Reverse Kelly criterion?
So y’all know how many investors and traders use the Kelly criterion to calculate the most optimal position size for a trade? I was thinking whether or not if it’s possible, since most people are more wrong than right, to use the Kelly Criterion instead to optimally short?
For instance, short all of Jim Cramer’s recommendations or some shitty sell side research (the structure of the short trade could prob use a bit of discussion) and then position each short based upon the Kelly criterion? A possible future step would then be to design a trading system around this strategy and structure some hedges to remove risk like beta. Theoretically, in the long run, wouldn’t the portfolio result in a net gain?
Of course, I didn’t really dig deep into the maths behind the Kelly criterion so idk if this is even possible, but this was just a sudden idea that came to me today and I want to know if it’s feasible just for curiosity sake.
Additional question: do L/S hedge funds (think, the usual suspects in the MM/SM field) use the Kelly criterion?
kelly just tells you how to size bets based on your statistical edge
ive never seen any thoughtful way to derive an estimate of your "statistical edge" (eg i am 52 vs 60 % likely to be right here!) in the long/short equity world, and if you talk to PMs, some are aware of the concept but practically no one uses it based on my internship experience
don't think the approach would be super diff for shortselling, it's just another bet where u try to figure out your edge and then use kelly to derive optimal sizing based on ur edge
kelly ignores: risk/reward ratio (uncapped downside on shorts etc means u need to size them all smaller than longs)
one way you could figure out your statistical edge w kelly is to look at your trading history, but that assumes you already have a large enough sample size
kelly is much more relevant for short horizon systematic strategies/high frequency stuff since u actually get a good amount of meaningful data on your statistical edge that u can precisely measure
if you are doing anything longer horizon than calling the next quarter, idk how you would get remotely meaningful data on ur statistical edge without waiting like a couple decades
I agree with the above, but I would just like to add further caution to blindly using Kelly. It’s basically sizing expected value vs. r/r. If you get overestimate the r/r it’ll tell you to put on a big position (even if you use fractional Kelly) and then you’ll get annihilated on the downside. Always think about the *noise* in your estimation of expected value and risk/reward. This is the same reason MV optimization doesn’t work. I’d encourage you to keep things simple and size by proportional expected (idiosyncratic) return where you choose a number of positions (+ other factors) to target vol.
yep
Eius sit impedit consectetur ad repudiandae et sunt. Esse deserunt deleniti impedit assumenda hic voluptatem possimus laborum. Nihil itaque quam perferendis dolor eveniet doloribus.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...