Back testing and walk forward is completed. I'm paper trading now in IB a weekly system with an average bars (weeks) held of 46 . Avg wins are 61 bars. 68 trades per year on average. Gaining statistical significance for this system is too long for my patience. Damn, any suggestions on speeding up the process?
Paper trading for a year!
- Thread starter DingoCC
- Start date
So you're telling me you lost the borderline once in a lifetime opportunity the market has offered since April?, you are a very patient person indeed
Did you complete at least 200 to 500 historical trades from your backtest?
Quite frankly forget about live testing and the so called "forward" testing, what is really important is the number of historical trades you have already tested.
Personally I never trade a system unless I have backtested it with at least 20 years of historical data, on the 5 to 15 min charts.
Yes, A walk forward over 12 years produced 823 trades. A five year back test produced 463 trades.
Paper trading tells you if your execution systems are working and possibly lets you verify your friction/slippage assumptions. It should not be used for statistical significance, and indeed can't practically be used that way because it takes too long. Especially for a weekly system.. forget about it. The walk forward test should be all you need for significance - combined with some intuition about overfitting. Are all your parameters learned walk-forward? If you are still worried about data mining you can look into cross validation.
What's the use of a walk forward test if the backtest reveals that the system is losing 8% a year on average, after 1200 (historical) trades?
Last edited:
Walk forward used two key params only. I'll look into cross validation.
Make sure when you guys adjust the system over your 20 years of backtesting data that you properly curve fit it /s
you guys are ridiculous.
you guys are ridiculous.
Supportive and helpful. :-|
Wow, love your patience.