Does confluence improve the winrate of a system?

We have a discussion with some friend on another forum about confluence : let's say I have two systems, one with a winrate of 56 % and the other with a WR of 63%. Now if we take trades only when both system are in confluence (both say to buy or to sell), will the overall winrate of the final system be improved over the two initial systems ?

thanks a lot

Jeff

• This "confluence" thing sounds similar to "voting ensembles" in machine learning.
– Flux
Aug 4, 2022 at 14:32
• Given the information provided, it is not possible to determine if the overall WR will improve, stay the same, or decline. Aug 4, 2022 at 14:57
• what infos do you need ? would you need the exact trades ? There is no universal rule about confluence ? I thought it could only improve WR
– Jeff
Aug 4, 2022 at 16:47
• machinelearningmastery.com/voting-ensembles-with-python "We can see the hard voting ensemble achieves a better classification accuracy of about 90.2% compared to all standalone versions of the model." does that mean the winrate is improved to 90 % ? that sounds insane
– Jeff
Aug 4, 2022 at 16:57

That is not how probabilities work. You cannot make any assumptions about the combined winrate without further knowledge.

Imagine a game/stock exchange with the following rules: you win if you buy on Monday or Tuesday; you lose all other days.

• Your systems might work this way: system 1 says "buy" on Mondays with a probability of 56% and on Wednesdays with 44%. System 2 says "buy" on Tuesdays with a probability of 63%, and on Wednesdays with 37%.

Your combined system says "buy" only on Wednesdays, giving you a winrate of 0%.

• Or your systems might work this way: system 1 says "buy" on Mondays with a probability of 56% and on Wednesdays with 44%. System 2 says "buy" on Mondays with a probability of 63%, and on Fridays with 37%.

Your combined system says "buy" only on Mondays, giving you a winrate of 100%.

We really cannot tell you which of those (or any other combination of probabilities) your systems fall into.

• Thank you ! So finally I made an algorithm that computes two very simple strategies (one with RSI and one with CCI) and the corresponding confluence, and what i've found out is that most of the time, the resulting win rate is the mean of both systems. Which makes sense somewhat and can be interesting as a kind of diversification
– Jeff
Aug 7, 2022 at 13:08