system expectancy question

Discussion in 'Trading' started by mdszj, Sep 15, 2018.

  1. mdszj


    Hey all

    I was trading a mutual fund system a couple years ago for which I calculated an average per trade (expectancy) of about 0.95%. This was based on a set of 465 trades. My win rate was 52% and loss rate was 48%. Average win was 3.74% and average loss was 2.04%. Over time the average per trade has ranged up and down, generally between about 0.8% and about 2% per trade. Average trade length was 11 days.

    I stopped trading it during a big downturn in the market but I am now thinking of starting to trade it again. Also, it seems like it is a pretty low expectancy figure. Was wondering how other trading systems measure up against this system as far as stats, especially average per trade.

    Any info appreciated.
  2. bpr


    0.95 expectancy is very good. It should not have big draw down.
    The average win and avg loss might be off. any change in there there will be a big change in expectancy.
    You need to divide the data in to many parts and calculate expectancy and other metrics independently and then compare with whole and see the correlation.
    if good correlation then expectancy is meaningful else it cannot be relied on.
  3. mdszj



    thx for replying - I have broken the data set into sections already and have found that when I get a bunch of losses and the system expectancy goes down, say when NYSI goes negative and the market corrects. The system expectancy goes back up when NYSI goes positive and the market rallies. That is where I get the variation in expectancy of different groups of trades that I mentioned above.

    I was typically in about 20-30 open positions at any time, so if the market took a dive, usually they were almost all (if not all) hit. Opposite applied if market rallied. This often resulted in "streaks" of a lot of wins in a row or losses in a row.

    Is that what you were referring to?
  4. bpr


    hey I just noticed your calculation is wrong
    expectancy is 0.47 and not 0.95 for this data
    52% win 48% loss
    avg win 3.74%
    avg loss 2.04%
    Now this makes sense and yes you will have draw downs.
    Regardless every system will have draw downs and streak of loosers.
    Money management is the only way to fight it off.

    here is a excel that I have created for Expectancy calculation.
    May be it is useful for you. Calculation.xlsx?dl=0
  5. Handle123


    Really difficult to come up with a truer assessment cause you been trading in a market after 2009 ? And this time period not having excessive drawdown, I have to say your stats are curve fitted even if you tried to have honest approach to testing. What can be expected as with any trend following method, you might have the largest position on till the top. What risk management do you plan on for a 30% or ??? slide? or bear market, have they made mutual funds where you can short them? Have you done studies on topping formations and hedging open profits? You have decent ave wins to ave loss, but market in strong bullish direction. Have you tested before 2009? Much larger sample size of 3,000 be more prudent and covering 20 years if using daily data, you want to be able to test when times are choppy or reversing, am guessing your win ave percentages would decline to 35-40? by extending more data. I hope I am wrong for your sake but it is what I have discovered with my studies of charting and testing.
    Good luck.
  6. pipeguy


    Measure its returns by the opportunity cost of the capital. What instruments do you think imply similar risk level? A grade corporate bonds or may be some blue-chips stocks. I would suggest that you can compare it with high quality company bonds
  7. mdszj



    thx for responding.

    bpr - Thx for the dropbox spreadsheet. I have all my trades documented in an excel spreadsheet and that is how I did the calcs. Just to make sure I am using the right expectancy equation, I used:

    Expectancy = (Avg winner x win rate) - (avg loser x loss rate)
    = (3.73 x 0.52) - (2.04 x 0.48)

    where avg winner = 3.73%, win rate = 52%, avg loser = 2.04%, and loss rate = 48%.

    I just recalc'd it and still come up with an expectancy of 0.96% per trade. Am I doing this right? Not sure how you came up with 0.47%, can you let me know?

    Handle123 - you are correct, my first trade was 6/29/17 and last one was 2/15/18, a total of 231 calendar days. I did not do much backtesting - pretty much all forward tests using real $$. I always used the same position size for each position. I started with $200 per position initially just to get some data points then as I gained confidence in the method gradually went to $1000 each position (in an acct of about 50K), which is about 2% acct risk per position assuming 100 % loss. I could not set any stops since they are mutual funds. I just scanned the results for all the trades and there are no losses greater than 10%. I noticed several trades with gains greater than 10%. There are no commissions to impact frequent buys/sells.

    This strategy is nothing more than buying a daily breakout from a descending trendline and selling the position when it breaks down from the (hopefully) ascending TL. ALmost all of the mutual funds are long only funds so I usually dont short or use inverse funds.

    pipeguy - since the avg trade length is 11 days, I feel like there is some opportunity cost, seems like a long time to be in a position for the return it provides.
  8. bpr



    the avg win and avg loss always need to be reward to risk ratio where risk needs to be made as 1 and then you calculate
    so in your case 3.73 : 2.04 = 1.83 : 1
    now using these 1.83*0.52 -1 *0.48 =0.47
  9. mdszj


    ok thx, good to know.

    So I guess the systemis not that great, guess I will keep looking for a better one.
  10. bpr


    currently I am trading a system with expectancy range 0.35 to 0.5 so 0.47 is not bad but you have to deal with drawdown sometime.
    Of course if you can come of with better expectancy then grt. Me too looking to improve my number but not easy.
    #10     Sep 16, 2018