the bankroll of the typical person accepting startup equity in place of cash, the Kelly Criterion may indicate that startups should usually not be more than hobbies for “normal” (non-rich, non-certainly-immortal, declining-utility-in-dollars) humans.
Let’s see… Somewhere Paul Graham says that >90% of startups will fail, so our Kelly odds are 9:1. What’s the return on a won bet? Well, the recent Kaufmann Foundation report on VC funds puts the single best VC funds at an overall return of ~8x but that’s not enough because that implies that we may not even break even if we lost ~9 investments for every 1 investment returning 8! (receiving 8 back on a 9:1 bet)
If startups are negative expected value, the KC is not useful: it presumes bets are positive expected value and the question is what fraction to bet at any time to avoid ruin. I suppose that treating them as lottery tickets and assuming you are risk-seeking might make it useful, but I don’t know how to do that.
Maybe time-value will help. Thinking of a LWer I know, he received the rough equivalent of a year’s salary when the startup ‘won’. But the startup itself took years and naturally wasn’t paying the salary a big competitor might, so it’s not obvious that he was better off in the end, which brings us back to the expected value question.
Yeah, but I don’t think that really applies to startups! (What is ‘the other side’? Are there people who offer shorts on arbitrary startups for less than millions?)
I don’t understand your calculation. Even the best VC funds probably make some losing investments, so to achieve an overall return of 8x, the winning startups must yield more than that.
If startups are negative expected value, the KC is not useful: it presumes bets are positive expected value and the question is what fraction to bet at any time to avoid ruin.
On the contrary, in some sense, that’s when the KC is most useful. The correct amount of money to gamble on losing propositions is 0!
My estimation of startups in general is that startups are a good way for exceptional individuals to capture much of the value they create. The problem is that it’s difficult to tell who is exceptional beforehand, especially if one can only measure sparkle and not grit, and also especially if one has not determined their own level yet.
In that vein, I am cautious about finding cofounders in ways like this.
The major value-add of professional VCs is that they are (should be) better at picking startups than most people. It’s very much possible for 90% of startups to fail while VCs still make money. (For one, successful startups can use much more capital; and the rest of the money is supplied by unsophisticated founders.)
I was under the impression that VCs often had significant industry contacts that the fledgling company would then have access to, and that advice for founders is to not sign a deal with someone who is only offering money. (Of course, that advice given by a connected investor is self-serving, and should be taken with a grain of salt.)
I’m not too clear how we would apply KC to startups (as opposed to specific contracts in prediction markets).
Let’s see… Somewhere Paul Graham says that >90% of startups will fail, so our Kelly odds are 9:1. What’s the return on a won bet? Well, the recent Kaufmann Foundation report on VC funds puts the single best VC funds at an overall return of ~8x but that’s not enough because that implies that we may not even break even if we lost ~9 investments for every 1 investment returning 8! (receiving 8 back on a 9:1 bet)
If startups are negative expected value, the KC is not useful: it presumes bets are positive expected value and the question is what fraction to bet at any time to avoid ruin. I suppose that treating them as lottery tickets and assuming you are risk-seeking might make it useful, but I don’t know how to do that.
Maybe time-value will help. Thinking of a LWer I know, he received the rough equivalent of a year’s salary when the startup ‘won’. But the startup itself took years and naturally wasn’t paying the salary a big competitor might, so it’s not obvious that he was better off in the end, which brings us back to the expected value question.
Yeah, I dunno.
KC does apply to negative EV bets. The formula emits a negative allocation (ie “take the other side”).
Yeah, but I don’t think that really applies to startups! (What is ‘the other side’? Are there people who offer shorts on arbitrary startups for less than millions?)
I don’t understand your calculation. Even the best VC funds probably make some losing investments, so to achieve an overall return of 8x, the winning startups must yield more than that.
I did say it doesn’t make a whole lot of sense to me.
On the contrary, in some sense, that’s when the KC is most useful. The correct amount of money to gamble on losing propositions is 0!
My estimation of startups in general is that startups are a good way for exceptional individuals to capture much of the value they create. The problem is that it’s difficult to tell who is exceptional beforehand, especially if one can only measure sparkle and not grit, and also especially if one has not determined their own level yet.
In that vein, I am cautious about finding cofounders in ways like this.
The major value-add of professional VCs is that they are (should be) better at picking startups than most people. It’s very much possible for 90% of startups to fail while VCs still make money. (For one, successful startups can use much more capital; and the rest of the money is supplied by unsophisticated founders.)
I was under the impression that VCs often had significant industry contacts that the fledgling company would then have access to, and that advice for founders is to not sign a deal with someone who is only offering money. (Of course, that advice given by a connected investor is self-serving, and should be taken with a grain of salt.)