Optimization Paradox: Avg Trade vs. Number of Trades

Discussion in 'Strategy Development' started by cunparis, Nov 27, 2008.

  1. Let's say you have a strategy which is positive and just for the sake of argument this strategy has (and I'm making these up so they may not be 100% correct but you'll get the idea):

    Option 1
    Win Rate: 60%
    Reward/Risk: 1.25
    Profit: $25k
    Trades: 300
    Avg Trade = $83

    Now let's say you see a way to optimize this strategy. You add in an extra condition and you get:

    Option 2
    Win Rate: 70%
    Reward/Risk: 1.5
    Profit: $12k
    Trades: 100
    Avg Trade: $120

    So we can clearly see two things:

    1 - Option 2 performs better in terms of risk and equity curve

    2 - Option 1 will make you more money

    So here's the paradox:

    If you take the second strategy, you could trade a bigger size. Let's say you double your size. Now you have:

    Option 3:
    Win Rate: 70%
    Reward/Risk: 2.0
    Profit: $24k
    Trades: 100
    Avg Trade: $240

    Now here's the delima: Option 3 is safer than Option 1 and earns the same amount of money, but if you look at return on a percentage basis, it performs less because you have to use twice your money. Now if your money is sitting idle it seems that this is the way to go. But if you could allocate your capital to other trades, then it seems that maybe Option 1 would be better.

    However, Option 3 has less trades than Option 1.. which means that you are in cash more often, and that cash is available for other trades. For example if Option 1 has 1 trade every 1 day, and Option 3 has 1 trade every 3 days, you would have (assuming it's a day trade) your money free for 2 of the 3 days and available for other strategies.

    So which option would you trade?
  2. How do you measure reward/risk?
  3. The Risk/Reward increases to 2.0 after doubling the size???

    Or wait... maybe he has a $Value stop or something... Of course, that'll make it a whole different system profile and be completely irrelevant to the examples the OP provided...

    The biggest paradox is the OP's stupidity....
  4. Ok, I made a mistake. The R/R ratio shouldn't change between Option 2 & 3:

    Option 3:
    Win Rate: 70%
    Reward/Risk: 1.5
    Profit: $24k
    Trades: 100
    Avg Trade: $240

    R/R ratio is simply the sum of the wins / sum of the losses.

    I know it's inevitable that some may feel the need to show they are better than everyone else, it's quite common here on ET. But I'm hoping we can keep this thread on topic without insults and make it interesting for everyone.
  5. cunparis,

    What is it you hope to achieve with these comparisons.

    Is it purely educational or are you hoping to apply it and to which instrument are you wishing to apply it to.

  6. You might also consider the magnitude of the draw downs (losing streaks).
  7. I ran across this situation while backtesting a strategy I'm working on and I'm currently debating which option to trade.

    For example I can set the criteria really tight but it gives 2 trades a month. I can set it really lose and it gives a trade a day. The challenge is to find the optimum.
  8. Hi,

    In my educated opinion I would say #1, all else being equal. Going from #1 --> #2 is a double whammy, as you are adding a parameter (losing a degree of freedom) AND drastically reducing the number of trades.

    How many total parameters do you have?

  9. You bring up good points. I have 5 variables. For example, one is the Bollinger Band period and another is the BB width.

    Actually Option 2 didn't really add a criteria, I think I mispoke. What I did was replace a criteria with another one that is more specific.

    What I find is that I can always add more criteria that will improve results but reduce the number of trades and therefore reduce the overall profit. For example, let's say for the setup. I could say to only enter if 20 day MA is going up. That might make the win rate & R/R ratio better but might give me less money. I probably should have just used that as my example in the beginning it would have made the question simpler.

    I realize that each variable (if I add the MA above then the 20 becomes a variable) will contribute to overoptimization and reduce the chances that it will work in the future.

    But it seems more logical to me to add the criteria and have lower risk, and then to put effort into a 2nd strategy and trade them both. I'm thinking two strategies like Option 2 are better than 1 like Option 1.

    Thanks for your input.
  10. This is a very good point. I observed today that one variation had 3x higher drawdown than another. Reducing drawdown is very interesting.
    #10     Nov 27, 2008