Conversation

SBF
@SBF_FTX
Replying to
@SBF_FTX
5) Non existential risk is..... different. a) One view: it doesn't matter much, you win some you lose some. Again here "risk" means "0 EV risk", so upside is as big as downside b) Another view: sure but if you lose a lot of them then you're back at existential risk
1
29
SBF
@SBF_FTX
6) So, ok, where does this leave us? Well it means that other risk matters to the extent, but not as much. So how much?
1
20
SBF
@SBF_FTX
8) Winning $3 is 3X as good as winning $3. Risking $3 is _more than_ 3X as bad as risking $1. In fact it's *9x* as bad. (Assuming law of large numbers.) Why? Well say you have EV = +1 and risk = 1. Then you do that 25 times. You now have EV = +25 and risk = 5 (!!!).
2
39
SBF
@SBF_FTX
9) That's how gaussians add. And if you have enough things without wacky tails, that's how _anything_ adds. This is equivalent to: variance goes as sqrt(time). So the real cost you pay to risk is the square of the risk.
3
39
SBF
@SBF_FTX
10) What does that mean? It means that EV / RISK is not the right metric. EV / (RISK * RISK) is. (Unless you're making many bets at once and RISK is correlated between them, in which case you add the RISK!!!)
2
36
SBF
@SBF_FTX
11) So risking 10% of your stack isn't 1/5th as bad as risking 50% of it is. Risking 10% of your stack is only 1/25th as bad. That's a lot less bad! BUT REMEMBER EV IS LINEAR. "RISKING" 10% IS REALLY BAD IF YOU'RE DEFINITELY GOING TO LOSE EACH TIME!
2
54