might’ve even happened at that point and he was just stealing customer money to cover it up.
not that i’d expect journalists to have been able to work it out; anybody who can understand kelly’s paper is making more money some place else.
might’ve even happened at that point and he was just stealing customer money to cover it up.
not that i’d expect journalists to have been able to work it out; anybody who can understand kelly’s paper is making more money some place else.
Regardless of his other arguments, it never makes sense to bet more than Kelly, because you move into _diminishing_ expected value… while also taking on more risk! You can do less than Kelly if you want to diminish your expected value, but at least that also comes with less risk.
right he also had a thing in there where i think he was conflating the log utility in kelly with human’s tendency to have a logarithmic view of the utility of money.
they’re not actually connected, they’re just both the words “log utility”.
You can argue that there is no such thing as a one-off bet, because there will always be more bets on other things for someone who isn't ruined, but that's the explicit assumption for that thread.
Anyway - anyone who has ever run any money knows you use a fraction of kelly, which if you have a brain makes you wonder why it is optimal to use a fraction of the optimal amount - which should make you realise that kelly is loaded with stupid assumptions and is basically useless beyond a philosophical construct.
and i'm fine with that. i don't believe it exists, especially for a business/fund like his, where even making that claim is its own sort of hilarity, but i'm fine with that for the purposes of the thread and this response:
> Kelly betting applies to repeated bets. ... In that situation, EV really does scale linearly with bet size, so if you have any edge you should just go big.
in which case you don't spend any time talking about how you're going to bet "X times kelly". you don't deny that kelly applies, and then compute some >1 multiple of kelly and defend that number by saying kelly doesn't apply. there's no fixed multiple of what you get out of the gambler's formulation that makes any kind of sense if you've abandoned the initial premise.
i suspect the only good, coherent defense of what he was saying on twitter is something like "he had articulate arguments but didn't present them because of twitter limitations"
not to mention the Matt Levine interview where Matt tells him he's basically describing yield farming as a Ponzi scheme and he concedes that's valid and doesn't even try to construct a counterargument for where the value comes from...
The scary thing is he understood that the box inherently is nothing of any value, except that the market has placed a value on it - yet still bought into lots of boxes that the market said had value till they didnt have and value. The gambler once again.