Backtesting - Execution Concerns

Discussion in 'Strategy Development' started by AllenTrader, May 4, 2004.

  1. I've been using a "trainer" trading platform and I have serious concern over the validity of my trades actually occuring in the real world. I've been paper trading anywhere from 5 to 20 ES contracts and usually take 1%-5% daily whereupon I "hang up my hat" for the day and start over to seek further validation of the strategy figuring with real money I'd be out at anything over 2% (pigs get slaughtered). I've spoke with the firm I'm using and they say that once 2/3 of the available volume is consumed my paper trades are executed. I'm new to futures trading and have been in the securities business 8 plus years. I stepped up and traded 100 and 250 ES lots figuring the increased volume on my part would help illuminate potential execution problems and create a more realistic execution. Daily returns fell somewhat, moreso in faster markets. How are fills determined with the various backtesting strategies out there? Are they just based on price, available volume, percentage of volume? Does anyone not affiliated with a provider have the answer?

    Allen
     
  2. ...you're testing with limit orders, I assume? IMO the most conservative thing to do is to BT with market orders and use your historical slippage. If in actual trading you do better with limit orders, fine.
     
  3. and you will a have a more conservative view of potential results.

    Limit orders may not be filled even for small size even when placed well in advance.

    I have worked with a firm that automates and trades a large number of systems and have come to the following conclusion:

    You are competing against automated market making systems who's sole purpose is to facilitate better fills for various other more directional systems by placing a band of orders around current price range to ensure better placement in the price time cue. Orders are placed, cancelled and replaced en mass as the price oscillates, until a fill is needed by one of the systems whereby it my have a better chance of a fill due to price time priority.

    In short:
    I have experienced the very real disappointment of poor fills on a "high %" systems ( many small gains/ occasional larger loss) that traded very profitably and consistently in tests...

    Try your approach with one lot over the course of a month or two.

    Best of luck.
     
  4. ...a trick question designed to enrage certain parties here: would you therefore conclude that depth of market is useless?
     
  5. nitro

    nitro

    You mean limit orders that are away from the market, i.e., _limit_buys _not_ at the ask, and _limit_sells_not_at_the_bid.

    nitro
     
  6. The trainer is in real time and is apparently identical to the real deal save for executions. (FFastTrade...no plug) I use stop limits for entry and regular "get the hell out of Dodge" stops for exits.
     
  7. Since slippage was not part of the question I assumed we were discussing limit orders away from the market.

    Don't market orders get filled with the exception of limit moves?

    Here is my reference for order descriptions:
    CME http://www.cme.com/edu/getstr/clslkord643.html

    "Limit Orders
    A limit order is one that can be executed only at a specified price or better.

    Market Orders
    A market order is to be filled at the best available price immediately upon receipt by the broker."

    ___________________________________
    Going forward:

    If a simulated buy stop limit is triggered and filled at the current ask I would feel more confident since the order would essentially be marketable, at least while liquidity remained available at the ask. The reverse would be true for sell orders.

    If the stop limit buy were triggered by trades at the bid and your limit were to join the bid I would not count a fill unless a lower price occurred while your order was live.

    I hope I make sense...

    If you deduct/ add a tick from either side of your round trip order in testing you'll see weather or not the approach is robust however you'll have to trade the same way in actual trading.

    Testing with limit orders + slippage does not solve the problem of unfilled profitable orders. The losses will almost always be filled. The price could reach your target with out you being in position. Price could then retrace and fill your limit order and subsequently keep going and hit your stop...


    I experienced similar scenarios. In testing 5 profitable trades occurred over a 1 hour period while in real trading 1 profitable trade occurred followed by a loss, all due to the order in which orders filled and the subsequent price fluctuations..


    I recommend the following: make the test more difficult ( add slippage or use market orders), if it still looks good, rigorously trade the system/ approach with small size and compare real results with simulated results. If the test corresponds closely to the actual results proceed with caution.


    Best regards,
    Alex
     
  8. I have mixed feelings about the value of market depth as a stand alone indicator primarily due to to hidden size...

    I do place value in time and sales and the impact of large trades on remaining posted liquidity at price areas around those I'm observing or intending to trade near.

    Sorry if that's ambiguous.
     
  9. ...thanks for your reply. I share your opinion. However Jack has posted that evaporation of depth outside the inside bid is significant. Just curious what you thought.
     
  10. Osman

    Osman

    as a rule of thumb, if the stock you're trading has high volume (in the millions) the less you have to worry about slippage. the lower the volume, the more you need to worry.
     
    #10     May 5, 2004