My first back test

Discussion in 'Strategy Development' started by wdbaker, Aug 15, 2002.

  1. wdbaker


    # of Trades 83.00 Ratio + 51 61%
    Cost $415.00 Ratio - 32 39%
    Max $14.50
    Min $(2.25)
    Points 121.31 $/Trade $73.08
    Total Rev $6,065.47 Pts/Trade 1.461558072

    # of Trades 82.00 Ratio + 50 63%
    Cost $410.00 Ratio - 30 40%
    Max $16.75
    Min $(2.75)
    Points 115.06 $/Trade $70.16
    Total Rev $5,752.97 Pts/Trade 1.403162317

    Trades the S&P e-mini on 5minute charts, this is the June contract
    What am I missing, this was done on an excel spreadsheet similar results on previous 5 contracts.

    Look forward to your thoughts, crititisism, slams etc...
    Is this thing ready to trade???

    Also the reason for the odd figures is that I used .00001 in place of a blank or zero on the spreadsheet so that I could tell the diff between a trade that came up zero or a blank no trade.

    Please don't laugh to hard :D
  2. maybe you are not missing anything.. run it live and then you will know for sure about fills, slippage, etc.

    edit: you did remember to count the losing trades right? :)
  3. wdbaker


    You funny Ha Ha

    By the way, you were looking for data before, found free five min data at anfutures, here is the link, both es mini and nq mini

    Just need to get the money to trade it, even if you add slippage it looks pretty good on a 1 contract basis.

  4. if it works, it shouldnt be hard to get money to trade it.. one contract is $2k margin ($3k to be safe). heck.. if it works, i'll code it, fund it, run it, and split the profit with you.
  5. Your commish is $5 each way so subtract $10.

    Each trade is worth $70 minus $10 >>> $60.

    Then there is slippage. Slippage is not just what you get after you pull the trigger. It also includes the changes that occur from the moment your signal is given until you pull the trigger. If you are a robot, or you have automated execution, that part is negated. I suspect that neither are true. All you need is 3-4 losers in a row to make you hesitate pulling the trigger on the 5th trade, resulting in even more slippage.

    That is not to say that all slippage works against you. Some slippage will result in price improvement. But it "seems" to happen so rarely that it is not "counted" on.

    Backtesting is great to at least give you an idea of the optimal potential of the method.

    Suggestion: program slippage. One tick = $12.50. If your method says long at a price, try adding 1-2 ticks to that number as your fill. For exits from a long, subtract 1-2 ticks. For shorts, obviously, subtract 2 ticks from your fill and add 2 ticks to the cover/exit price.

    Optimally, the method produces about $60 per trade after commish, before slippage. If you lose just one tick to each side of the trade the profit is $35. If you lose 2 ticks per side, the profit is $10 per trade.

    The only way to really know is trade it live. Your mission will be to trade as close to optimal as you can.

    Because the bottom line is....
  6. Aaron


    If you used the June contract data you show the results for in developing your strategy, then you have to be concerned about curve fitting. If so, you should now take your strategy and test it on out of sample data and see if it still performs as well.