Need advice with position sizing

Discussion in 'Risk Management' started by hlpsg, Jul 23, 2008.

  1. hlpsg


    If you had a system that has a winning percentage of around 83%, but it loses twice as much on average as it wins, what would be a your plan on position sizing for such a strategy?

    During backtests, I've not yet seen two consequetive losses in a row, but it could always happen in real trading.

    Any advice pls?
  2. danielc1


    How many times do you have a trade?
    How much do you want to make per year/month/day?
    What is your max. drawdown tollerance?

    A fixed fractional strategie for position sizing is hard to beat. I would go with that.
  3. hlpsg


    Thanks for the response. It trades about 12 times a year.

    Target profit of 30 - 50% annually, max drawdown perhaps 20%.

    The max drawdown could probably be reduced futher if I diversify the instruments, but I need to first fix my position sizing problem.
  4. (0.83)(1) - (0.17)(2) = 0.49..... ~0.50. If I calculated that properly, your optimal bet percentage is 49% or 50%. You could start out at 20% or 25% and then increase it up to 50% as you get more comfortable.
  5. hlpsg


    Are you using Kelly's formula to calculate that? I got around the same figure when calculating with Kelly's formula.

    And that's 50% of whatever my current capital is (after taking a win or loss), correct?

  6. Yes and yes. Start smaller and then work your way up. The consecutive-loser will eventually catch up with you. Resist the urge to bet more than 50% after it happens. You could get hit with a triple-consecutive loser.
  7. danielc1


    If you drawdown tolerance is max 20% a year, you can never, ever put a bet size of 20% or more per trade. If you have two losers in a row, you are down 40%... How would you psychology react to that?

    I would try to smooth out the equity curve(and emotions) with a bet size between 3% and max 6% of your capital. With 6% pushing it...
  8. hlpsg


    Thanks, I appreciate looking at this from all angles, so I welcome all input.

    My drawdown on total capital (assuming I bet 50% of capital) is roughly 15% after one loss.

    The chances of me having 2 losses in a row are 0.17 * 0.17 = 3%. That's quite a high chance.

    Should I suffer 2 losses in a row, I'd be down close to 30% of my initial capital. So that's more than I was initially prepared to tolerate, so I might have to rethink this.

    The chances of 3 losses in a row are 0.5%. In real life, I'd say it's probably higher than that, let's be conservative and double it and say 1%. That's still quite significant.

    Should I suffer 3 losses in a row, I'd be down 38.5% of my initial account.

    Should I bet 1/2 Kelly, i.e. 25% of my account, these are the rough numbers:

    Probability of one loss = 17%
    Loss on total capital after 1 loss = 7%

    Probability of two consequetive losses = 3%
    Loss on total capital after 2 losses = 14%

    Probability of three consequetive losses = 0.5%
    Loss on total capital after 3 losses = 20%

    Thanks everyone for helping me think through this. Really appreciate it.
  9. danielc1


    You risk is then 7% of your capital per trade, right?

    This is acceptable if your system gives you the results you think, but high.

    What you also could do is making another positionsizing calculation for your winnings in the market. If you for example see your capital as a base and your winnings as markets money, you could risk 14% on markets money(50% bet size in capital of that money) and risk only 7% on your own capital.
  10. ronblack


    I think you will find this free article by Michael Harris very informative:

    In addition, Michael Harris in his new book Profitability and Systematic Trading has a section about position sizing, explains the fixed fractional method and offers some very good tips and insight into optimal methods.

    As far as consecutive losers note that these may not come in a row, as Michael Harris explains in the book, so the probabilitiy of you ruining the account varies with respect to that sequence.

    #10     Jul 24, 2008