How do you backtest algos?

Discussion in 'Order Execution' started by mizhael, Aug 26, 2010.

  1. Lets say you are exploring some algo trading strategies, and lets take VWAP for the time-being.

    To put VWAP into test in real trading environment is expensive, because VWAP is only meaningful when you have a reasonably large number of contracts to trade. Therefore we cannot use 1 contract or 2 lots, etc. to test. We probably have to put up 100 contracts or lots at least for testing...

    That's costly.

    Same in backtest. You have to do 100+ lots instead of 1 lots, etc.

    But then how do you backtest the market interaction realistically?

    Any thoughts?

    Thank you!
  2. You don't give up, do you?

    You cannot backtest participant interactions. Forget it. You can simulate them. Learn the difference. Simulation won't help you. Too wide variations.
  3. wenzi


    You can't.

    As your size grows, you will know.
  4. Agreed!
  5. I knew I cannot... but maybe some smart people out there know how to do that?