Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
There are many threads about walk forward optimization, but something is still unclear to me.
My question is basically as follows: is the Walk Forward Optimization *always* a PLUS in comparison with traditional optimization? In other terms, is there any chance that a WFO is NOT necessary and/or useful?
Just a specific case, to make my point clear.
I’m currently testing a (purely automated/mechanical) strategy with very few parameters, and most of them seem to be unrelated to any kind of periodical change in the market I‘m trading (QM futures contract). I basically started asking myself: is it necessary/useful to include these parameters in the optimization, if they don’t seem to be correlated to any specific change in market behavior? (i.e. they basically seem to casually change from one period to another).
One of my mechanical strategies uses a fixed stop loss and evaluates the opening session gap size before taking action. I would usually prefer to set the stop loss just once, by considering the general MAE of the historical backtest. As for the gap size, I usually cannot see any particular change between the sizes of 2002 gaps and the ones related to today’s markets. So, does it make sense to include these parameters in the optimization?
What about a strategy that has NO variables/parameters at all?
I would be more than happy to listen to your opinions and clarifications.
Can you help answer these questions from other members on NexusFi?
Here is my experience, take it for what it is worth...
1. Whenever I optimized up to the present day with all data (a traditional backtest), and then started trading it, results were almost always bad.
2. When I do walkforward optimization, results are much better, but just because a strategy passes walkforward doesn't mean it will work going forward, or will work for a specified period of time.
3. Whenever I have done what I thought was "no optimization," I eventually found out that I was doing some type of optimization, just a little bit disguised. So, then it reverted to point 1.
You mention "a strategy that has NO variables/parameters at all." If this was the first and only strategy you have ever tested, and it performs well, I say congratulations, you may just have something wonderful.
If your "NO variables/parameters at all" strategy is your second, third, etc. attempt, then you have optimized to a degree - it just doesn't look or feel like a traditional backtest optimization.
I was particularly speaking about very simple strategies, like (for example): buy at "x" fib extension bkout and exit at the end of day.
I don't have enough of data to demonstrate this (so I might be wrong), but in some cases, when you take a very simple strategy as it is, it seems to work beautifully for a specific market (and probably only for that one). And even if you would like to optimize and apply a WFO on it... what kind of parameters are you supposed to submit to the testing engine?
As for the specific case I mentioned above: you may think of the "x" variable as a parameter (to be optimized), or simply use one the fib levels like it is usually understood and used by hundreds of traders (and I personally wouldn't think of an optimization for that).
If I thought an entry which had no parameters was valid, I might look at different exits which did have parameters, or maybe only take the entry under certain market conditions (bear, bull, flat, volatile, rangebound, etc).
Sometime when you see an indicator work superbly for a short time, it may just be random chance, or a cherry picked example. I love vendors who sell such indicators, because they look like the Holy Grail, based on their example cases. But when put to the test, most indicators are only accurate about 50% of the time. At least in my tests!
Can I ask you why you are asking this question? Do you fully understand the NT WFO procedure and all the related stuff that comes with it?
I find myself in a very similar predicament to you with a strategy I have been working on for a long time. My strategy is purely intraday. I have just let it run and run on various instruments for weeks and it seems to do what it says on the tin. I have not used the WFO (yet) and I admit that while I do understand in sample/out of sample testing concept, I dont fully understand the exact NT WFO process/procedure myself. Need to get my head around that though because everyone seems to say its a must for any auto strat.
PM me if you wish but either way i would be very interested in more of your thoughts on this..
You're absolutely right, but this is not always the case. I'm speaking about strategies tested on a 10-12year base, 5min charts, thousands of trades. The accuracy should not be in question.
Also, an objection to your point: adding conditions (like exit rules and so on) wouldn't also mean adding complexities and then decrease the degree of freedom of the overall system?
Yes, I think I understand the WFO process, even though I use it on TradeStation (and not on NT).
The point in my question is: is there any reason (or added value) in adding complexity to a strategy that seems to already work on its own? What's the meaning in adding a parameter, just to make the WFO process run in a meaningful way?
I do think that WFO is a tool, and not a goal in itself.
Also (my impression, but this is exactly what I'm asking your opinion about), I suspect that some unchanging (or randomly changing) market conditions don't require any specific optimization. By applying a WFO in cases like that, you would probably insist on tracking a market adaptation that doesn't really exist. The final outcome would risk to be unrelated to the real reasons that make your stategy work. This is really dangerous: your stategy may produce random results, for better or worse, but definitely out of control.
In my opinion, one of the best books related to the subject is Jaekle - Tomasini, Trading Systems. A new approach to system development and portfolio optimization.
The authors give special emphasis to the fact that in a robust system for every single out-of-sample period the suggested parameters don't change a lot, i.e. their change from one period to the one that follows is smooth and very often it's very small.