I see a potential problem with this that might not happen immediately, but is likely to pop up sooner or later. An EA needs to be constantly supervised, because unless it is self adapting (neural network style) then it will eventually need to be tweaked. Not every EA can work with every market condition, and you will end up losing money. The trick here would be to have good MM, and if your EA is constantly doing the 3 trade thing, then you need to either tweak your EA, or get another one.
IMO, a better solution would be to find the conditions that are causing something like that to happen to your EA, and adjust the EA. Test it for the long haul to make sure it works (for $ value) making more than it loses, because you will lose some trades. It is the only 100% guaranteed thing to happen in trading. It would be better to plan for winning trades anyway, and have a good way to limit losses before they wipe out your account.
Strategy tester reports consecutive losses in terms of average and maximum streak.
In your example, if an EA likes to lose 3 times, that is just the average. It can lose 1, 2, 3, or even 5 trades consecutively. And if the EA reverses when it should not, then you get another losing streak.
The idea you describe is somewhat similar to predicting at which point the ZigZag indicator will not repaint. You can use neural nets or statistical methods (such as Poisson) for this, but they have a margin of error. The only way I can think of that gives 100% accuracy when 'optimized' this way is data snooping, which is not good.