92.9	94.7	96.1	97.0	97.7	98.2	98.7	99.0	99.2	99.4	99.5	99.6	99.7	99.8	99.8	99.9	99.9	99.9	99.9	99.9
2.06	73.3	82.0	87.4	90.8	93.3	95.0	96.2	97.2	97.9	98.4	98.7	99.0	99.2	99.4	99.5	99.7	99.7	99.8	99.9	99.9	99.9	99.9	99.9	99.9
2.08	73.3	82.5	87.7	91.2	93.6	95.1	96.4	97.4	98.0	98.4	98.8	99.2	99.3	99.5	99.6	99.7	99.8	99.8	99.9	99.9	99.9	99.9	99.9	99.9
2.10	73.9	82.5	88.1	91.4	93.7	95.4	96.5	97.4	98.0	98.5	98.9	99.2	99.4	99.5	99.7	99.7	99.8	99.9	99.9	99.9	99.9	99.9	99.9	99.9
2.12	73.9	83.1	88.2	91.8	94.1	95.6	96.8	97.6	98.2	98.6	98.9	99.2	99.4	99.6	99.7	99.7	99.8	99.8	99.9	99.9	99.9	99.9	99.9	99.9
2.14	74.3	83.2	88.7	91.8	94.1	95.8	96.9	97.7	98.2	98.7	99.0	99.3	99.5	99.6	99.7	99.8	99.8	99.9	99.9	99.9	99.9	99.9	99.9	99.9
2.16	74.4	83.8	88.6	92.1	94.4	96.0	97.1	97.8	98.4	98.8	99.1	99.3	99.5	99.6	99.7	99.8	99.8	99.9	99.9	99.9	99.9	99.9	99.9	99.9
2.18	74.6	83.9	88.9	92.4	94.5	96.1	97.2	97.8	98.5	98.9	99.2	99.4	99.5	99.7	99.7	99.8	99.9	99.9	99.9	99.9	99.9	99.9	99.9	99.9
2.20	74.8	84.1	89.1	92.6	94.8	96.2	97.3	98.0	98.5	98.9	99.3	99.5	99.6	99.7	99.8	99.8	99.9	99.9	99.9	99.9	99.9	99.9	99.9	99.9
Table used with single systems to get an idea of profitability for a period. It uses profit factor and number of trades during a period to estimate the percentage of time the period will be profitable.
For example,  system has a profit factor of 1.50 and it trades 50 times per-month then expect this system by itself to be profitable about 89% of the time. If you wanted the system to be profitable 95% of the time the table indicates that you'd have to increase the trading frequency to about 90 trades per-month.

acrary   Registered: Apr 2002 Posts: 700

07-01-04 05:38 AM Quote from mind: i also tried out the edge test. i used shapre ratio of my system and compared it with randomly takekn daily returns from the makret itself. systems i liked where on top of the list and made the effort useless. peace Doesn't look anything like what the edge test was designed to do. I'm including a summary for those that don't want to review some of my old posts. To do the edge test you use a single method at a time. First you backtest on the data you're using to develop the method. Then, when you're satisfied with the overall results you separate the trades by long and short by year. It'll look something like this: 1996 Long +3.00 hold 1 day Short -2.00 hold 2 days Long -1.00 hold 2 days Short +4.00 hold 1 day etc. Then you process the year of 1996 and pull out random individual trades with the same length of hold (being careful to avoid reuse of any one day). Ex. Long -2.00 hold 1 day Short +1.00 hold 2 days Long -2.00 hold 2 days Short +1.00 hold 1 day etc. When you get done you total up results of the long and short trades for the random pass for the entire period. ex. Long -4.00 Short +2.00 You do this random pass thousands of times (Monte Carlo) and rank each each of the passes so that you have a distribution from 1% - 99% for both longs and shorts for each year of the tests. Ex. Long 1% -16.00 etc. Long 99% +21.00 Then you compare the total you have for your tested trades versus the distribution to rank where your trades are as compared to the random trades. (do this for both longs and shorts for each year). If both longs and shorts rank 70% or better (20%+ better than random) then you might be looking at a edge. Do the same test with out of sample data and shorten the time period to 3 months (so that you can view multiple forward time periods). If the numbers continue to be 70% or better on both longs and shorts then you probably are trading with a edge. You do this test every 3 months after you start trading it to make sure the edge is not deteriorating. If it drops below 70% then stop trading it. (Note:  stop trading the system below 60%, wait to reuse it above 70%)  My experience with it has been very good and worth looking at during the development of a trading method phase.

acrary   Registered: Apr 2002 Posts: 700

09-21-05 12:35 PM Thank you Magna for re-opening this thread. Here's a table that I've used with single systems to get an idea of profitability for a period. It uses profit factor and number of trades during a period to estimate the percentage of time the period will be profitable. For example, if I have a system with a profit factor of 1.50 and it trades 50 times per-month then I'd expect this system by itself to be profitable about 89% of the time. If I wanted the system to be profitable 95% of the time I can see I'd have to increase the trading frequency to about 90 trades per-month. I printed it out in wordpad in landscape mode for reference. Attachment: baseline.txt This has been downloaded 859 time(s).

Chriz   Registered: Jul 2005 Posts: 88	09-21-05 01:53 PM Acrary, thank you for providing the table. Im interested in the math behind. Can you explain the calculation or point me to a page where i can find further information?
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

09-21-05 02:13 PM It's just a Monte Carlo sim using 100,000 passes per table entry and plugging in the profit factor as a multiplier for the winning trades. Sum the winners and losers until you've reached the number of trades in the pass. Then see if the pass is a winner or loser. The percentage in the table is the number of winners in relation to the 100,000 passes.

acrary   Registered: Apr 2002 Posts: 700

09-26-05 05:57 AM Quote from bulat: Would you consider this to be a tradeable pattern, or would you avoid such patterns that work as a combination of many identifiers that fail when tested separately? -bulat In your example you're showing 1-2 daytrades per-year. No matter how good they are I think that would be a waste of time. I wasn't planning on posting about edges in this thread however I'll give you the basic process for mining them since you're so persistent. First find out what is going on in the market you want to trade in the timeframe you plan on holding a position. If you want to daytrade with one trade per-day then find out all the different ways the day has played out in the past. ex. trend day, two-way day, reversal day, etc. Once you've done this you should have an idea of which type of day is most common and which is most profitable. Then define something which could be of value to trade one of the market types. An example might be in a reversal day to find out how often the market makes a low of the day in the first 15 min. of the session. If it happens often enough to be of interest then you go on to the next step. Take every period for which the target is found and create a table of outputs with 1 for the target and 0 for non-targets. Then pre-process all the inputs into the target and convert them to binary inputs. (A common mistake is to take open, high,low, and close data -- analog and assume you can find relationships with the target). For ex. yesterday close > day before yesterday close. If found mark the input as a 1 if not present mark it as a 0. Do this for as many identifies as you can. This may present a hundred or more binary inputs leading to the target for each day of the data. Then you'd pass the data into a backprop neural net and have it train on the data. (you'll need to set aside some data for out of sample testing). Once it's trained to hit at least 90% correctly test the NN on the out-of-sample data. If you hit at least 85% correctly then you can do one of two things. If you're a discretionary trader, setup the NN and preprocess the inputs every day and use the net to predict whether tomorrow has the target (in this example the low of the day is within 15 min. of the start of the session). If so use it to trade to the upside as long the net remains 85% correct. If you're a systems trader then go back to the net and look at the weights of the net to see which of the binary inputs were most important in hitting the target. Use the inputs to create a backtestable system based on the patterns. A system might be when xyz pattern exists then buy next bar above the lowest bar as long as the time is within the first 15 min. of the day. Set the stop to one tick below the low. If the system tests profitable enough to be of interest then move on to the next step. Next, take the trades and test them against random trades pulled from the same year (the edge test). Rank the trades versus random for each year of the backtest. If the trades score consistently above the 70th percentile then you can guess you've found a edge-based system. If not, then you have to assume you've found a temporal characteristic in the data that can be exploited for some period of time. If it's edge based then all you need to do is adjust the trades for market volatility and apply a money management strategy. Check the trades on a periodic basis to ensure the edge continues and plan what to do with your next million. If it's not edge based you can still trade it but you need to setup a objective bailout method such as running a monte carlo sim and determining the bailout point to be say the 95% level of the predicted max drawdown point. Your trading would be more defensive using a non-edge based method as well. Maybe you'd split the trade size in half and have a 15 min. or 10% of daily range as a filter to adding the second position (letting the position prove itself) as long as the volatility was large enough to justify the scaled entry.

acrary   Registered: Apr 2002 Posts: 700	10-13-05 01:16 PM Ok, I've given up on the whole word processing "professional looking" project. I don't have the desire to become a expert on the Microsoft products. I also noticed on the work I've completed that I kept going through notes and pulling out more and more related material. I think I could probably create a 3 book series based on the material. I also realize most people won't go further than one or two systems and then either bank some coin or give up and realize it's too hard. So, for part 2 of this journal I decided to work towards a goal. The goal is to replicate the performance of Monroe Trout in the New Market Wizards book. In the beginning of the interview Jack Schwager described Trout's 5+ year performance numbers. A 67% annual return, 87% of all months profitable, and a max drawdown of just over 8%. This should be fun (at least for me). I'll do this with 5 models or less and try to explain some stuff as I go along.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 01:24 PM From the first part of this journal I posted some about weighting different systems using the modified sharpe ratio. Here's a run using the first model. I'm using only monthly performance numbers to come up with the weights. Obviously if only used one model the weights would be unimportant. The modified sharpe ratio posted is the monthly modified sharpe ratio. Attachment: 1modcor.txt This has been downloaded 469 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 01:58 PM The weighting between the models is then passed into a money management program. To make the process easier to understand I'm using a fixed per-cent risk model for the selected portfolio. Here is a run using just the one model and projecting the performance out 12 months. For the first pass I'm just using 1% per-trade risk. I'm doing 100,000 passes in the Monte-Carlo run. Looking at the numbers you can see we project a return of 36.8%, a average drawdown of 10.9%, and 69.5% of the months were expected to be profitable. None of the criteria were met using one model so obviously more than one will be needed. Near the top of page one is the minimum funding to trade the model in the portfolio and here's how it's calculated. How much do I need to fund a account? M = Min. Required Capital A = Average loss P = Per-cent risk DD = Drawdown to recover at 95% of Monte Carlo Sims in decimal ex. .2 = 20% M = (A/P) * (1/(1-DD)) A = average loss of 1,000 P = 1% risk per-trade DD = 19% drawdown at 95% level M = (A/P) * (1/(1-DD)) M = (1000/.01) * (1/(1-.19)) M = 100,000 * (1/.81) M = 100,000 * 1.236 M = 123,600 I use the 95% level as a cutoff because I'd stop trading a system at that point. The projected per-cent months profitable is from the sheet I previously posted showing the effect of number of trades with profit factor to figure out the expected percentage of profitable months. On page 2 are the Monte-Carlo runs. From this you can see there is about a 94% chance of this system being profitable at the end of 12 months. Since the first model didn't get close to our goals I'll add a second model and you'll see how that works. Attachment: 1modmmg.txt This has been downloaded 346 time(s).

acrary   Registered: Apr 2002 Posts: 700

10-13-05 02:44 PM We're back to the weighting program this time using two models. You can see the modified sharpe ratio was improved by adding a second model (better than either of the single models by themselves). In the two model results there are a couple of things I'll go over. The first is the weighting. The position in the weighting is reflective of the position of the model under model #. In this case the best weighting to achieve the optimal modified sharpe ratio is 3 units of model 1 for every 1 unit of model 2. I call them units because they aren't contracts (they are volatility normalized results). Notice in the two model result there is a 12 month rolling correlation. To do this I use a minimum of 48 months of results. The rolling correlation starts after the first 12 months and then keeps figuring them for each month until the end of the data. The total of all the rolling months is then divided by the number of months tested. The number in the report is the average correlation. I also figure the standard deviation of the correlation for each two model pair and add/subtract it from the average to figure out how stable the correlation is. In this case the two models are negatively correlated with a pretty low standard deviation. It would take a more than 3 standard deviation month for them to be highly correlated. This would be a very good candidate for trading with the first model. Under the money management contract multiplier notice everything is related to 1.0. I did this to improve the processing in the money management software. You'll see these numbers plugged in there. Attachment: 2modcor.txt This has been downloaded 323 time(s).

acrary   Registered: Apr 2002 Posts: 700	10-13-05 02:52 PM If you want to create a program here's how I calculated the correlation between two systems. Attachment: corr.txt This has been downloaded 486 time(s).

acrary   Registered: Apr 2002 Posts: 700

10-13-05 03:21 PM Quote from makosgu: The correlation calculation here is straight forward. With correlation tho, I presume that the objective would be to have n systems with 0 correlation. The pairwise correlation (N^2) would be desired to be zero... However, given the sample size of returns (several months), is it introducing any sample errors? What I mean is that if I were to assess the correlation, should I be taking a random sample from the population of the systems returns and then evaluating the correlation or does it all just collapse to this simple case calculation... Very interesting stuff... For sure! I have considerable quant background but have kept much of this stuff separated from trading (well actually all aspects separate from trading). It is a knowledge and application weakness... MAK! For this run I believe there were 37 samples that were averaged. I chose 48 months minimum because you need 12 before you can do a rolling average. Then at least 30 more to get some sort of base close to a normal distribution. Of course there is sampling error introduced with such small numbers but it's the best I can do. I could add more periods but it doesn't change the correlations much. What I was trying to do is assess the overall correlation (average) and the distribution (std. deviation). If you have models that trade on a daily basis it would probably work better to do the test on daily results instead of monthly. These models are pretty selective so their trade frequency isn't anything special. In a practical sense what I'm interested in is the average correlation + 1 std. dev. being below .5. I have a program around here somewhere where I processed trades at different correlations and found that below +.5, it's better to trade separately. Above +.5 it's better to save the better model and either combine the second model with it or discard it if it can't be integrated.

acrary   Registered: Apr 2002 Posts: 700

10-13-05 03:45 PM Here's the money management sim for the two combined models. I ran the report once and it hit the profit objective so I cut back the % risk per-trade to 3/4%. Notice the contract multiplier is the same as on the weighting report. With the addition of the second system the expected return moved up to 49.6%, the expected drawdown dropped to 9.8%, and the profitable months moved up to 74.9%. The combination of two uncorrelated systems has helped. One thing to note about Monte-Carlo tests is they tend to offer pessimistic results as compared to the real-world trades. This happens because when two models are structurally non-correlated the chances both are in drawdown at the same time are virtually nil. Monte-Carlo assumes both are just as likely to be in drawdown at the same time as any other outcome. Since the objectives aren't met, it's on to using 3 models. Attachment: 2modmmg.txt This has been downloaded 352 time(s).

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:12 PM Here's the 3 model weighting report. The third model I added is the same as the first one except it is run on the NQ market instead of ES. It also is at a higher timeframe so it trades less often. I added it to show that by changing markets and timeframe but using the same methodology you can sometimes get good results. Notice the reduced modified sharpe ratio on model 3. This is what happens when the trading frequency declines. Also notice that the overall correlation between 1 and 3 is slightly correlated. Also notice the standard dev on the correlation moves the correlation +1 standard dev. up to +.32. While it's ok, if I was looking to trade in realtime, I'd look to see if I have another model with better correlation stats. In the big picture the 3 model test shows the modified sharpe ratio is moving up, so I expect the number of winning months to go up as well. Attachment: 3modcor.txt This has been downloaded 293 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:46 PM With the 3 models the projected profits moved up past the profit target using .75% risk per-trade but the max dd was above the goal. I reran the mmgt test dropping the fixed % to risk per-trade to .6%. The result was all three of the performance goals came closer. The projected average return moved up to 58.9%, the average drawdown dropped to 9.1%, and the per-cent of profitable months moved up to 79.4%. Notice how every time I hit the performance goal I reduce the risk per-trade. This is done to lower the expected drawdown. In testing I've found if you use fixed % risk per-trade the best way to lower the drawdown is to lower the average losing trade. At this level the only tool is the % risked. I'll need a fourth model to see if we can meet the goals. I'm going to get something to eat and be back. Attachment: 3modmmg.txt This has been downloaded 232 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:34 PM Now we're up to 4 models. Here's the weighting report. I love to look at these reports when they start getting up in numbers of systems because it tells you so much about how your portfolio is being managed. Once again the test reveals the best combination of models is using all 4 models. I like to look at the 12 month rolling correlation with +1 std. dev. in the two model results and also the two model modified sharpe ratio's to get ideas on what I need to work on to improve my performance. Attachment: 4modcor.txt This has been downloaded 248 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:50 PM Once again, here's the 4 model mmgt report. I once again lowered the % risked on each trade. This time down to .5%. The projected returns stayed about the same with lower drawdowns and more months projected to be profitable. The projected return was 58.1%, with expected average drawdown to be 7.7%, and expected per-cent months profitable to be 83.6%. Attachment: 4modmmg.txt This has been downloaded 226 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 07:05 PM I realize not everyone will have the talent or desire to develop software to do these types of tests. Because I do want to help individual traders, I'm willing to have these tests run on your stuff on up to 5 models (I have a helper that's pretty bored right now). To do this I'd need the following for each model. 1). A monthly total profit/loss for at least 48 months. 2). A trade list with date followed by a comma and the total closed profit. I'll setup a email account somewhere where anyone that wants this done can send the files as attachments. If I get too many requests, I'll let you know. I'll post the files from model131 as examples. Here's the monthly file. Attachment: model131.txt This has been downloaded 310 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 07:07 PM And here is the individual trades for model131. Attachment: model131.txt This has been downloaded 294 time(s).

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:12 PM Here's the 3 model weighting report. The third model I added is the same as the first one except it is run on the NQ market instead of ES. It also is at a higher timeframe so it trades less often. I added it to show that by changing markets and timeframe but using the same methodology you can sometimes get good results. Notice the reduced modified sharpe ratio on model 3. This is what happens when the trading frequency declines. Also notice that the overall correlation between 1 and 3 is slightly correlated. Also notice the standard dev on the correlation moves the correlation +1 standard dev. up to +.32. While it's ok, if I was looking to trade in realtime, I'd look to see if I have another model with better correlation stats. In the big picture the 3 model test shows the modified sharpe ratio is moving up, so I expect the number of winning months to go up as well. Attachment: 3modcor.txt This has been downloaded 293 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:46 PM With the 3 models the projected profits moved up past the profit target using .75% risk per-trade but the max dd was above the goal. I reran the mmgt test dropping the fixed % to risk per-trade to .6%. The result was all three of the performance goals came closer. The projected average return moved up to 58.9%, the average drawdown dropped to 9.1%, and the per-cent of profitable months moved up to 79.4%. Notice how every time I hit the performance goal I reduce the risk per-trade. This is done to lower the expected drawdown. In testing I've found if you use fixed % risk per-trade the best way to lower the drawdown is to lower the average losing trade. At this level the only tool is the % risked. I'll need a fourth model to see if we can meet the goals. I'm going to get something to eat and be back. Attachment: 3modmmg.txt This has been downloaded 232 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:34 PM Now we're up to 4 models. Here's the weighting report. I love to look at these reports when they start getting up in numbers of systems because it tells you so much about how your portfolio is being managed. Once again the test reveals the best combination of models is using all 4 models. I like to look at the 12 month rolling correlation with +1 std. dev. in the two model results and also the two model modified sharpe ratio's to get ideas on what I need to work on to improve my performance. Attachment: 4modcor.txt This has been downloaded 248 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:50 PM Once again, here's the 4 model mmgt report. I once again lowered the % risked on each trade. This time down to .5%. The projected returns stayed about the same with lower drawdowns and more months projected to be profitable. The projected return was 58.1%, with expected average drawdown to be 7.7%, and expected per-cent months profitable to be 83.6%. Attachment: 4modmmg.txt This has been downloaded 226 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 07:05 PM I realize not everyone will have the talent or desire to develop software to do these types of tests. Because I do want to help individual traders, I'm willing to have these tests run on your stuff on up to 5 models (I have a helper that's pretty bored right now). To do this I'd need the following for each model. 1). A monthly total profit/loss for at least 48 months. 2). A trade list with date followed by a comma and the total closed profit. I'll setup a email account somewhere where anyone that wants this done can send the files as attachments. If I get too many requests, I'll let you know. I'll post the files from model131 as examples. Here's the monthly file. Attachment: model131.txt This has been downloaded 310 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 07:07 PM And here is the individual trades for model131. Attachment: model131.txt This has been downloaded 294 time(s).

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:12 PM Here's the 3 model weighting report. The third model I added is the same as the first one except it is run on the NQ market instead of ES. It also is at a higher timeframe so it trades less often. I added it to show that by changing markets and timeframe but using the same methodology you can sometimes get good results. Notice the reduced modified sharpe ratio on model 3. This is what happens when the trading frequency declines. Also notice that the overall correlation between 1 and 3 is slightly correlated. Also notice the standard dev on the correlation moves the correlation +1 standard dev. up to +.32. While it's ok, if I was looking to trade in realtime, I'd look to see if I have another model with better correlation stats. In the big picture the 3 model test shows the modified sharpe ratio is moving up, so I expect the number of winning months to go up as well. Attachment: 3modcor.txt This has been downloaded 293 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:46 PM With the 3 models the projected profits moved up past the profit target using .75% risk per-trade but the max dd was above the goal. I reran the mmgt test dropping the fixed % to risk per-trade to .6%. The result was all three of the performance goals came closer. The projected average return moved up to 58.9%, the average drawdown dropped to 9.1%, and the per-cent of profitable months moved up to 79.4%. Notice how every time I hit the performance goal I reduce the risk per-trade. This is done to lower the expected drawdown. In testing I've found if you use fixed % risk per-trade the best way to lower the drawdown is to lower the average losing trade. At this level the only tool is the % risked. I'll need a fourth model to see if we can meet the goals. I'm going to get something to eat and be back. Attachment: 3modmmg.txt This has been downloaded 232 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:34 PM Now we're up to 4 models. Here's the weighting report. I love to look at these reports when they start getting up in numbers of systems because it tells you so much about how your portfolio is being managed. Once again the test reveals the best combination of models is using all 4 models. I like to look at the 12 month rolling correlation with +1 std. dev. in the two model results and also the two model modified sharpe ratio's to get ideas on what I need to work on to improve my performance. Attachment: 4modcor.txt This has been downloaded 248 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:50 PM Once again, here's the 4 model mmgt report. I once again lowered the % risked on each trade. This time down to .5%. The projected returns stayed about the same with lower drawdowns and more months projected to be profitable. The projected return was 58.1%, with expected average drawdown to be 7.7%, and expected per-cent months profitable to be 83.6%. Attachment: 4modmmg.txt This has been downloaded 226 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 07:05 PM I realize not everyone will have the talent or desire to develop software to do these types of tests. Because I do want to help individual traders, I'm willing to have these tests run on your stuff on up to 5 models (I have a helper that's pretty bored right now). To do this I'd need the following for each model. 1). A monthly total profit/loss for at least 48 months. 2). A trade list with date followed by a comma and the total closed profit. I'll setup a email account somewhere where anyone that wants this done can send the files as attachments. If I get too many requests, I'll let you know. I'll post the files from model131 as examples. Here's the monthly file. Attachment: model131.txt This has been downloaded 310 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 07:07 PM And here is the individual trades for model131. Attachment: model131.txt This has been downloaded 294 time(s).

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:12 PM Here's the 3 model weighting report. The third model I added is the same as the first one except it is run on the NQ market instead of ES. It also is at a higher timeframe so it trades less often. I added it to show that by changing markets and timeframe but using the same methodology you can sometimes get good results. Notice the reduced modified sharpe ratio on model 3. This is what happens when the trading frequency declines. Also notice that the overall correlation between 1 and 3 is slightly correlated. Also notice the standard dev on the correlation moves the correlation +1 standard dev. up to +.32. While it's ok, if I was looking to trade in realtime, I'd look to see if I have another model with better correlation stats. In the big picture the 3 model test shows the modified sharpe ratio is moving up, so I expect the number of winning months to go up as well. Attachment: 3modcor.txt This has been downloaded 293 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 04:46 PM With the 3 models the projected profits moved up past the profit target using .75% risk per-trade but the max dd was above the goal. I reran the mmgt test dropping the fixed % to risk per-trade to .6%. The result was all three of the performance goals came closer. The projected average return moved up to 58.9%, the average drawdown dropped to 9.1%, and the per-cent of profitable months moved up to 79.4%. Notice how every time I hit the performance goal I reduce the risk per-trade. This is done to lower the expected drawdown. In testing I've found if you use fixed % risk per-trade the best way to lower the drawdown is to lower the average losing trade. At this level the only tool is the % risked. I'll need a fourth model to see if we can meet the goals. I'm going to get something to eat and be back. Attachment: 3modmmg.txt This has been downloaded 232 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:34 PM Now we're up to 4 models. Here's the weighting report. I love to look at these reports when they start getting up in numbers of systems because it tells you so much about how your portfolio is being managed. Once again the test reveals the best combination of models is using all 4 models. I like to look at the 12 month rolling correlation with +1 std. dev. in the two model results and also the two model modified sharpe ratio's to get ideas on what I need to work on to improve my performance. Attachment: 4modcor.txt This has been downloaded 248 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 06:50 PM Once again, here's the 4 model mmgt report. I once again lowered the % risked on each trade. This time down to .5%. The projected returns stayed about the same with lower drawdowns and more months projected to be profitable. The projected return was 58.1%, with expected average drawdown to be 7.7%, and expected per-cent months profitable to be 83.6%. Attachment: 4modmmg.txt This has been downloaded 226 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 07:05 PM I realize not everyone will have the talent or desire to develop software to do these types of tests. Because I do want to help individual traders, I'm willing to have these tests run on your stuff on up to 5 models (I have a helper that's pretty bored right now). To do this I'd need the following for each model. 1). A monthly total profit/loss for at least 48 months. 2). A trade list with date followed by a comma and the total closed profit. I'll setup a email account somewhere where anyone that wants this done can send the files as attachments. If I get too many requests, I'll let you know. I'll post the files from model131 as examples. Here's the monthly file. Attachment: model131.txt This has been downloaded 310 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 07:07 PM And here is the individual trades for model131. Attachment: model131.txt This has been downloaded 294 time(s).

acrary   Registered: Apr 2002 Posts: 700	10-13-05 07:42 PM Glad that worked! Ok, with 4 average models I didn't achieve the goal. So of course I've got to add a fifth model. Just to make sure it achieves the goals I'm using one of my models that I trade. It's going to change the weights dramatically, but it should be interesting. It's probably going to take about 20 min. for the weighting program to do it's crunching.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 07:57 PM Anybody that's wants their stuff tested, here's the email address. [e-mail] (lol) I put in dashes to prevent autobots from putting me on autospam (I've seen others do it so I'm guessing here). Remove the dashes to send stuff to me.

acrary   Registered: Apr 2002 Posts: 700	10-13-05 08:22 PM Here's the weight test for 5 models. Notice the modified sharpe ratio for model 5. Also notice how it boosts the overall sharpe ratio for the five systems dramatically. Now also notice how none of the 5 systems has a strong correlation. Even when taking into consideration the standard deviation none of them go above +.4. Notice on the weighting for the money management how model 5 becomes the basis for all others to weight against. This is what happens when you bring in a great system. Attachment: 5modcor.txt This has been downloaded 315 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 08:32 PM Quote from bjs211: Acrary, First, thank you for doing this...very helpful to see simple examples done step by step. Could you confirm for me that I understand the position sizing element properly...here is my take: 1. Determine the optimal relative position weights for each model by calculating the combined maximum mod sharpe ratio. (I use solver or goal seek in excel) 2. Take the largest weighting and set it equal to 1. All the other models will have position sizes that are a fraction thereof. So for example, if the weightings for 3 different systems came out to be 5, 3, and 1 respectively-- I would use a 1 "unit" position size for system 1, .6 "unit" position size for system 2, and .2 for system 3. Assuming I was using 1% fixed percentage position sizing methodology, this would equate to 1% equity risked for each position put on for model 1, .6% equity risked for each position put on for model, and .2% for a position in model 3. Am I looking at this the right way? Thanks again, bjs Yes, that is exactly how it works using fixed fractional money mmgt.

acrary   Registered: Apr 2002 Posts: 700	10-13-05 08:48 PM Quote from bjs211: Thanks for the quick reply. Also, if you don't mind, it would be valuable to hear your recommendations on selecting a money management approach. I realize that this would be getting a bit off topic at the moment, but if you'd come back to the topic when you think it fits, I'd appreciate it. I'm sorry i had to run the mmgt test over (reduce trade size again). In general if you have a small account fixed ratio is probably better. Once it's grown you can switch to fixed % at risk. After you've done that for awhile you'll get some new ideas. I don't plan on posting about what I'm currently using (which is why I'm willing to post on older material). As you can see, to use fixed % mmgt you need a substantial account if you want small drawdowns.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-13-05 09:11 PM Well here's the mmgt projected report for the 5 models. The goals: 67% average return --- beat 72.1% 87% of all months profitable -- beat 92.6% 8% Max drawdown -- beat 7.6% at 95% probability (once in 20 years) Notice how the addition of one really good model changed everything. The average annual drawdown is only expected to be 4.8%. Remember this is probably pessimistic because of the limitations of Monte-Carlo testing. The % risked per-trade was also reduced to just .3% so as you can see you don't need to risk alot to gain alot. Of course the proof is not in the model but what would have happened in historical trading. Next I'll post the historical report that uses the weights, and % risk to see how accurate the prediction model is to what have happened. Attachment: 5modmmg.txt This has been downloaded 297 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-13-05 09:29 PM Here's the historical report for the same period with all of the weighting numbers plugged in. Risk per-trade .3% Annualized return 75.3% (slightly better than the model) % winning months 94.0% (slightly better than the model) Max DD -7.1% (slightly better than the model) Going forward the returns are dependent on the stability of the individual models. One of the nice things about using non-correlated models is even if one starts falling apart at least one of the others will most likely pick up some of the losses. Attachment: hist.txt This has been downloaded 271 time(s).

acrary   Registered: Apr 2002 Posts: 700

10-13-05 10:32 PM The last post of the night. I wanted to just show how to use the model to get a boost in returns while reducing the drawdowns. I'm tired so this is going to be a little simplified. I call this the free money ratio. If you noticed on the minimum capital required to trade the 5 systems it only amounted to about $187,000. Since the account had $500,000 all the extra money isn't really working very hard. The free money ratio is: FR = account funding / minimum capital required in this case: FR = 500,000 / 187,000 FR = 2.673 To use this you multiply the original projected drawdown @95% confidence by the FR to come up with a new target drawdown. In this case the 95% drawdown was 7.6% * 2.673 = 20.31% Using this new drawdown target we re-run the mmgt report using a larger risk amount until it's close to the new target. We also use only the min. account required for the initial capital. In this case I upped the risk per-trade to .75% and set the initial capital to $187,000. You can see from the report the 95% drawdown level is 18.2% so I could have upped the risk per-trade a little more. What we're doing is saying of the original 500,000 most of it (313,000) is not going to be actively traded. Because of this it will have no risk and no return. The rest will be actively traded at the free money ratio. The return at the higher risk is 287.2% on the $187,000. While the drawdown at the 95% level is 18.2%. We then convert the %'s into dollar returns. Return = 187,000 * 2.872 = 537,064 expected profit Risk = 187,000 * .182 = 34,034 expected max dd at 95% confidence we then use the return and risk with the total account return = 537,064 / 500,000 = 1.074 or 107.4% expected return risk = 34,034 / 500,000 = .068 or 6.8% expected drawdown @ 95% As you can see by doing this the return was boosted and the drawdown reduced. You can do this everytime the account hits a new equity high and compound the return to much higher levels. This works because returns do not grow at a linear rate. I'm tired and I have to get up in a few hours to trade the DAX so I hope someone got some ideas from all this work. Take care until the next session. Attachment: free.txt This has been downloaded 331 time(s).

acrary   Registered: Apr 2002 Posts: 700	10-18-05 02:10 PM Yes, the two subjects are mutually exclusive. I wanted to show the benefits of building systems this way and then switching to the process of finding the models to plug in to achieve the results. My style of development requires a paradigm shift away from tactical trading (buy when x crosses y, etc.) to a strategic vision. Tactics are still there but only to supplement the strategic goals. So far it doesn't look like there's much interest in this material apart from professionals so I may take a break.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700

10-18-05 02:14 PM Quote from man: alan i wonder about one thing. i have never seen a sharpe of 1.6 with such a great sterling (return p.a./ mdd). from a 1.6 equity curve i would expect a sterling of 3, maybe 4 or 5. but 14? hmm. i hope i did not get anything too wrong here through my fast cross-reading your material ... peace The modified sharpe ratio I posted is the monthly sharpe...not annualized.

steve46   Registered: Mar 2003 Posts: 3575

10-18-05 02:33 PM Alan: I always enjoy hearing from you on this subject. My approach is similar in many ways, probably because I have listened to your prior comments. I observe the following: You look to minimize risk first, then you focus is on consistency of returns. This is what professionals do. Amateurs in contrast look for historical tendency, and throw money at it hoping that risk will "take care of itself". It never does. On the system side, I have followed a rule that has helped me greatly and is probably of more interest to the retail folks. What I have done is to find repetitive market behaviors and incorporate both the behavior and "failure" into my systems. As you already know, if the repetition is frequent enough, and you have programmed correctly, you will then catch one or the other all the time. This works well IF both behaviors have significantly different results. The obvious example is a breakout after a test of a price level (like say a 50 period EMA) or a failure and move down to nearest support. personaly, if I were looking to make a dollar trading, after reading Alan's comments I would be mobilizing all my resources to try to learn to use a Monte Carlo Engine (I like @Risk by the way). Good luck guys, Steve

acrary   Registered: Apr 2002 Posts: 700	10-20-05 01:58 PM It's pretty obvious some of you are expert model builders. Please contribute! Here's a historical report that was run from one of the submissions risking only .4% per-trade. Pretty humbling to see someone on this site with this kind of talent. Attachment: history.txt This has been downloaded 700 time(s).
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-20-05 03:04 PM Quote from slacker: What software are you using for the MonteCarlo analysis? For the MonteCarlo work I've written my own routines. I don't know of any software that simulates a non-normal trading type of distribution. The fat tails can make a big difference in the overall profit/(loss). You described a very simple way to size your position. If you used this approach with your MonteCarlo testing would it appear to be too conservative and risk adverse? For the trade sizing I think it accurately reflects the type of returns expected in the "real" world. I don't think it was too risk adverse. The MonteCarlo tests in the top and bottom 5% are not realistic because they don't take into account the effects of correlation (which is a big part of my strategy). Would you change this basic position size approach if the market appeared to be more favorable to you? Have you found any way to match position size with market conditions instead of only using account size? For example, if last 4 trades were winners, increase the size of the next trade for example... Another example, if volatility of the market > x and trendstrength > y then increase size... At this level it's assumed you've already made your trade decisions. All this is doing is maximizing the return and reducing the risk based on the account size and models. If you want to vary trade size based on other information it would be passed to this process. For instance if you want to scale in with three separate trades then the three trades will all be used to determine how much risk to put on for each of those trades.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-20-05 03:17 PM Quote from MarketDog: I have a question for you in regard to your correlation and weighted models work. Does each model added trade one instrument such as ES? Or does the model trade an edge across several instruments? Each model represents one stream of trades. As I posted, this could be one model in multiple markets represented by multiple streams or multiple models in one market. What if you have a trading model that trades stocks in the Nasdaq? Would it be appropriate to keep that as one model? Or possibly break the stocks into their perspective sectors and have numerous sector models where you could use your correlation and weighted designs to add each sector model only if improvement is made to overall return and risk? You could use it either way. Instead of looking for many non-correlated models/markets you could just do pairs trading based on two non-correlated models for each security, basket of securities, or market sector. If you had 10 sectors you could then give 10% to each pair and recalculate the pool every so often or you could take the results from the 10 sector pairs and run them through the weighting test to see how much each should get from the pool and allocate funds to each after every trade.
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-20-05 03:34 PM Quote from rokafella: Dang... thanks for the scare! So what's in the next chapter? I'm planning on showing how to setup a research platform using excel. After that it'll be what most of you are waiting for..."How to build your first system"
		Edit/Delete • Quote • Complain

acrary   Registered: Apr 2002 Posts: 700	10-20-05 03:36 PM Quote from Walther: I would like to hear more about what do you mean by a strategic vision ? I do not want to jump to conclusion but do I smell some forward projecting around here ? You will when I start on building a system from scratch. No forward projecting here. It'll mostly be like "it's so obvious how come I didn't think of that?"