This thread has been long dead. Recently i just research into something very s imilar to this, i found some merit in this actually. In gambling this mm called negative progressive betting, and they may have many variant.
Some premises:
- even with positive expectancy system, you still can not predict the event of consecutive ( or net series of) losses, thus drawdown. So kelly optimal formula cannot be used safely to determine risk per trade. Theoritically, the chance of consecutive loss increase as the frequency of trades. In reality, the consecutive event can just occur in the begginning and long enough to wipe out your account.
- the tenet and other premise of using negative progressive In trading are tese:
"if all other factors are equal, one win trade should be greater than a loss trade"
"since we cannot prefectly foresee the distribution of win loss sequence, trade in small size initially. E.g 0.1% or event 0.01%
"the bigger than exponential factor (1 for pure martingale, 0 for fixed lot), the better the recovery, but Quicker the drawdown grow. So why not choice something in between, rather than the two extreme to adjust/tune in with your trading setup.
When start trading with this mindset, the system expectancy is secondary, the primary task is to control sequence of trade event. I see nothing wrong with this, you just trade differently. Plus there are interesting aspects when combined with different RR setup.
Also i found this to achieve 20-50% anum quite achievable with accepatable drawdown( say 10-20 %) from historical data on original system that is not profitable at all on recent 3 years period.
Still the best, take all above with skeptical view, and do your own homework if get interested.