backtesting my acutual trades

Discussion in 'Automated Trading' started by mrmoose, Jun 12, 2009.

  1. mrmoose


    I an a discretionary trader, and I want to eventually automated my exits. I first want to back test various exit strategies against my actual trades. Does anyone know of a system where I can import my trades at the end of day to be tested against real time tick data?
  2. I just want to clear this out before I comment...

    Are you asking for a software that can reverse-engineer a trade log and "fit" it into a trading model?
  3. mrmoose


    Yeah in other words I want to take what i actually did for a month and see if I could beat my results by using an automated exit.
  4. 1. A software does not exist.

    2. This goes back to the old school Tradestation 4.0 - 2000i days, which is like 10 years ago, where they would check your entry/exit and analyze them with the tops/bottoms of the bar. For example...

    Let's say your long exiting price (sell) was at $95. The particular bar you are testing was a daily bar, and the high of the bar you exitted was $100. Then the software would assume that the best possible exiting price was $100 and calculate the "efficiency" of the exit.

    In another words. You entered at 90 and exitted at 95. You had $5 profit. But the high of the bar was 100... so your hypothetical max would be $10. (100 -90), hence the efficiency of your trade would by 50%.

    This measure is very trivial, because getting the top/bottom of the bar is extremely "hard" to do. And also, the usage of a bar high/low is trivial.

    3. I have never heard of a for-sale software that can do what I asked. If I felt it significant to run a test like what I suspect, I would have to code it on my own.
  5. I do not know of commercial off-the-shelf software that does this... the closest would be like MAE indicators in existing backtesting software, but this software is not setup so much for grey boxes.

    however... if you have logs of your orders, it's not too difficult to do this. From trades is harder because as GannGalt mentions you would sort of half to try to re-find your trade and assume you could get the same fill... it's better to work off the orders.

    also you would need to have already recorded the tickdata or purchased it for same time periods.

    Then you would just load up your order file into your strategy, and parse the orders for each tick.... anytime the symbol/date/time matched, you would send that order... and then pair this entry with all the variations of different exits you want to backtest.