What is my optimal bet size if I am running $600K with portfolio margin and I am only shorting UVXY? How would I got about getting the best position to withstand the occasional huge spike?
I like that you are shorting UVXY (has been a favorite of mine for a while), but sizing with Kelly criteria should relate to your specific trading strategy IFF if you have enough history and variance confidence to expect that to play forward. For example, I had a trade using UVXY that I used Kelly criteria until the forward playing ceased matching the history (ie wheels came off my wagon), forcing me back to drawing board! -- Am very curious of what you may be doing.
How much of your account can you afford to lose on a worst case spike loss? Size your position according to that. Also to consider is getting two worst case losses in a row or very close together. The market has a habit of doing that.
Responding to the last two posts: Well, I would think that the history of UVXY would serve as a model for history and variance, maybe go back to the change to 1.5X leverage. I guess from that you calculate SD? ... I would place my loss, ordinarily, as 2% of the account but given that VIX reverts to the mean and -- eventually, and that could be a long time -- UVXY slides back down toward zero (unlike other long or short stocks), I'm wonder if I should raise that for this particularly trade. ...
%% Exactly\ but consider something else. WHAT if you made much more than you thought you would or could+ what he is is trading looks like a nat gas chart. T Boone Pickens went broke once with his nat gas co. So if i blew up an account once or went broke trading something nat gas; make sure it never happens in 2nd worst case. Good points even though i would make sure i never blew up a 2nd accounts or ground down 2 accounts in to nothing. But i am over 33.3 years so that makes a difference/LOL
Kelly Criterion, Optimal F is all based on a model that as Stephen above mentioned assume forward risk/return would be kept the same as the past. Now assuming You are not using any system, you could use an approach that close/reopen the same trade every week on whole data history to calculate it. If data history is long enough You will find an expectancy and OF not that high, actually an r close to what is actually said to be considered a safe value. However If You are using a single model, You will find soon that is risky betting so much on a single system and expect same variance into the future risking your whole account value. Some use a fraction of the account, half optimal f to smooth this risk etc. What is interesting is optimal f at the portfolio level but difficult to calculate it. And even on a portfolio level dd will be terrific and very difficult to carry on even for iron traders.. one statistic that should be accounted in the model is the damage of the inevitable Maxdd that at least in the single systems composing portfolio is almost assured into the future. Whole point is that as You realize risks, add them into equation, the optimal f will be a value very closed to what is commonly used and considered safe. Is micro adding of new ideas, model, strategies, asset classes etc to the portfolio, assuming that are not correlated(big assumption) the whole secret to get what kelly and Vince tried to make. Theoretical vs reality equation is still an academic unanswered not contemplated model due to the rigorous mathematical approach used. What if your stop is cancelled by your broker? Risk like that can wipe out your account in minutes using above Kelly or optimal f academic models.