What rules do you use for your IS and OOS ranking?
6 replies
kasinath
3 years ago #267908
I am curios to know how people are filtering generated strategies, and using IS/ISV/OOS.
Please share how you assign these, and how you filter them.
- are you using in sample data on the rough parts of price action?
- Or only and isv on the not so rough parts are you using only in sample and oos?
- are you using multiple isv? Why?
- Do you filter based on stagnation? Or just Ret/DD ratio?
- Are you more strict with your IS or your OOS?
- Etc
Obviously, it all depends on what you are trading, what your goals are, the time period, etc, so please include that information in your reply!
Waid
3 years ago #267914
kasinath
3 years ago #267920
Thanks for sharing.
Here is one of mine.
Goal: Profitable Intraday Trend Follower
Symbol: BTCUSDT
Deposit: $500
Time Period: 2017.08 – 2020.12
Time Frames: M30+H4
Method: Breakouts and Trailing Stops. No Take profits
In this time period there are multiple market regimes (bullish, bearish, sideways), so I wanted to be sure that both my IS and OOS data include data from all regimes. To do this I used the alternating IS-OOS option available in SQX.
For filtering, I use a mix of filters, some for IS, some for OOS, and some for the entire range.
Would love to hear feedback from everyone: What problems do you see with this filtering, and how would you improve it?
Waid
3 years ago #267924
Hi Kasinath,
Just for curiosity, since you do IS and OOS testing by the “stripe method” (equally time with IS and OOS), why are you choosing Ret/DD(IS) >3 and Ret/DD(oos) > 1.5 as your filters?
hankeys
3 years ago #267925
IS/ISV or OOS – always differently
ranking by PF (bigger 1.3), RDD (min. 0.5 per year), number of trades (min. average 2 by month) , win% (min 30%, max 70%)
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
Waid
3 years ago #267936
I am thinking one thing:
filter with RDD lowest demanding setting
=
filter with high fitness (with strategy quality ranking(fitness) = RDD)?
For instance, set the fitness to 0.85
kasinath
3 years ago #268201
why are you choosing Ret/DD(IS) >3 and Ret/DD(oos) > 1.5 as your filters?
Thanks for the question, and sorry for the delayed response.
In general, I expect the Ret/DD on the IS data to be higher than the OOS, since the algo was trained on this data. I am less strict on rules for passing OOS data, so there will be less chance of overfitting across the entire data set.
My specific numbers (3 vs 1.5) vary depending on the market ,and style of strategy I am generating.
In my initial generation, I do not include Ret/DD, to see what kind of numbers show up in the generated strategies, to get a chance of what to expect, then i make a decision based on that.
hope this helps!
Viewing 6 replies - 1 through 6 (of 6 total)