nightoftheiguana2000
<mailto:nightoftheiguana2…@…com?subject=Re%3A%20Bob%20Dancer%27s%20LVA%20-%2019%20MAY%202015>
wrote:
"If your goal is bankroll growth with zero risk of ruin, keep your
bankroll over variance/edge/2 bets (or half the Kelly number if you know
it)."
Unfortunately, in the real world of casino play, there can be no such
thing as zero risk of ruin (RoR) with indefinite play because there is
always a minimum bet size, and it's possible for a long series of losses
to reduce your bankroll to that minimum bet. But also in the real world,
very few people keep their gambling bankroll completely separate from
their other funds. If you were to lose your gambling bankroll would you
quit gambling, or would you replenish it from other earnings?
Kelly can tell you what your optimum bet should be, but it can't make it
possible to bet less than the table/machine minimum. Sorokin won't tell
you directly what your bet size should be, but it can tell your RoR with
a given bet size, so if you can put a number on your emotional tolerance
of RoR then it can be used to find your optimum bet size. Don't forget
that the somewhat arbitrary division by 2 for Kelly is related to
emotional tolerance.
Also, variance is not a good indicator of risk. Consider these two games
which have roughly the same player advantage and variance:
Deuces Wild (full pay): ER=100.762% Variance=25.91 RoR=3.81%
All American Poker: ER=100.722% Variance=26.82 RoR=5.25%
In both cases, the RoR is per Sorokin assuming a starting bankroll of
5000 betting units (e.g., $25,000 on a 5-coin dollar machine). The
difference in RoR is much greater than might be suggested by the small
differences in ER and variance.
So is it really a good idea to use variance for the Kelly calculation?
How about a Kelly calculation that uses the Sorokin RoR instead of
variance? Then you would have be best of both types of calculation.
[Non-text portions of this message have been removed]