Friday, May 3, 2013

TF System

Last month was the first month in memory that I made no trades.  I have been building and testing strategies in NinjaTrader and that work continues.  I just haven't found anything I'm satisfied with to actually start trading.  Hopefully that happens soon.  It's really really hard to find something that nets more than a few ticks of profit per trade per contract with minimal drawdowns!

As mentioned a month ago, I decided to focus more on TF since that is the best bang for my commission dollar.  That said, I thought I'd share some data for one strategy that I recently finished coding but am still hoping to improve and make better. If you have any thoughts about whether you'd trade this, let me know.

Initially just backtesting the past year, results looked pretty good.  But then when I tested 3 years of data, the drawdown became unbearable.

Curious what system optimized for min. drawdown would look like for past 5 years yielded:
There are only 2 variables in this system which determine settings for profit and stop on each trade and any open trades are exited at close each day. I haven't done anything yet with walkforward testing but that may be something to look at to improve results.

Or perhaps the next idea I test will be much better?  Who knows...

10 comments:

  1. What is the problem with the system? That 5 year track record is excellent for a purely mechanical system, isn't it?

    ReplyDelete
    Replies
    1. I don't know. Consider that no commissions or slippage are included in the results. So that would diminish the avg. $46/trade for the 5-year optimization down to, perhaps, $25-35 per trade? And could you trade this knowing that in past 5 years there were 2 long drawdown periods, one just under a year and the other over a year?

      Delete
    2. Yeah I guess I didnt look at the time scale, didn't realize those drawdowns were so low. And you do use still factor in slippage when you use limit entries?

      Delete
    3. No I don't figure any slippage on limit orders. I'm not sure how much to include for market orders? Do you (like if you enter long at close of a bar, does Ninja take into account you would buy at ask price of next bar's opening print)? But stop orders you have to assume at least 1/2 to 1 tick per trade per side I think. Really, I'm finding that I always look at the per trade Dollar avg first whenever I run a backtest. If that isn't at least a tick, I consider it a losing system instantly.

      Delete
  2. PF is too low. I don't trade anything less than 1.5/1.6. Average trade is kind of ok. What slippage & commissions did you use?

    ReplyDelete
    Replies
    1. just read you didn't use any commissions and slippage. Not tradeable precisely because of prolonged drawdowns as you mentioned. Potentially you can use this with other systems (that you will develop in future).

      When I started developing strategies, there were some ideas that were like this and I kept them on backburner. 6-9 months later, I would have a thought and I will re-visit the idea and modify it a bit - my PF will go up sometimes from 1.3 to 1.8. So, I say lets keep it in your bag and move on. Maybe you come back to it later.

      Delete
    2. Good points thanks anon. Agree higher PF would be nicer - but to come up with a system that yields higher PF, now that is the hard part...

      Delete
  3. Since these results are in-sample, optimized results, they cannot be relied upon to judge this strategy, other than to say the going forward performance will almost certainly be worse.

    ReplyDelete
    Replies
    1. True. I should "incubate" like you do Kevin!

      Delete
  4. Keep up the good work MBA,
    My 2cents advice, u should consider putting volatility and liquidity in play. When volatility is high, there are chance u won't get fill if ur formula using market order or have high chance of getting big slippage when liquidity is low.

    It's really in real market everything is not as good as backtesting or demo account.But I didn't have solution for that.

    Wish you best of luck

    ReplyDelete