Let's say you see a very favourable trade setup. For example, something with a 75% win rate and a 5:1 payout. Obviously, the Kelly Rule recommends betting way too much to be sane. My question is what should be the absolute maximum bet size for really favourable conviction plays. If we assume that no trade can ever be known to have a win rate of over 75%, due to the inherent uncertainty of markets, what is the maximum optimal bet size for these slam dunk trades with high conviction and excellent potential payouts when they work? Assume a maximum drawdown tolerance of 20%.
Too low win rate to take this bet with risk more than 2%, at least for me. For 95% win rate I will go up to 10%. Forget about Kelly. This betting system is for gamblers.
It really depends upon how much conviction you have that the favorable setup will yield the anticipated results. If I really believed I had odds like that that I would easily bet 20% of total wealth. You almost could not bet enough on a situation like that if the odds were truly predictable/fixed. The problem is in trading, they never are.
If you want to be safe and not risk ruin, then 5-10% maximum. I used to bet more when I had real conviction in a trade and was on a winning streak and feeling confident, but there are just too many situations that can crop up with correlations etc that causes risk to be greater than you may have anticipated. Suppose you generally risk 2% and you push it up to 6%,8%,10% you have to be ready to cut your size right back down when you start losing. This is the primary reason bet scaling is so difficult- most push their bets when winning, but after a few losers they do not cut back down (or even worse betting even more!).. and so end up blowing the whole account. If you have a strong edge do not be too afraid to push the boat out a little bit- you don't know how long your advantage will last, so make hay while the sun shines...
Assemble a statistical sample consisting of all the historical occurences of your setup. Make it as exhaustive as your resources permit. Check the worst case scenario that has unfolded in the past. Apply a leverage that makes that worst case comfortably below your pain threshold of 20%. Remember, the worst outcome is yet to happen.
Agreed. There is also no problem with Kelly. The problems usually arise because of faulty inputs (such as believing you have a 75% chance of receiving 5-1). Joe.
I would suggest looking at it slightly differently. Take the maximum amount you are willing to lose and set it into a different account solely for the purpose of doing Kelly betting on these types of plays. As that account grows (and, if your Kelly is positive and your set-up continues to pay out, it will grow), rebalance every once in a while so that your Kelly betting account doesn't become so large that if you lose it, you will experience regret, while, since you are betting the theoretical maximum when you do bet, you won't experience "winner's remorse" at not betting more when you win. I would assume that such a setup would not be very frequent, so you wouldn't have to worry about overtrading.
I don't think it's always the case that the "worst outcome is yet to happen". For example, if you recalculate Kelly after every trade and it is declining, you might bet Kelly on every trade from now until you stop trading and never experience a larger percentage loss than you did on a prior trade when your Kelly percentage was larger. Yes, you may experience a larger point loss, say you have a 50 point loss on a trade and your previously largest loss was a 45 point loss, but the actual percentage of your account lost on the 50 point loss trade would be smaller, since the Kelly percentage bet on that trade was smaller. Also, you could just as accurately say that your largest best outcome is yet to happen, too, unless your trading strategy is to simply trade for a fixed number of points per winning trade.
For the 20 or 30 system I use that have a 95% win rate, I go as high as 12%. The 99% ones, I go to 15%.