Bernie: "Combatting" variance/ negative fluctuations..6 Deck games

[Bernie]

Credible blackjack authors emphasize the complete mathematical randomness of standard deviation,variance, and fluctuations; thus we can't predict when or how severely negative fluctuation will occur. In the literature and on blackjack forums there are multiple accounts of highest losses when max bets are placed
at very high, player- favorable counts (highest losses and wins both usually
always occur under high counts; it's likely that the losses are just more retained in our memories). I'm a rookie AP (total of approximately 275 hours/7 months)
and have experienced these high count losses (presumably due to negative variance) and realize they're to be expected. An axiom of advantage play is to
continue placing proper bets as dictated by the count even when we're losing
repeatedly; the expectation is that we will win over the long haul, mathematically speaking. As a relative novice at BJ, this is still one of the most frustrating aspects I encounter, despite that it's expected and that we are not
to give in to the emotional barbs of losses. I suspect you've heard all these ramblings about losses numerous times. My inquiry regards methods to reduce effects of variance..ie..are there mathematically sound methods to
combat negative fluctuations? Several literature sources suggest the following as potential tools to reduce negative variance (if variance is indeed
random and thus uncontrollable, are they better termed as "means to enhance our controllable advantage" ?)

1. adherence to games with good rules to begin with (I always do this)
2. minimally acceptable penetration levels (75%- 6 deck; Ido this as well))
3. higher than adequate bankroll requirements
4. wonging / exit strategies
5. reduced ROR's ( < 1 and up to 2%)
6. play fewer decks (difficult to locate decent single and/or double decks

due to preponderance of 6:5 payoffs and/or poor penetration)
7. Spreading to multiple hands (cover?)
8. one author suggested "smaller bets"; doesn't reconcile with "getting the

money out there" when called for
9. Comp hustling
10. Risk-averse indices

I'd appreciate seasoned players comments/experiences on the above. And for any mathematicians, is there a way to attack variance that hasn't been
discussed in the literature yet ? I'm curious as to the potential effects of "short-term card observation" as a tool for this. IE, a run of high cards
(as we're hoping for at this count level) that doesn't reduce the TC below
+4 or +5 but we assume a relatively higher proportion of low cards are
next in line? Voodoo?..thanks..

[Don Schlesinger]

> Credible blackjack authors emphasize the complete
> mathematical randomness of standard
> deviation, variance, and fluctuations;

Well, that's what the terms mean! :-)

> thus we can't
> predict when or how severely negative fluctuation will
> occur. In the literature and on blackjack forums there
> are multiple accounts of highest losses when max bets
> are placed
> at very high, player-favorable counts (highest losses
> and wins both usually
> always occur under high counts; it's likely that the
> losses are just more retained in our memories). I'm a
> rookie AP (total of approximately 275 hours/7 months)
> and have experienced these high count losses
> (presumably due to negative variance) and realize
> they're to be expected. An axiom of advantage play is
> to
> continue placing proper bets as dictated by the count
> even when we're losing
> repeatedly; the expectation is that we will win over
> the long haul, mathematically speaking.

So far, so good.

> As a relative
> novice at BJ, this is still one of the most
> frustrating aspects I encounter, despite that it's
> expected and that we are not
> to give in to the emotional barbs of losses. I suspect
> you've heard all these ramblings about losses numerous
> times. My inquiry regards methods to reduce effects of
> variance; i.e, are there mathematically sound methods
> to
> combat negative fluctuations?

Short answer: no!!

Longer answer: Understand that there is a difference between winning as much as you can, and reducing variance, and that, in most cases, the two are contradictory ideas. There is always the Holy Grail of blackjack, which is to seek out games with surrender, which both increases expectation while decreasing variance, but that is the exception to the general rule that, to win more, you have to risk more.

> Several literature
> sources suggest the following as potential tools to
> reduce negative variance (if variance is indeed
> random and thus uncontrollable, are they better termed
> as "means to enhance our controllable
> advantage" ?)

Yes. And, again, it's important to understand the difference between the two. Virtually everything that you've written below can enhance your expectation, but, in almost all cases, it will probably increase your variance at the same time.

> 1. adherence to games with good rules to begin with (I
> always do this)

The better the rules (das, split to four hands, doa, etc., the higher the variance). Again, surrender is the exception.

> 2. minimally acceptable penetration levels (75%- 6
> deck; I do this as well))

The deeper the penetration, the higher the variance.

> 3. higher than adequate bankroll requirements

Has nothing to do with variance.

> 4. wonging / exit strategies

Wonging has higher variance than play-all, if you bet optimally.

> 5. reduced ROR's

Simply implies a higher bankroll, or a lower Kelly fraction. If you bet less, you reduce variance and reduce e.v., as well.

6. play fewer decks (difficult to
> locate decent single and/or double decks

Probably increases variance.

> 7. Spreading to multiple hands (cover?)

Increases variance.

> 8. one author suggested "smaller bets";
> doesn't reconcile with "getting the
> money out there" when called for

Well, it reduces variance! :-) But, I suspect that you're beginning to understand that this really isn't your goal.

> 9. Comp hustling

Reduces expenses, not variance of playing the game.

> 10. Risk-averse indices

Aha! You'd think that this ought to reduce variance, but the effect is minor, because, if you bet optimally, using r-a indices permits you to bet slightly more, so we're back where we started from. See the BJA3 discussion.

> I'd appreciate seasoned players comments/experiences
> on the above. And for any mathematicians, is there a
> way to attack variance that hasn't been
> discussed in the literature yet?

Not that I know of, or I would have discussed it. :-)

> I'm curious as to
> the potential effects of "short-term card
> observation" as a tool for this.

Fuhgeddaboudit!

> IE, a run of
> high cards
> (as we're hoping for at this count level) that doesn't
> reduce the TC below
> +4 or +5 but we assume a relatively higher proportion
> of low cards are
> next in line? Voodoo?..thanks..

Voodoo. You're welcome. :-)

Don

[jblaze]

Maybe I can offer an easy way to understand why seemingly good rules and games increase variance. It does seem counterintuitive, I agree, but it makes sense if you can understand.

The underlying theme of blackjack is risk management. You want to optimize your profit to risk ratio. SCORE provides a measure for this. As a player, you pick a risk of ruin you are comfortable with taking. "Kelly" betting offers the mathematical best risk reward ratio for you to take. However, many people cannot stomach 'full Kelly'. (And again, 'true Kelly' changes with each decision and is impossible to achieve in a real life setting).

Now say you find some relatively crummy game to play, like H17, ndas, sp2, 60% pen. If you decide to play (which you shouldn't), you say to yourself, OK well let me sim my method of playing and find a bet ramp that gives me my ROR. So you go ahead and do this, and you get some crummy SCORE for your crummy game.

But then you get tired of making pennies, so you find a better game, say S17, das, sp4, 80% pen. Now, if you use the same ramp, I believe your variance might actually decrease (though maybe not, I'm not exactly sure). Your ROR will certainly decrease. HOWEVER, you are not in it to decrease your ROR. You are in it to maximize your SCORE, that is, your profit to risk ratio. So, now you readjust your bet ramp to offer the same ROR as your previous game. And you should be betting much more, and consequently winning much more, for the same ROR. So, your variability will increase, your SCORE will increase, your win rate will increase, and your ROR will remain the same.

Hope that helps. In summary, good games allow you to bet more for the same risk, but this increases your variance.

[Don Schlesinger]

Nice explanation! Well done.

> Now say you find some relatively crummy game to play,
> like H17, ndas, sp2, 60% pen. If you decide to play
> (which you shouldn't), you say to yourself, OK well
> let me sim my method of playing and find a bet ramp
> that gives me my ROR. So you go ahead and do this, and
> you get some crummy SCORE for your crummy game.

> But then you get tired of making pennies, so you find
> a better game, say S17, das, sp4, 80% pen. Now, if you
> use the same ramp, I believe your variance might
> actually decrease (though maybe not, I'm not exactly
> sure).

No, I don't think so. The better pen is going to give you higher frequency of the better counts. So, you'll bet more money. So, your variance will be larger. Remember, variance has nothing to do with edge, but rather just bet sizes squared times frequencies, summed vertically.

Don

[Bernie]

> Nice explanation! Well done.

> No, I don't think so. The better pen is going to give
> you higher frequency of the better counts. So, you'll
> bet more money. So, your variance will be larger.
> Remember, variance has nothing to do with edge, but
> rather just bet sizes squared times frequencies,
> summed vertically.

> Don
Appreciate the advice from both of you..thanks