Estimating slippage in SPY backtest

Discussion in 'Strategy Development' started by logic_man, Oct 9, 2010.

  1. My strategy uses conditional orders to set off market orders in the SPY, so how far, on average, should I assume the SPY market order gets filled from the price I used to trigger it? .02? .01?
  2. For your backtesting, are you using EOD data or tick data?

    Most backtesting is done with EOD data. Slippage is lost in the noise when evaluating a strategy using EOD data.

    So what is backtesting for? I limit it to qualifying a strategy for prototyping using paper trading. Even paper trading has its limitations. The platforms that I have any experience with will "execute" your paper trade if the price hits your specifications. However, no consideration is given to the quantity specification of your order compared to the available quantity in the marketplace.

    After qualifying with backtesting and prototyping with paper trading your next step is live trading. Only then will you know what the market has in store for you.

    Now if you are engaging in high frequency trading backed by supercomputers to operate your algorithms, well that's another story all together.
  3. It would be tick data. I had heard there were backtesting programs which could estimate the amount of slippage for a given instrument, e.g. the SPY. Common sense would tell me that no retail trader can move the SPY price and I know from observing it over the years that it rarely moves more than .01 in the couple of seconds following my order triggering, by which time my order should be executed, but I don't want to be overly optimistic in looking at the backtest results, either.

    Agree that the actual market-based results are all the matter in the end.

    The supercomputers will have to wait, unfortunately.
  4. You may not move the Market as an individual. But take your system and have a big fund trade it, or have a bunch of people trading the same signal, you'll get slippage thats not directly related to you.
  5. Yes, which is why I have attempted to make my algorithm completely unique, which, I am aware, may be a pipe dream and, even if accomplished, could end up being unique but random rather than unique and meaningful, in which case I'm trading a unique piece of crap.
  6. your slippage will be dependent on the type of strategy you're running and your latency to market (i'm assuming since you're asking this question you're not very big). regarding the former, unless you're trading economic releases, you're not going to have much slippage (< .01). regarding the latter, if you're less than 40-80 ms, same as above.
  7. OP, Who are you trading/executing through? What type of hardware are you running it on and what type of internet/connection to your broker do you have?

    Your slippage will vary depending on all of the above.