Anyway to simulate "Competition of Shares"

Discussion in 'Automated Trading' started by GetBusyLiving, Feb 28, 2010.

  1. I try to get the most accurate results when I test my strategies and I barley ever back test. I forward test all my strategies using Market Replay from Ninja Trader at 500 speed (it works pretty well).

    I want to get even more accurate. I was wondering if there is a way to "simulate" competition for shares. I know in simulation mode if the shares are there it will grab them, but in real life someone else could have grabbed those shares before you.

    So anybody know of a way to simulate "competition" in forward testing?
     
  2. do you have a way to set for optimistic/pecimistic execution? Can you tell it to pay through by 1-2 cents?

    I record data and then backtest through live data with actual size on the L1 and L2 so I'm not sure what your settings are.

    I would suggest that you tell your simulator to pay through the bid/offer by a few pennies and if it still works then your executions will probably be better in real life.
     
  3. is there a way to reduce your reliance on signals/events?

    It sounds like the system waits for an event, than reacts...

    Rolling into and out of positions in advance of signals might limit the damage that other competition to your fills may cause.

    Reduce / increase your position in relation to the given risk that the signal / event will trigger, before that trigger happens.
     
  4. Do some research on a topic called Monte-Carlo.

    Be careful though. You can't get more "accurate" since you don't know what the market does or says. Market replay is not forward testing. It is back testing. Forward testing is when you run your strategy while the market is running.

    The idea of forward testing is to see how your strategy works under brand new circumstances that it has not yet seen. Thus if you run market replay multiple times you are in fact back testing.

    Simulating the grab for shares is not a simple topic and I doubt a tool such as Ninja trader do it properly. Not to say that Ninja trader is bad. I am saying Ninja trader is not geared for that market. The reason why I am skeptical is because people pay big money to do this. It is not simple stuff because you are trying to predict your footprint.
     
  5. Thanks for everyones input. I will try a couple things out.