vpFREE2 Forums

Bob Dancer's LVA - 2 DEC 2014

#1

Albert Pearson wrote:
>I agree with the principles, I disagree with the math. How do you go from playing with a 99% accuracy to giving up 1% on the game ? I'm no math wizard. but I suspect that if I was playing at an accuracy of 99%, I'd probably be giving up closer to .01% than 1%. A.P.

And I agree completely with that. If a semi-skilled player makes an error in 1% of the dealt hands, those are mostly close decisions with small difference in EV, so the loss in actual payback is very small. That's why the Player History in OpVP tracks accuracy as the percentage of potential EV achieved instead of just the percentage of hands played computer-perfect for max EV. It's also why the program has features to evaluate the precise EV of a strategy chart and then check your play against the chart. It turns out that, for most games, a chart with no penalty considerations will yield within 0.01% of perfect max EV play. ~Dan

#2

I am sure this is what Bob means by 99% accurate. Not that 99 out of 100 hands are right but that for every $100 in coin-in you make $1 worth of error.

Accuracy measured this way means you are giving up 1% of EV and thus it's costing you 1%. My goal when training is an error rate of 0.1% accuracy. If I can get to that level then I am good. I know Bob likes it closer to 0.0001% :slight_smile:

As Dan points out good VP software will measuring accuracy this way (most do).

[Non-text portions of this message have been removed]

#3

I imagine kiwiboy/wayne is right (post appended below); a context of 1% ER cost seems more likely than erring on holds 1 out of 100 times, given the overall article focus. But, then, I'm not sure Bob kept straight what he meant.

He also remarks, "I retain in my memory 99% accurate strategies for perhaps 25 different video poker games".
Clearly this is a reference to % of hands played correctly ... not a suggestion that the strategies yield only 99% of optimal ER (I doubt Bob would commit the latter to memory :wink:

Bottom line, the thrust of the article is on the money, but seems to fall short of Bob's normal execution (of all gambling writers, he's certainly the most precise and exact ... at least in my book)

---In vpFREE@yahoogroups.com, <waynes@...> wrote :

I am sure this is what Bob means by 99% accurate. Not that 99 out of 100 hands are right but that for every $100 in coin-in you make $1 worth of error.

Accuracy measured this way means you are giving up 1% of EV and thus it's costing you 1%. My goal when training is an error rate of 0.1% accuracy. If I can get to that level then I am good. I know Bob likes it closer to 0.0001% :slight_smile:

As Dan points out good VP software will measuring accuracy this way (most do).

[Non-text portions of this message have been removed]

[Non-text portions of this message have been removed]

#4

I think the error rate of playing in a casino, caused by such distractions as noise, alcohol, sticky buttons, spotlight glare, and so on, is almost always much greater than the error rate of not remembering the computer perfect maxEV strategy for every hand.

Computer perfect maxEV strategy is fine in a vacuum, indeed mathematically in isolation it is the optimal strategy for EV, however, casinos are far from a vacuum. Given the distractions and frankly slot machines and lighting that are designed to cause errors and the various oddball promotions and punishments and even violations of civil rights being employed, often a simpler strategy or even a min cost strategy is more optimal in the actual environment of a casino. For example, on paper a royal flush may pay 800 bets, but in reality there are various additional hidden costs that bring that return closer to 400 or even 0. Now some may think that makes the game unplayable, but actually it doesn't, if an optimal strategy is employed. Note that the optimal strategy is dependent on the total environment of the game, computer perfect in vitro maxEV may be a place to start, but quite likely it is not the optimal strategy in vivo.

[Non-text portions of this message have been removed]

#5

nightoftheiguana wrote:

the optimal strategy is dependent on the total environment of the game, computer perfect in vitro maxEV may be a place to start, but quite likely it is not the optimal strategy in vivo.

Another advantage of at least heading toward min-cost from max-EV is
that usually it involves less variance. I'm surprised the phrase
"Kelly expected value" isn't more popular. I'd like to see a video
poker analyzer that allows for the logarithmic value of each hand to
be evaluated, so that what bankroll it takes to, say, keep 3 aces over
a full house in Double Bonus, according to the Kelly Criterion, could
be easily seen.

#6

Ditto. Somebody should update VPSM with an input for bankroll and an output of the Kelly optimal strategy. Or maybe somebody can pay the wizard to update his strategy generator likewise. In a pinch values can always be estimated with the Certainty Equivalence (EV - Variance/2xBankroll). I'm sure there are lots of errors generated by players assuming an infinite bankroll and not taking into account their current bankroll.

[Non-text portions of this message have been removed]