Backtested vs. forward performance

Discussion in 'Automated Trading' started by andrewbee, Oct 4, 2010.

  1. Hi,

    Who here has run an automated system that performed as well going forward as it did in backtesting? Over at least a few months anyway. If you did, how long a period did you backtest over?

    I have had a lot that failed (call them seminars), some that broke even, and some that made money (albeit less than the backtested performance).

    Thanks
     
  2. I have systems that I developed 10 years ago and they are still profitable. I backtested them for 8 - 12 years before that. The problem is that they do not generate many trades and I hesitate risking a big position. They are all EOD position trading systems. I have found out that as I moved towards intraday timeframes the forward performance deteriorated faster. Apparently and surprising, the same happened to EOD trend-following systems. Maybe it is me, not a general ruel.
     
  3. It takes a long time to tweak your backtest engine and understand your backtest data vs. your live data and execution system. My backtest data is fairly accurate to what real world results I get.

    I don't really use simulation/walk forward, just turn the strategy on if backtest numbers are decent and tweak a bit after real live testing.