When to give up EV for reduced variance?

#1
When does it make sense to give up EV for reduced variance? I'm thinking about insuring BJ or a 20.

I guess I'm looking for two answers:
One in theory
Two in practice
 

Gamblor

Well-Known Member
#2
In general anything where you have a big bet out, and are still close to the index, I would be gun shy about making the doubles and splits. Some of the more common situations are:

-Doubling 8 v 5
-Doubling 10 v 10,A
-Doubling 9 v 7
-Splitting 10's v 5,6
-Splitting 88's v 9


Similarly would tend to take a surrender more often at a shade under the index:

-Sur. 14 v 10, 14 v 9, 13 v 10, 15 v 8, 16 v 8, etc.,
 

Zach Black

Active Member
#3
In Ian Anderson's "Burning the Tables", he suggest an insure all approach as a low cost form of cover for high level players. SW gave analysis that the cost (EV) of this cover is minimal.(4-5% of the EV)

I insure 19,20 & BJ around a TC of 2 to reduce variance when the index is 3 for hilo.
 
#4
Look up risk averse indices. That is what they are all about, giving up EV on certain hand match ups with weak correlation so you can win more money long term by betting more on each hand.
 

rrwoods

Well-Known Member
#5
The "big picture" answer is to remember that EV and edge are different things, and that you should give up edge and reduce variance when doing so would allow you to bet more money and therefore increase EV.
 
#6
lowering variance

The biggest way to lower variance is to bet a fraction of Kelly resizing, so one is not resizing their bets as often.

Example:
Full Kelly resizing increases N0 * 9
Half Kelly resizing increases N0 * 1.87

Half Kelly has 75% the growth rate of Kelly with a lot less variance/lower N0.

With bigger, non replenishable banks; $100,000+, 1/4 to 1/8 Kelly lowers N0 close to the N0 for flat betting.
 

iCountNTrack

Well-Known Member
#7
Mr. Ed said:
When does it make sense to give up EV for reduced variance? I'm thinking about insuring BJ or a 20.

I guess I'm looking for two answers:
One in theory
Two in practice
What you really need to look at is the ratio of the expectation value to the standard deviation on that play. The ratio has been coined by Don Schlesinger as the desirability index.

for instance if we look at 8,2 vs 9 (1 deck, s17)

The EV of hitting is 0.120803495385 ± 0.935247682187
And EV of doubling is 0.174427797338 ± 1.90495414525

Correct play would be to double if based on EV alone, however if we look at the ratio of EV to SD we get

hitting EV/SD = 0.1291673828076
doubling EV/SD = 0.091565352254

so if one wants to include risk, the correct play would be to hit.
 
#10
iCountNTrack said:
What you really need to look at is the ratio of the expectation value to the standard deviation on that play. The ratio has been coined by Don Schlesinger as the desirability index.

for instance if we look at 8,2 vs 9 (1 deck, s17)

The EV of hitting is 0.120803495385 ± 0.935247682187
And EV of doubling is 0.174427797338 ± 1.90495414525

Correct play would be to double if based on EV alone, however if we look at the ratio of EV to SD we get

hitting EV/SD = 0.1291673828076
doubling EV/SD = 0.091565352254

so if one wants to include risk, the correct play would be to hit.
What would happen if one construct a BasicStrategy with this approach ?
 
Top