Walk Forward efficiency ratio

Vonasi · Thu, 23 Nov 2017 23:23:24 +0000

Am I right in thinking that a perfect overall walk forward efficiency ratio would be 100% with 50% being an acceptable level and below 50% unacceptable. Over 100% would be questionable. So an absolutely perfect walk forward test would be 100% overall efficiency ratio with say five IS and OOS periods all 100% ? I’m still trying to decide what is a good test result and what is not. I have a daily trading strategy that has 83% overall Efficiency Ratio with IS/OOS returns of 85% 118% 59% 86% 95% when WF is run with a dummy optimization value. Only 122 trades but with a Winning Trades percentage of 97% and a Gain/Loss Ratio of 90% Should I be pleased with this result? I’m getting a slight feeling of ‘too good to be true!’ so I am starting to question my knowledge of WF ????

This topic has 2 replies, 3 voices, and was last updated 7 years ago by jebus89.

Tagged: walk forward

Viewing 3 posts - 1 through 3 (of 3 total)

11/24/2017 at 12:23 AM #53781

Vonasi

Am I right in thinking that a perfect overall walk forward efficiency ratio would be 100% with 50% being an acceptable level and below 50% unacceptable. Over 100% would be questionable.

So an absolutely perfect walk forward test would be 100% overall efficiency ratio with say five IS and OOS periods all 100% ?

I’m still trying to decide what is a good test result and what is not. I have a daily trading strategy that has 83% overall Efficiency Ratio with IS/OOS returns of 85% 118% 59% 86% 95% when WF is run with a dummy optimization value. Only 122 trades but with a Winning Trades percentage of 97% and a Gain/Loss Ratio of 90%

Should I be pleased with this result? I’m getting a slight feeling of ‘too good to be true!’ so I am starting to question my knowledge of WF ????

Vonasi
Moderator

Topics: 192
Replies: 5088
Been thanked: 1811 times

11/24/2017 at 8:06 AM #53789

Nicolas

You have understood the general principle and the ratios that you get are perfectly correct! Congratulations.
122 orders do not seem to be much, but that depends mainly on the observed period, the strategy and the timeframe used of course.

An OOS ratio above 100% is indeed suspicious, but it is necessary to analyze why such a performance took place? Greater volatility? Macro economic announcements? Global crisis ? There is sometimes a good reason and we can remove many doubts about the robustness test.

Nicolas
Keymaster

Topics: 81
Replies: 20863
Been thanked: 4774 times

11/24/2017 at 8:14 AM #53792

jebus89

Im not a pro or a very experienced member on this forum but i have a feeling that 122 on daily timeframe is… bad? I mean with daily candles you can go a few years back, if you only get 122 trades im guessing thats 0.1 pr day or something like that? Tends to happen when a strategy is indeed, too spesific (curve-fitted) for the backtest data that you have. Also the results just seem to good to be true hehe just my 2 cents 🙂

jebus89
Participant

Topics: 118
Replies: 586
Been thanked: 238 times
Author

Posts

Viewing 3 posts - 1 through 3 (of 3 total)

Find exclusive trading pro-tools on