Riddle

New comment
 

Hi every one.

suppose you have two strategies both working like random and the strategies use totally different tools(indicators or whatever you use). 

strategy A ==> win rate=50% and RR ratio=1:1

strategy B ==> win rate=50% and RR ratio=1:1


If a signal is issued by both A and B then the possibility of the signal ending up in profit is: (0.5*0.5=0.25) ==> 25% 

Because signals are independent and we can use the formula:

P(A∩B) = P(A) · P(B)

Isn't it strange. You have two strategies without any edge over market but their combination bring you an edge of 75%.

 
Yashar Seyyedin:

Hi every one.

suppose you have two strategies both working like random and the strategies use totally different tools(indicators or whatever you use). 

strategy A ==> win rate=50% and RR ratio=1:1

strategy B ==> win rate=50% and RR ratio=1:1


If a signal is issued by both A and B then the possibility of the signal ending up in profit is: (0.5*0.5=0.25) ==> 25% 

Because signals are independent and we can use the formula:

P(A∩B) = P(A) · P(B)

Isn't it strange. You have two strategies without any edge over market but their combination bring you an edge of 75%.

I think its still 50% unless the signals are opposite

 

I guess you meant 

signal ending up in loss is

instead of

signal ending up in profit is


:)

Yes, the hard part here is to be sure that these are really independant.


This idea is being developed further: Construct a portfolio of N EAs, which in total deliver white noise and trade it (now using martingale is safe!)

https://hudsonthames.org/caveats-in-calibrating-the-ou-process/

Caveats in Calibrating the OU Process - Hudson & Thames
Caveats in Calibrating the OU Process - Hudson & Thames
  • hudsonthames.org
This is a series where we aim to cover in detail various aspects of the classic Ornstein-Uhlenbeck (OU) model and the Ornstein-Uhlenbeck Jump (OUJ) model, with applications focusing on mean-reverting spread modeling under the context of pairs trading or statistical arbitrage. Given the universality and popularity of those models, the techniques...
 
Obviously there is a flaw in reasoning. Because you take 100 of these and the edge is the bankruptcy of all other market participants.
 
Lorentzos Roussos #:

I think its still 50% unless the signals are opposite

İ think the problem is:
 possibility of both signal simultaneously is almost zero.
 
Yashar Seyyedin: suppose you have two strategies both working like random and the strategies use totally different tools(indicators or whatever you use). 

strategy A ==> win rate=50% and RR ratio=1:1
strategy B ==> win rate=50% and RR ratio=1:1

If a signal is issued by both A and B then the possibility of the signal ending up in profit is: (0.5*0.5=0.25) ==> 25% 

Because signals are independent and we can use the formula:

P(A∩B) = P(A) · P(B)

Isn't it strange. You have two strategies without any edge over market but their combination bring you an edge of 75%.

No! The strategies are independent of each other. Neither has an effect of the other.
 
Yashar Seyyedin #:
İ think the problem is:
 possibility of both signal simultaneously is almost zero.

Well you can have a tolerance level and you can test these .

Let both strategies run in the tester , pull their trades out , mark wins and losses and time and strategy.

Then feed them back in from the file you exported (no need to constantly trade in the test) use a custom criterion, give the genetic algo some selection criteria to work on and see how they combine .*the mt5 tester

 
Lorentzos Roussos #:

Well you can have a tolerance level and you can test these .

Let both strategies run in the tester , pull their trades out , mark wins and losses and time and strategy.

Then feed them back in from the file you exported (no need to constantly trade in the test) use a custom criterion, give the genetic algo some selection criteria to work on and see how they combine .*the mt5 tester

İ am too lazy to do that. Let's get back to the sellers' dying thread.
 
Yashar Seyyedin #:
İ am too lazy to do that. Let's get back to the sellers' dying thread.

It's not that boring or difficult , the problem is would you enter after the "second" strategy fired ? And then you'd have to have some sort of a "box" around the first signal bound by time left and right (bars) and price top and bottom and then the tester would have to figure out what the size of the boundaries is for any improvement 

It'd be more fun though to export one strategy , mark it's winners and losers and then try to find what "relates" to either one of them
 
Eugen Funk #:

I guess you meant 

signal ending up in loss is

instead of

signal ending up in profit is


:)

Yes, the hard part here is to be sure that these are really independant.


This idea is being developed further: Construct a portfolio of N EAs, which in total deliver white noise and trade it (now using martingale is safe!)

https://hudsonthames.org/caveats-in-calibrating-the-ou-process/

Too scientific 
New comment
Reason: