I guess you meant 20, 6.5 and 3.5 (whose average
is 10) or 20.5, 6 and 3.5.
With the 5% fee on the cashout, the question is nontrivial.
The general rule is that, assuming that the value of the wealth (bankroll) is the proportional to the logarithm of its size, the risk-adjusted profit is calculated as follows:
EV - variance / (2*wealth),
where the
variance is the mean square error. E.g. with the payouts of 20, 6.5 and 3.5 BIs, the variance of the picking is ((20-10)^2 + (6.5-10)^2 + (3.5-10)^2) / 3 = 51.5 "square buy-ins".
By declining the cashout, you save up the cashout fee of 0.5 BIs, but you assume risk that's equivalent to a penalty of 51.5 / (2*bankroll) BIs, and that risk penalty should be less than 1/2 BI to justify the prize picking.
Thus the cashout is worthwhile in this spot if and only if the bankroll (with the potential cashout prize added) is less than 51.5 BIs, i.e. if the BR before the bink was less than 42 BIs. (Was your question the ultimate question of life, the universe and everything?
)
I know that it's hard to calculate the variance of the pick in real time; the rule of thumb is that it's generally higher (worse) when only 1 of the spinner prizes is much higher than the average, than when 2 of the prizes are slightly higher than the avg (e.g. 14, 13 and 3 BIs, 14>13>10, variance = 74/3 ~ 24.7 BI^2).
Still, I normally decline the cashout almost always if the avg prize is less than 20 BIs, and also decline it if the avg is 20-30 BIs and 2 of the spinners are higher than it; otherwise I tend to take the cashout.
Last edited by coon74; 06-29-2018 at 01:03 AM.