Fill Engine

Discussion in 'Strategy Development' started by maninmoon, Mar 9, 2011.

  1. Does anyone have experience of creating a fill-engine for simulating limit orders - and then has compared that to live trading results?
     
  2. Sure. what is the question ?

    Tom
     
  3. Hi Tom,

    Have you found a close correspondance between your simulated fills and what you get in real life?

    Moon
     
  4. Yes, sure. Still not completely clear the question though. :)

    By "simulated", do you mean a "virtual execution" (like in backtest) or in "paper trading" ?

    Also what do you mean exactly by "correspondence" ?

    Anyway, the general answer (in my case) is no, i have not found significant differences. What matters is the strategy, small randomic slippages (in favor or against you) statistically mean really nothing.

    Tom
     
  5. Hi Tom,

    Thanks for quick response. By simulated I mean backtest. BUT in particular here I am talking about backtest of a strategy that uses LimitOrders not MarketOrders.

    MarketOrder backtests I understand pretty well. But simulating when you might get a fill on a LimitOrder is a different animal.

    I have been reading your old thread about your CT system - very interesting.

    Moon
     
  6. Thanks a lot Moon.

    I have also a more recent one, that i am compiling now:
    http://www.elitetrader.com/vb/showthread.php?s=&threadid=208265&perpage=6&pagenumber=12

    About orders, personally i use <b>only lmt orders</b> (i give in my bot also the possibility to override this setting, but usually the feature has little use).

    With lmt orders you can have precise computations. The fill price of a mkt order is at market discretion.

    In real trading, if an order is not immediately filled or is partially filled, your automated system will probably repeat it (depending on strategy). It's not something which is going to have much importance.

    What matters is the overall strategy. Small slippages that should ever occur in entries/exits can be sometimes slightly favourable, sometimes slightly unfavourable, but that can be considered a sort of random, accidental perturbation, usually negligible compared to the importance of the overall game and hedging mechanisms.

    Tom
     
  7. Tom,

    Excuse me - I should have been more specific. I am thinking about resting "limit" orders, anf quite High Frequency.

    Moon
     
  8. Well, i think you may want to enforce a "fill timeout", and cancel the unfilled ones (i do the same).

    I don't believe much, personally, in the myth of very high frequency. It's much better to <b>overlay multiple medium frequency</b> games. You can attain a much higher frequency with an incomparably safer approach and risk diversification. :)

    Tom
     
  9. Ah I see.

    What would you call "medium frequency" in TPD?

    And are you sytems intra-day i.e. Do you flatten position each night?

    Moon
     
  10. maninmoon and fullautotrading,

    Thanks for an interesting thread.

    What rules do each of you use in your "fill engines" to specify whether a limit order fills or not?
     
    #10     Mar 15, 2011