I ran a simulation
I modified the simulation to it's running the 121x "red" plinkopot line:
Quote
r = random.random() * 65536
if r >= 26333 and r < 39203: p = 0.3
elif r >= 14893 and r < 50643: p = 0.5
elif r >= 6885 and r < 58651: p = 1
elif r >= 2517 and r < 63019: p = 1.4
elif r >= 697 and r < 64839: p = 3
elif r >= 137 and r < 65399: p = 5
elif r >= 17 and r < 65519: p = 13
elif r >= 1 and r < 65535: p = 47
else: p = 121
and generated some plots of the average log bankroll growth against percentage of bankroll risked.
It takes a lot of rolls to get a good smooth curve, presumably because of the high variance of the 121x payout.
First attempt:
Second attempt:
Both show that risking somewhere around 40% of the bankroll per game is optimal, but that risking half that isn't anywhere near half as bad.