A great book on testing and optimisation - page 5

New comment
 
Mak:
The WFA does not allow you to avoid tinkering.


Well, it depends on what you call fitting :)))))
 

Yeah, I haven't seen the discussion develop all weekend: My ISP had a crash - still not fixed (I'm writing from the office).

2 VelesFX: Honestly, I don't see why the output strategy quality from your sophisticated optimization can be statistically better than Pardo's proposed one.

2 Mak: well of course the fitting can never be avoided - simply because the dataset is always finite, and certainly 5-6 optimizations within WFA are still not conducted on a statistically significant dataset allowing one to say "now the strategy will always work since my model covers all possible market behaviors ". VelesFX, that's what I was talking about when I said "diversification across different market behaviours".

 
Mathemat писал (а):

2 VelesFX: To be honest, I don't yet see why the quality of the strategy output from your sophisticated optimisation could be statistically better than the proposed Pard


I was just very unclear about FT being only a panacea for overfittinga ...

И

The only obvious way to increase the resilience of the system is to increase the number of trades during the optimization process. ....

И

The only way to eliminate "randomness" in the system's real time operation is through diversification, but not through "diversification" which you claim is possible with WFA, I simply call it "diversification" as stability (just think of the terms :)))))). The real multi-model, multi-tool, multi-parameter diversification, when the system consists of 20-30 EAs, well, the more the better.

I think so :)))))

Alexey, are you going to take part in the Championship? Do you have something working?
 
VelesFX:

The only obvious way to increase the resilience of the system is to increase the number of transactions during the optimisation process. ... ..

И

The only way to eliminate "randomness" in the system on the real is diversification, but not "diversification", which you think can be achieved with WFA, I simply call this "diversification" as stability (you need to define the terms :))))).
Only God who makes quotes knows which way is obvious and which is completely useless. For example I think that multimarket testing in model development that works only on one instrument is not very necessary - and can even be harmful, because it can reject the really working models. What is the sense in making an EA working only on gold to test it also on currencies and indices, if gold behaves completely differently than they do? VelesFX, are you sure you have in your mind all the possible ways to increase the number of deals?

I'm really wrong with the terms, it's more of a sustainability thing. ...

About participating in the championship: well, yes, I would like to, but I'm afraid I won't have time to put up something really sustainable.
 
These methods are the only and obvious in my opinion only!!!!!
About the multi-market testing - today I've found parameters for one trading model on USDCHF, the system shows excellent curve (Small drawdown, Good expected payoff, Good profitability) with these parameters on the optimization test window. So if I test this model on AUDUSD, USDCAD, USDJPY, GBPUSD on the same test window (time frame is the same) then the system draws not a bad curve.
USDCHF

USDCAD

AUDUSD


USDJPY


GBPUSD


There is very low correlation between gold and indices (maybe)), therefore it's useless to test an Expert Advisor for gold on indices, but one can always try. I don't conduct WFA at all, I don't believe in it, I prefer to use "rough" optimization!!!!!
About taking part in the championship: well, yes, I'd like to, but I'm afraid I won't have time to put up anything really sustainable.
I have to take part by all means. And forget about sustainability, it's for earning money. The Championship is a gamble!!! The organizers have created all the conditions for equalizing the chances of winning between competent developers and outright amateurs, probably to create some intrigue. If I'm not mistaken, the Expert Advisor with an error won that year, so it is not stability that matters, but luck :))))).
 
VelesFX:
If I am not mistaken, the Expert Advisor with the error was the winner last year, so the main thing is luck and not stability :)))))
You are obviously wrong in such a simple statement to check. I dare say the rest of your claims about the Championship are untrue.

To make valid claims, you should at least familiarise yourself with public information:
 
VelesFX: And forget about sustainability, it's about making money. The Championship is a gamble!!! The organizers have created all the conditions to equalize the chances of winning between competent developers and outright amateurs, probably for the sake of intrigue.
I don't want to display something unstable, although there won't be any cheating if you tell others about it, in case the expert is in the lead. One would like something that can be put on the real with practically no changes. That's when it will be convincing for myself as well (apart from quite a considerable cash bonus, of course). Well, we'll see, there are still three weeks to go...
 

Mathemat ,

if you can read English, let me email you what you need,
or at least very interesting

 
No problem in English, I've translated a few books after all. Send it. Mail it to me.
 
geometrr:

Mathemat ,

if you can read English, let me email you what you need,
or at least very interesting

Excuse me, can you send it to my email? SSD@DP.UKRTEL.NET

Sincerely S.D.
123456789101112...22
New comment