Support Board
Date/Time: Fri, 27 Dec 2024 18:27:29 +0000
Post From: replay back testing vs real-life
| [2014-01-15 17:27:10] |
| vegasfoster - Posts: 444 |
|
Replay/backtest will never be 100% accurate to live, sad but true. By and large backtesting and optimization are myths. It really just tells you if something is completely no good, but a strategy can test good and still be lousy live based upon a multitude of factors, the most significant being slippage on market orders, no-fills on limit orders, randomization and emotions. On the technical side, I'm not the right person to answer why simulation and replay would be different, but in lieu of a better response, I believe simulation accounts for the current bid/ask whereas replay does not, i.e. there is no level 2 in replay. I've previously requested level 2 in replay many moons ago, but it has not been implemented. Someone correct me if that is wrong. Also, make sure your intraday storage time unit for your data/trade service is set to 1 tick. |