Question for APs

lk,kl

When to raise your bet is more important in shoe games than when to deviate from basic strategy. What count you use should be based on what you're comfortable with. If you look in Wong's Professional Blackjack, you will note that in his baseline game the difference between Halves and Hi Lo is you win $17 instead of $16 for doing an awful lot more work. If you are a marathon man playing long hours using a third level count means you will play less hours than if you used Hi Lo because you will tire more easily and make more errors. Again, to paraphrase Wong because he puts things into perspective so well, "You'd do better working on an act that allows you to throw an extra green chip on your big bets than to worry over which count to use." I've known and traveled with a lot of counters and the ones using Hi Lo and seeking the best cuts possible always outperformed counters using second and third level counts but placing less importance on penetration.

I've also generated indices using Arnold Snyder's algebraic formula, SBA and Wong's software. Many indices between the three sets of numbers differed by 1 or 2. I then ran 4σ simulations repeatedly with a 6 deck game with a win rate in the $86 range. The difference in win rate between the 3 sets of indices was around thirty two cents. The point is, these are trivial considerations.

What count are you using and what count are you looking at?

I have used Zen for a very long time. On rare occasions I will help someone who wants to get into counting and I use Hi-Lo for that. If they say they want to do it full time I try to talk them out of it, especially if they're only interested in straight counting. After my last "students", I stayed with Hi-Lo for about a year, only going back to Zen out of boredom. I did not see any significant difference after around 600 hours of play. Not scientific, I know, but it does back up what others have said.

I wouldn't worry about it. You're better off sticking with the good ol' Hi-Lo count. It's easy, as it's Level-1. You can play it more accurately for longer plays. It already has a good betting correlation, which is what you want, especially in multi-deck play. Find a good game with liberal rules and good penetration, and that will be worth more than some tweak.

km,;l';

jljljjlkj

Your questions were answered. Buy a sim program.

Quote: Sonuvabish

Let me put it a different way. There a couple changes I'm thinking of making. The first one was the simplest, adding zero difficulty to how I count. Another results in .015 increase in playing efficiency and no change in betting, little more complicated. Another adds the same betting correlation as the first, while only decreasing playing efficiency by .02, but is the most complicated and don't really like this one. Any idea the effects on expectation? Cuz I don't.

Sorry I don't know the values of that exchange outright. Correlations to optimal do not work to value exactly linearly. They are a decent estimate, but at the levels you are worrying about I think you need a LONG simulation to see the differences. I would further guess that parameters outside of the system effect the values noticably at the levels you are speaking of. (Penetration and rules might change the values of each bit of correlation in the payoff to you.)

hkjhj

Quote: Sonuvabish

Thanks, Mike. I wasn't privy to much of that. Now I really don't know what to do! Maybe we should just vote! Kinda how I always did it before, except I was the only one voting. It's worked out so far. If you had to choose one of those changes...anyone?

Gun to my head, without a simulation, I would guess you should take the betting correlation. Different magnitudes of correlation are a lot, but in shoe games with aggressive spreads BC is where most of the money is. (shoe games without EXTREME penetration)

(Disclaimer: I am predisposed to BC because I am SOOO bad at playing corrections)

Quote: endermike

Gun to my head, without a simulation, I would guess you should take the betting correlation. Different magnitudes of correlation are a lot, but in shoe games with aggressive spreads BC is where most of the money is. (shoe games without EXTREME penetration)

(Disclaimer: I am predisposed to BC because I am SOOO bad at playing corrections)

I'm not sure I agree with this.

Knowing your exact edge isn't worth that much. Whether your edge is 2% or 2.2% is more or less irrelevant -- your bets are going to be limited by the amount that you can get down without generating heat, not by knowing your edge to an extra significant digit.

On the other hand, if you can tweak it in a way that eliminates false positives or false negatives, that is worth more. I'm not sure how well the concept of "betting correlation" captures this -- these are two very different concepts which get mixed together when calculating correlation.

In other words, I'd rather have a count that is consistently off by 0.2% in the edge calculation, than one which is perfect almost all the time, but occasionally tells me that I have a 2% edge when, in fact, the house has the advantage (or vice-versa), even though the correlations might be the same.

Quote: AxiomOfChoice

I'm not sure I agree with this.

Knowing your exact edge isn't worth that much. Whether your edge is 2% or 2.2% is more or less irrelevant -- your bets are going to be limited by the amount that you can get down without generating heat, not by knowing your edge to an extra significant digit.

On the other hand, if you can tweak it in a way that eliminates false positives or false negatives, that is worth more. I'm not sure how well the concept of "betting correlation" captures this -- these are two very different concepts which get mixed together when calculating correlation.

In other words, I'd rather have a count that is consistently off by 0.2% in the edge calculation, than one which is perfect almost all the time, but occasionally tells me that I have a 2% edge when, in fact, the house has the advantage (or vice-versa), even though the correlations might be the same.

I split 10s the other day only because I was alone at the table. The point is, ploppies dictate how I play far more than the pit does. I couldn't care less about heat, unless that particular game is important to me for some reason. That's not advice as to how anyone should play, that is just a fact about my style. The issue in this thread is definitely not heat...not to take away from your legitimate point.

I was under the impression higher BC decreased false positives/negatives. I thought that a 100% BC meant that the count always is a perfect indicator of the deck composition as it relates to player advantage/disadvantage; a 99% BC indicates that there is a 1% probability that the value of the deck composition does not correlate to the count. You seem to be saying it means something else altogether, but I am not clear as to what. I am certainly open to hear explanations, because I have read some stuff on the subject that goes way above my head, so I know that my understanding of it could be flawed (probably is flawed). I am also not sure how you could have a count that is consistently wrong by a certain %...why not just compensate for it, so it is always correct?

Quote: Sonuvabish

I was under the impression higher BC decreased false positives/negatives. I thought that a 100% BC meant that the count always is a perfect indicator of the deck composition as it relates to player advantage/disadvantage; a 99% BC indicates that there is a 1% probability that the value of the deck composition does not correlate to the count. You seem to be saying it means something else altogether, but I am not clear as to what. I am certainly open to hear explanations, because I have read some stuff on the subject that goes way above my head, so I know that my understanding of it could be flawed (probably is flawed). I am also not sure how you could have a count that is consistently wrong by a certain %...why not just compensate for it, so it is always correct?

I'm not 100% sure as to what is being calculated by "bet correlation". You can consider a count to be a mapping from deck composition to a number (which can then be converted to an edge). There is also a function which maps deck compositions to the actual edges. I assume that BC refers to the statistical correlations between those two functions, but I'm not 100% sure because I've never seen it written out that way (or I've glossed over it)

Assuming that this is the case, then a correlation of 1 would indicate that your count always exactly determines the edge. But a correlation of, say, 0.96, could occur in 2 different ways. The edge from your count could always be a little bit off from the actual edge, but never by very much. I would contend that this doesn't matter much, because, as I said above, I don't particularly care if my edge is 2% or 2.2% -- it won't change anything with my betting.

On the other hand, you could get the same 0.96 correlation with a count that always gives you the exact right edge except for a few rare cases where it is way off, and this can be bad. I guess, once the correlation gets high enough, I'm more interested in reducing the number and/or size of big errors, than I am with reducing the size of of frequent small errors.

Quote: AxiomOfChoice

I'm not 100% sure as to what is being calculated by "bet correlation". You can consider a count to be a mapping from deck composition to a number (which can then be converted to an edge). There is also a function which maps deck compositions to the actual edges. I assume that BC refers to the statistical correlations between those two functions, but I'm not 100% sure because I've never seen it written out that way (or I've glossed over it)

Assuming that this is the case, then a correlation of 1 would indicate that your count always exactly determines the edge. But a correlation of, say, 0.96, could occur in 2 different ways. The edge from your count could always be a little bit off from the actual edge, but never by very much. I would contend that this doesn't matter much, because, as I said above, I don't particularly care if my edge is 2% or 2.2% -- it won't change anything with my betting.

On the other hand, you could get the same 0.96 correlation with a count that always gives you the exact right edge except for a few rare cases where it is way off, and this can be bad. I guess, once the correlation gets high enough, I'm more interested in reducing the number and/or size of big errors, than I am with reducing the size of of frequent small errors.

Interesting. I didn't look at it from that perspective before. But it seems to me, now anyway, that it would probably happen both ways, not one or the other. Your actual advantage will always depend on neutral cards, and the frequency at which other cards came out. For example, in hi-lo, if all the 9s and 2s came come out before the rest of the cards, your count will be greatly exaggerated--towards the end, you could conceivably be max betting in a negative expectation. So I think that any count could be way off, it's just more likely to be only a little off. I would imagine the likelihood of each error decreases with an increase in BC. I bet everyone is just as fuzzy as me on PE. These concepts are pretty esoteric.