Home > Technical Topics > Strategy Development > Trading system results overview

Trading system results overview

  1. I would appreciate some remarks on the results of a trading system for YM i've been testing.
    The following results are for basically the same system based on some range breakouts. the first one is based on daily data and the second one is based on intraday day.
    especially i would like to discuss the difference in the "profit factor" and "payoff ratio" parameters in the different time frames - is it something you've seen in the past: how these ratios drop dramatically in intraday system as opposed to daily systems.
    Above all, do you think it's realistic (in terms of slippage and commissions), the daily system can be traded manually but the intraday must be automated naturally.

    Both systems were checked on a period of 20 months.

    1st System - Daily Signals
    netProfit: 1222 (in YM points)
    profitFactor: 1.7642
    payoffRatio: 1.3232
    numOfTrades: 42
    winPrecent: 57.1429
    avgTrade: 29.0952 (in YM points)
    avgWinner: 117.5417 (in YM points)
    avgLosser: -88.8333 (in YM points)
    avgHoldTime: 4.3844 (in days)
    avgHoldTimeWinner: 4.7672 (in days)
    avgHoldTimeLosser: 3.8739 (in days)
    max DD: 334 (in YM points)

    2nd system - intraday signals
    netProfit: 2913 (in YM points)
    profitFactor: 1.2929
    payoffRatio: 1.2649
    numOfTrades: 643
    winPrecent: 50.3888
    avgTrade: 4.5303 (in YM points)
    avgWinner: 39.6914 (in YM points)
    avgLosser: -31.3785 (in YM points)
    avgHoldTime: 196.8865 (in minutes)
    avgHoldTimeWinner: 213.0154 (in minutes)
    max DD: 702 (in YM points)
    avgHoldTimeLosser: 180.1230 (in minutes)

    do you think any other parameters should be checked?

    theoretical "Buy and Hold" for the period tested would have given 131 YM points.

    thank you all for your time
     
  2. i would look at sharpe ratio or sortino ratio of daily returns expressed in percentage terms.

    profit factor in intraday day is a different story i would think. in general i find a flawed concept. should only be an addon to other figures, not of main interest IMO.
     
  3. thanks,

    what would you consider as a suitable sharpe / sortino values for daily/intraday system?
     
  4. sharpe 2.0+, the more plus the better. but you must take into consideration as well that stock systems might suffer from different biases, like survivorship- or indexmembership-bias. fees and slippage are important as well. what kind of stocks? what liquidity and so forth.

    i am currently not trading singlestocks on intraday basis, but will hopefully have something started by the end of the year. will stick to top fifty liquid names, maybe making the net position completely market neutral by use of futures. not sure as of now. my minimum sharpe will be 2.0 based on fees of 1 cent per share and 1.5% spread for shorting. something like that. it is conservative, but not too far of i would say. and i will test on five minute data, assuming i get the close of the next bar after my signal bar. that should do fine in terms of slippage.
     
  5. additional: your time period is too short. two years is nothing for a daily system. it is even too short for intraday. i would at least do analysis on five to six years. preferably longer to cover for some pre-bear market behaviour as well.
     
  6. sorry. i do not where i had my head. you are trading YM not single stocks, forget my comment. with futures i am pleased with sharpes of 1.5+ ... on intraday basis.

    i would get some sp future data and test on real long term as well.
     
  7. System 1 is based on 42 trades, this is way too short.
    You need atleast 10 times that for testing, if not more.

    System 2 is atleast tested over 600+ trades, however
    the win loss ratio of 39 : 31, is too small, you need atleast
    46:31 (ie 1.5:1) when you have a 50% win system.

    With such a small ratio: slippage and commissions and the odd mistake wont leave you with much left over.


    My primary system has an average win:loss ratio of 1.75:1 and a win rate of 45%.
     
  8. appreciate your post,
    quick question for you about the number of trades.
    i understand what your saying about the low number of trades, although i've seen people stating that 30 trades are starting to be statistically sufficient, but how do you take into consideration the number of trades in relation to the time period?
    the systems were tested on the same time period: the 1st produced 42 trades and the 2nd 643. shouldn't there be some way to factor this?
    thanks,
     
  9. actually this is quite interesting, and i have another question for you businessman (and everybody else)
    according to what you said about he 1.5:1 ratio, if you had the choice, what would you prefer:

    system A:
    winner 50%
    and ratio of 1.25 avg win to avg loser

    system B:
    winner 35%
    and ratio of 2 avg win to avg loser

    thanks
     
  10. The more trades the the better.
    If you can get 2000 trades to test with then thats ideal.
    More data wont hurt you, it can only prove your system is
    not robust which will save you $$$$.

    Ofcourse time period is important too, however the longer
    the type of trade the more years you have to test.
    A long term trend following system would have to be tested
    over decades.
    A swing trade or once a day system, between 10 and five years.

    For intra day system atleast three years.

    You also want to look at different types of markets for YM:

    Up markets eg 1999,2003
    Down markets eg 2002, 2001
    Low volality markets eg 2005
    High volality markets eg 2002
     
  11. I dont think you can make any money with system B,
    youll just about break even.

    And even if you had a 5:1 ratio, i would find it really hard
    to trade a system that only wins 35% of the time. Youd have
    to live through 10 to 15 losers in a row quite often.
     
  12. i hear what you're saying.
    anymore thoughts, would be much appreciated
    thanks
     
  13. No offense AskQuestions but profit factor and payoff ratio are both incredibly simple metrics and if you don't understand them, I highly recommend reading up on them until you can easily answer your question on your own. You definitely want at least that level of understanding before trading any system. Not to mention that when you ask others for help, you need to be wary of getting incorrect info e.g., what you got from man... FYI, profit factor intraday measures the exact same thing that it does on all other time frames -- how many dollars/euros/whatever your system has made for each one lost. Period. It's that simple, and it is NOT flawed. However, both Sharpe and Sortino ARE flawed because they FAIL to work as intended under certain circumstances that are not apparent unless you understand exactly how each is calculated. Study them in detail and you'll see what I mean. Also take a close look at the equity curves... they'll give you way more info than any single metric or group of metrics can.
     
  14. i am not offended, and i appreciate any response, and i thank you for your time.
    i do understand what profit factor and payoff ratio mean, and they are indeed simple measures.
    what i wasn't familiar (and therefore asked for more info) was on sharpe and sortino. i still dont understand what should be considered as "Rf is an expected return of another "standard" asset" (http://en.wikipedia.org/wiki/Sharpe_ratio)
    and would appreciate any comment.
    could you elaborate on what are the "certain circumstances" in which the sharpe and sortino fail?

    last one, what exactly are you looking for when checking the equity curve?
    thanks
     
  15. Hi. Personally, I've had several systems fail on me in the past couple year so I've kinda learned like everyone else, from mistakes. I'm still learning but here's some generalities I've applied, at least to 1-3 minute timeframes. Maybe it will also give you a warmer feeling about what you've developed;

    1) Robustness - Basically, how well does your system perform on all (YM, ER2, ES, QM, and NQ).

    2) Make sure you have commissions and at least 1 tick of slippage built in your model

    3) Personally, I like Profit Factor greater than 1.5 and Sharpe > 4

    4) Equity curve must be smooth

    5) Ideally, I would like Longs and Shorts to contribute 50/50

    6) Starting with a $25K account and trading only 1 contract, Annual Return greater than 30%, no compunding.

    7) People have diiferent pain levels so try and determine if you can psychologically deal with your systems max consecutive losses and max system drawdown.

    Just some thoughts. Good Luck.

    --jstox
     
  16. Only thought I would add to the previous post speaking of number of trades during your test period is this ... be cognoscente of the number of trades in your sample vs the average win..... you may find that as the number of winners goes up your average win goes down ... way down and you run the risk of getting chopped up. If the average win goes down to a level where you will end up in the negative when slippage and commission is taken into account you wanna know that ... sounds simple but ends up being an after thought to many folks.
     
  17. appreciate your reply

    some questions:
    1. the robustness issue sound important but do you really think the same system concept should apply to all major futures (YM, ER2, ES, QM, NQ)? are you personally familiar with such concepts?

    3. about the Sharpe: i've asked this before but didn't get any answer. how do you calculate this value? are you using some software that does it for you? in the definition here (http://en.wikipedia.org/wiki/Sharpe_ratio), one should define what is the expected return on another "standard" asset, and i'm having hard time figuring what it should be.

    7. very good advice. i'll keep that in mind
    thanks everyone,
    any more???
     
  18. 1) From my experience the results should at least breakeven when trading the other futures. Maybe some minor tweaks needed but it's a good indication of an over-optimized system.

    3) I use Amibroker, automatically calculates the analysis stuff.

    7) History does repeat itself. Yep, that ones critical :)

    --jstox
     
  19. Hi,

    Just wondering where you get your historical intra-day futures data from?

    Thanks
     
  20. Thanks AskQuestions,

    Looks good.. I like that you can buy symbols separately.
     
  21. jstox,
    do you recommend Amibroker as a backtesting platform?
    pros/cons?
    much appreciated,
     
  22. For my systems, I use the "modified sharpe ratio" mentioned by Acrary in his system development thread. The formula is:

    modified sharpe = mean / variance

    or more explicitly:

    modified sharpe = average trade return / standard deviation of trade return

    Like another poster on this thread said, expect that as your backtesting trade quantity goes up, your average return, largest loss, and historical max drawdown will usually have a negative drift.

    I understand the negative drift intuitively like this: if an extreme negative event for my system has something like a 1 in 1,000 chance of happening, I probably won't see it if I'm only looking at a 500 trade sample. Even if I'm looking at a sample of 10,000 trades, odds are that there is an even worse extreme event that has a 1 in 20,000 chance of happening and I can't see its effect in the backtest because my sample is too small.

    With this in mind, you can try to factor in the negative drift based on the size of trading samples your backtest is using. A related area of statistics that you might find helpful for implementing this kind of sample size based "fudge factor" is the "confidence interval", which I'm sure you're already familiar with. Don't forget about the fat tailed distribution of returns if you choose to implement something like this though.
     
  23. thanks for sharing.
    what would you consider as a tradeable system in terms of "modified sharpe" values? and do you find any difference between EOD systems and intraday systems?
     
  24. I look at how each system contributes to the modified sharpe of a basket of systems. If it improves the overall modified sharpe of the basket, then it is probably good to trade if the other metrics aren't bad because it makes a positive contribution towards a smoother equity curve. Each system shouldn't necessarily control an equal portion of your margin though.


    I think it is a mistake to pick one time horizon and ignore the others because doing so blinds you to a lot of information and causes you to miss opportunities. Many people assume that just by trading the short time horizon that they will implicitly get the benefits of the longer term moves, but that is an incorrect assumption. So my answer is yes, I do find a difference between EOD and intraday systems. Each are making decisions based on different information, and if your systems are based on TA signals, then they are modeling fundamentally different volatility and liquidity characteristics.
     
  25. I agree. 42 trades is not enough.
    I could get a profit factor and win ratio like that flipping coins with 42 trades.

    The lowest # I would use would be about 200 trades, and only if the profit factor and win rate was very very good. 1000 trades is solid to get some good stats if applied to lots of market conditions.

    My best systems are in the 65% win, PF 2.5, 6% max DD range and thats with several thousand trades, over years of data, tested against down markets, up markets and sideways markets.


    System 2 - not good enough to trade. You need more "buffer" in case you missed something.



     
  26. I think sharpe ratios, etc, are over rated.

    Just post a picture of the equity curve (by time) along with the max DD #.

    This will visually tell me more than any sharpe ratio or any other number can.

    This is all about understanding the "behavior" of your system.
     
  27. When you are running through 10,000+ different variations and system combinations, you can't look at every equity and drawdown chart, so what metric do you use for optimizing? I optimize for the modified sharpe and eyeball the more interesting equity and drawdown charts (chiefly the optimal one). I'd be interested to hear how other people do it.
     
  28. thank you dragon for sharing

    are you referring to futures intraday systems?

    i'm attaching the curve of the intraday system. any comment are welcome.
     
  29. Hmmm .... good question. Amibroker is awesome. It does everything I need plus a helluva lot more than I'll ever use. With the interface built-in for the IB datafeed, my monthly total operational cost for trading is $0.

    Do I recommend? Can't really say. This is very powerful and flexible software. The learning curve can be steep and a few people have given up.

    In my case, the invested effort was worth it.

    --jstox
     
  30. I agree with jstox, I have used Amibroker for four years and its better than anything else that I tried. Its a snap to write up a trading idea and test it out. Works great for autotrading at IB. And the price is low plus free data from IB.
     
  31. thanks,
    how about historical database (especially intraday futures), i understand you need to provide it yourself, right?
     

  32. i second every comma of that post. perfect. pro thinking.
     

  33. sorry mister for giving out "incorrect info", but profit factor in my humble opinion is flawed since it does not at all take smoothness of the equity curve into account. not at all. and this flaw becomes more apparent the more often you trade.

    a system that makes two trades every day. one looses 1%, one makes 1.1%. that results in a profit factor of 1.1 with a hit ratio of 50% and a payoff ratio of 1.1; no big figures for 28% annualised return with no down day. sharpe ratio would be infinite since there is literally no vola in a system that makes 10 basispoints each and every day.

    high sharpe and sortino figures rarely tell a wrong story if the number of trades is sufficient, let's say above 200. sharpes above 2.0 rarely show a strange equity curve.

    profit factor is for single traders. sharpe is for funds and i dare to say multisystem pros. but again, that is just my humble opinion.
     

  34. do not forget to annualise both the return and the standard deviation. (you annualise them in different ways, if you are not aware of that.)

    my threshold for intraday systems is a mod Sharpe of 1.5 for a single market. 2.0 for a portfolio of the same system on several markets.

    nevertheless sometimes a system with a sharpe of 1.0 might be so uncorrelated or even negatively correlated to the rest that it boosts portfolios sharpe more than another system with a higher sharpe. note that additional sharpe becomes more and more challenging the higher the level already is. to move from 1.0 to 2.0 is way easier than from 2.0 to 3.0 or even from 2.0 to 4.0, which would be the actual equivalent percentage wise.
     
  35. thanks
     
  36. I do that often.

    My summary results are loaded into a spreadsheet where I sort by total profit, %win, average profit per trade, max DD, etc,.

    I find the "juicy" area of the parameter sets. Then I reference those equity curves. I often do the reverse though, and start by eyeballing the EQ charts.

    I have a tool that allows me to flip through EQ charts in a web browser super fast, so I can spot the really good looking ones. Next to each EQ chart, is the set of params that created it so I can cross reference.


     
  37. I only tade NAZ/NYSE stocks. currently.


    Can you post an EQ chart that has TIME has the X axis? So we can see the frequency of the trades day by day?

    The problem with this type of EQ chart is, that all those trades could have occured in a single month out of a 5 year time period.


     
  38. Sorry mister man but profit factor doesn't try to quantify the smoothness of an equity curve. It simply measures how many dollars/euros/whatever a system has made for each one lost. And it does that just fine. So how is it flawed if it doesn't do something it wasn't designed for? That's like saying a car is flawed because it can't fly. LOL!! P.S. Sharpe and Sortino, on the other hand, are flawed because they don't always perform as intended. Just google around and you'll find plenty of papers on it.
     
  39. the only confusion on sharpe ratio happens in the tradestation reports. everywhere else it seems to be quite clear.

    SR = (annualised return - risk free rate)/annualised standard deviation of returns

    mod SR = annualised return / annualised standard deviation of returns

    "return" can be daily, weekly or monthly changes. i would recommend using daily, since they show the picture clearer with shorter time frames. nevertheless when you calc sharpe on ten years of data, your sharpe derived out of daily or monthly data will show very similar results. which is not the case with just one year if data, since you cut out some vola with the monthly observations.

    for the following i use the excel terminology. i hope i don't make a typo ... :)

    you annualise daily returns by "=POWER(1+avgReturn; 256) -1", monthly returns by "=POWER(1+avgReturn; 12) -1".
    that assumes that you calculate your returns as a percentage, like: (NAV(t0) - NAV(t-1))/NAV(t-1).

    if you calculate your returns as log-returns, like "LN(NAV(to)/NAV(t-1))" then you have to annualise by "avg return x 256" for daily data and "avg return x 12" for monthly.

    annualising standard deviation is done by "STABW(hereYouHaveToFillInTheColumnWhereYourReturnsAre)*SQRT(256)" for daily data and the same with "...SQRT(12)" for monthly.


    now, since sharpe ratio only looks at daily results, it does not care if you achieve them with high hit ratio and low payoff or vice versa. and think about your personal utility function. sharpe is probably much more similar to it than any other figure. which is the reason why i use it.
     

  40. judging system by means of profit factor IS flawed. the higher the frequency of trading the more.

    i know a couple of papers that put out the flaws of sharpe ratio. most of them are either completely academic or at least fail giving something better.

    it is quite interesting that you accused me of giving wrong information in tghe first place, while you yourself do not really bring up an argument but reference to google. now please put me on your ignore list, you enter mine in a second.
     
  41. You're kidding, right? If your understanding of something as simple as PF is so flawed, how can you ever hope to understand Sharpe? It is WELL KNOWN that Sharpe can mislead... even Sharpe himself said: "Sharpe ratios can give a false sense of precision and lead people to make predictions unwisely." So please do remain uninformed so I can take the other side of your trades. And keep me on iggy... at least until you get both of your neurons talking to each other.
     
  42. About the only problem I've seen with PF is that it doesn't take into consideration the distribution of trades. If a large chunk of the profits come from the top 10% of trades then the overall results from day-to-day trading will be overly optimistic.
     
  43. Would you say that an AM radio has a problem because it doesn't receive FM? PF doesn't try to boil an equity curve down to a single number... it's a very simple measure of profit/loss and that's about it. Useful only as a screen in conjunction with other metrics and certainly not a substitute for looking at an equity curve.

    P.S. What you wrote about PF not taking "into consideration the distribution of trades" is also true for Sharpe in a way, because Sharpe penalizes large positive returns.
     
  44. That's exactly how I use Sharpe and PF, first things I look at when researching. If they don't meet my numbers then I know I'm wasting time going down the checklist.

    When I'm done with my checklist and have identified the Holy Grail, then I have to go through my code to find the programming bug. :mad:

    --jstox
     
  45. Don't disagree with anything you've said. I was just trying to understand where man was coming from. I'm always up for learning something new. Perhaps he'll provide a little detail on his problem with pf if someone expressed more interest in shedding light rather than heat on the subject.

    As you pointed out the Sorentino ratio is probably a better measure than sharpe due to the performance penalty.

    Pardo in his book "Design, Testing, and Optimization of Trading Systems" had a interesting idea for measuring the robustness of a system by using a corr. coeff. between the equity curve and what he called "perfect trader". I've never tried it but it sounds interesting.
     

  46. i have the impression that with increased number of trades profit factor at least sometimes is diminishing while all other figures are great. i tested a smooth equity curve with a sharpe of 2, but a profit factor of merely 1.3. on the other hand i have a system with a profit factor of almost 3 with a sharpe of less than 1. if you look at the equity curves you quickly see that your visual impression is literally always in line with sharpe but really not always with PF.

    and as i said before. i think as a trader you are finally neither hurt by low hit ratio nor by low payoff but what you really feel is sharpe. give it a second thought.

    whenever you enter any pro-shop, what people are really talking about is sharpe. my problem with sortino is that you tend to overweight the importance of a few positive outliers. i saw a presentation of rentec recently: they publish sharpe and return, nothing else.

    anyways, there are other ratios that are advanced, i think omega is one of them. that would be a very different discussion. but between sharpe, hit ratio, payoff, profit factor, draw down, MAR ratio, sterling ratio there is absolutely no doubt where the true value lies.

    it is just not the ways traders think. look at tradestation reports. all these dollar figures. that is single trader thinking. dollars in an account and not % of capital. that is the main source of difficulty. traders usually think in dollars and not in %, therefore every figure that moves away from dollars is suspsicious. i highly respect that, no offense intended, but when you are talking to a real pro, who is running multiple systems or strategies, you realise they know their sharpe very well. just consider you have one system, than you add another one and another one. how do you measure the overall effect? how much better have you become? consider now you have two additional systems but you can, for whatever reason, only trade one of them. which one do you choose? sharpe (or if you will sortino, omega, what have you) answers this question elegantly. PF, hit ratio, payoff ...?

    having said that i respect that it is not by any means necessary to even ever have heard of sharpe to be a wonderful trader.

    BTW my problem with all draw down figures is that a single figure out of a whole distribution is statistically weak ... it is interesting. plus, i think that with the overall shape of the return distribution being unchanged, max dd MUST increase over time. so whenever you panick after five years when your system makes a new historic dd you might be in fact observing something inevitable ...
     
  47. I found recovery factor to be a very useful figure besides Sharpe. It takes implicitly into account not only all the individual trades and their variability, but also the time sequence in which they occurred because max drawdown is considered in it. The same set of trades could have produced a very different equity curve if they had occurred in a different order.

    If you look at an EQ curve with a very high RF you can see immediately that it's highly correlated to its linearity.

    I found RF to be the best stand alone figure to express the quality of a curve. Alternatively, I look at avg % x trade, % winning trades and payoff ratio considered together. But I like RF best.

    To the thread starter: I looked at the EQ curve you posted and it is too jagged to me, I personally wouldn't trade it. And the other system has too few trades, as it has already been said.

    GS
     

  48. valuable post.
     
  49. appreciate your opinion.
    thats an interesting point about RF.
    what would you consider as an attractive RF value for your systems (please share with us if its daily/intraday stocks/futures etc)
    thank you
     
  50. maybe the line of reasoning below will help set straight the PF bashers/sharpe praisers.

    profit factor is flawed.
    sharpe ratio is not flawed.
    warren buffet's track record has a high profit factor.
    warren buffet's track record has a terrible sharpe ratio.
    therefore, warren buffet's strategy must be flawed.

    lol!

     
  51. I currently trade equities but that's actually not very important. RF is simply final net equity / max drawdown, both in $ or %. So you choose your time horizon and your max DD level, relative to what you have. Personally, I like to see RF above 10 as a bare minimum on at least 10 years of daily data. This implies a 10% max DD.

    GS
     

  52. buffet has a modified sharpe of 1.54, with annual return of 22,43%. over thirty years with a multi-billion account you won't here too many comments by any pro that this is anywhere near to "terrible". one of the best, if not the best, quant shop out there, rentec, has a sharpe of 2.6 on their flagship medallion fund. very naive to think you match them when you are trading the same sharpe on a 50k account.

    so, i am afraid, your post is as humorous as ungrounded.

    other hint. if you had a backtest on intraday data like that:

    1999 0.5%
    2000 6.5%
    2001 -6.2%
    2002 10.0%
    2003 21.0%
    2004 10.5%
    2005 6.4%

    would you love to trade it?

    these are the last seven years of berkshire hathaway. now, if you think that i think buffet doesn't do a great job, please return to the beginning of this post. after the third loop you are allowed to read on ...
     
  53. hey man cut your losses. the more you try to show how smart you are the more you show how dumb you are. cut you losses. you are obviously not a pro as you think/claim you are.
    i can correct you on this post and on several of your previous posts, but by doing so i will only be doing one thing. i will teach you something. bye.


     
  54. which you prove by the fact that you could correct me on several occasions but you don't because you don't want to teach me something?

    ya. very impressive line of argumentation. indeed.
    how dumb am i, how smart are you. and how easily
    you proved that.
     
  55. Hi GS,

    Do you really find AFL language a snap??
    Do you have a programmng background?

    I agree that its very easy to code up simpler backtests in Ami,but I personally found the looping code a bit daunting..

    Any suggestions appreciated
     
  56. It is better to keep your mouth shut and appear stupid than to open it and remove all doubt.
    --Mark Twain


     
  57. Taowave
    I know some programming but not a lot. I just learned AFL by looking at examples on Amibroker web site. I mainly like the versatility and speed of Amybroker so it was worth my while to study AFL, glad I spent the time to do so.
     
  58. This is exactly why people need to stay away from metrics they don't understand. The examples "man" has given are perfectly plausible and shouldn't surprise anyone. Unless, of course, they're under the mistaken impression that Sharpe and PF measure the same thing.

    Profit factor is a very simple, "bottom line" metric that is calculated by: total wins/total losses. It (as a metric) is NOT "diminished" with an "increased number of trades" AT ALL and will only decrease with more trades if the system's profitability decreases with more trades. It doesn't take variability of returns into account... just the bottom line.

    Sharpe is a measure of reward vs. variability and the denominator of the formula can penalize large gains as well as losses. So a profitable system with a month of exceptionally high performance would be penalized by Sharpe but still have a high PF, while another system steadily returning a tiny amount over the risk-free rate would have a high Sharpe but a low PF (and bottom line).
     
  59. I fail to see why PF and Sharpe need to be used on an either/or basis. Black and white thinking, to quote spike500.

    Given their differences, what's wrong with casting an eye at both at different stages? PF for individual systems in development, Sharpe for the basket of systems in production.

    Neither metric conquers all. Funds describe their returns via several different lenses.

    AM
     
  60. agree with your spirit. disagree with you on the current issue. the point is that whenever you start real intensive research you need to focus on one main output figure. in my case i have one strategy that consists of several single strategies and we choose at any point fo time automatically which one to trade in which parameterisation. now in order to do so we need a single figure by which we rank. you might have some additional figures like minimum of trades, but these are merely secondary variables. now using sharpe or figures alike for that purpose serves quite well. profit factor fails. period.

    does that make pf wrong or meaningless as a secondary thing to look at? surely not. each and every additional aspect is great, welcome, "good" if you will. but this is not the discussion we are having here.

    i have the feeling that many people with not too much experience are not aware of these circumstances and follow these sometimes a little weird and bashing discussions, left with the believe "finally they all are right ... somehow". but this is not the case i am afraid. some simply know more than others. (and while i personally claim that i know more about the current subject than some others within this thread i am completely aware about my minor position regarding others on the board. over the years i think people will realise that for example the kind participation acrary showed will never happen again. a pro opening up like that with no other interest for getting something back than a feeling of community ... once in a lifetime.)

    why am i so insistant on this whole thing and not shut up? the initial problem of each journey is the target. if your target is vague ... the outcome will be. same in trading. what is your utility function? which figure is it that you want to maximise? i dare to say that almost every trader will finally maximise their sharpe. they might use five other different figures and may not even know sharpe as such, but they maximise their sharpe. why? well return can be easily changed by leverage. draw down? same thing? profit factor (i skip that ... reread the thread :)) ...
    sharpe does not change with leverage (actually it does a little due to some weakness of the concept, yet i find it bearable and the advantages offset the dis by far.).

    usually what single system traders finally use very often is the judgement by looking at the equity curve. is this a strong concept, actually stronger than sharpe as a single figure? no doubt. our internal pattern recognition engine is telling way more than any set of figures. but, as someone pointed out earlier, when it comes to many many tests, you need some figures to use as filter before ... or, to be mor precise, o n e figure ...
     
  61. man, can you please give the source of your information on Buffet's Sharpe ratio? if you calculated it yourself, can you please show the calculations? did you just use data from 1999-2005?

    The first link below is from a Wall Street Journal article written after 2005, and it says Buffet's sharpe ratio is 0.15.
    The second link is a research paper that says Buffet's sharpe ratio is 0.786 using data from 1980 to 2000.

    http://www.pfblog.com/digest/5063_buffetts_batting_average_is_sinking.shtml
    http://66.102.7.104/search?q=cache:...ffet+"sharpe+ratio"&hl=en&gl=hk&ct=clnk&cd=90

     
  62. did it myself. all years since 1966 out of the report on buffet's site. it is just annual data, but at 40 observations that is bearable.

    1965 23,80%
    1966 20,30%
    1967 11,00%
    1968 19,00%
    1969 16,20%
    1970 12,00%
    1971 16,40%
    1972 21,70%
    1973 4,70%
    1974 5,50%
    1975 21,90%
    1976 59,30%
    1977 31,90%
    1978 24,00%
    1979 35,70%
    1980 19,30%
    1981 31,40%
    1982 40,00%
    1983 32,30%
    1984 13,60%
    1985 48,20%
    1986 26,10%
    1987 19,50%
    1988 20,10%
    1989 44,40%
    1990 7,40%
    1991 39,60%
    1992 20,30%
    1993 14,30%
    1994 13,90%
    1995 43,10%
    1996 31,80%
    1997 34,10%
    1998 48,30%
    1999 0,50%
    2000 6,50%
    2001 -6,20%
    2002 10,00%
    2003 21,00%
    2004 10,50%
    2005 6,40%
     
  63. omg man you are using accounting book values to calculate the sharpe ratio.

    http://www.berkshirehathaway.com/letters/2005ltr.pdf

     
  64. did i get something wrong? i mean it is basically a stock, so what would you use?

    well, that was obviously a more stupid question ...
     
  65. first column is the stock price, second is the book value.

    return 23,62% 20,18%
    stdv 29% 16%
    mSharpe 0,82 1,25

    stock book value
    1988 59,32% 20,10%
    1989 84,57% 44,40%
    1990 -23,05% 7,40%
    1991 35,58% 36,90%
    1992 29,83% 20,30%
    1993 38,94% 14,30%
    1994 24,96% 13,90%
    1995 57,35% 43,10%
    1996 6,23% 31,80%
    1997 34,90% 34,10%
    1998 52,17% 48,30%
    1999 -19,86% 0,50%
    2000 26,56% 6,50%
    2001 0,48% -6,20%
    2002 -3,77% 10,00%
    2003 15,81% 21,00%
    2004 4,33% 10,50%
    2005 0,82% 6,40%

    sorry. my fault.

    i do not have longer stock returns back. but probably overall the sharpe will be more near 1 than 1.5 since 1966.

    overall i would keep up my argument up, though it is obviously not as strong as before.
     
  66. haha, so as it turned out, it was your post that was "humorous as well as ungrounded" :D

    nobody point out his errors anymore, please. let him bask in his own arrogance and stupidity.

     
  67. i am afraid my mistake will not cover yours.
    and i am afraid that your comment on buffet
    which you essentially used to raise questions
    on the usefulness of shapre ratio still stands
    as ... humorous and ungrounded.

    unfortunately that has not changed.
     
  68. and timmyz, please put me on ignore. the two of us are wasting each others time plus the attention of others. thnx.
     
  69. This discussion was never about the best single metric to use... and the debate about PF began when you ignorantly proclaimed it flawed. You continued to demonstrate your witlessness by claiming that PF as a metric somehow degrades as trade frequency increases, by being clueless of how PF and Sharpe can diverge, and by relentlessly digging yourself a deeper hole here. Are you as resolute about not admitting you're wrong when a trade's not working out?