Another question: is this idea based on the assumption that ProQuant backtests are wrong whenever they indicate a high W/L ratio (e.g. 90%), but right whenever they indicate a low W/L ratio (the example I see being given is 10%)?
That is to say, when you generate a 90% win strategy, you do not believe it will win 90% of the time, but when you generate a 10% strategy, you do believe it will lose 90% of the time?
P.S. Additionally, have you taken into account that by swapping the signals you will not necessarily achieve the inverse W/L ratio as if a position closes within spread you will lose money regardless of whether you went long or short?