Contrived Data for Backtesting

Discussion in 'Data Sets and Feeds' started by Norm, Jan 7, 2006.

  1. Norm


    Does anyone use non-market data for backtesting? I mean data that was artificially generated, such as via a random number generator?

    Or, has anyone put together a sort of torture test from real market data by a cut and paste process to assure that all relevant market conditions are captured within a single data set (or maybe a series of data sets).

    In either instance, the goal is to have a data set (or maybe a series of data sets) that provide a good backtest and thus avoid the need (at least for preliminary testing) to obtain real market data. This would be most significant for long periods of tick data, which are not always readily available.

  2. Do you believe these random generated data can be used to replace real data?
    I personally don't believe you can use random data, because the market is driven by humans, so the moves are based on mass behaviour and not at random.
  3. Norm


    spike 500 ,

    I agree that truely random data would not be very representative of market data. However, the other side of the coin is that old market data is not necessarily representative of new (future) market data.

    The ideal solultion might just be contrived data that contains all possible relevant market conditions. This contrived data could be artificially generated or might be a sort of composite containing real market data segments.