Deck Estimation

[assume_R]

How good does everybody think their deck estimation is? Are you able to tell the difference between 1.6 and 1.8 decks of penetration for example?

I'm starting to think I am not as good at it as I thought.

I want to buy some used cards from casinos to practice, but I heard from a pit boss that a lot of casinos stopped selling those hole-punched cards.

 

[Jack_Black]

How good does everybody think their deck estimation is? Are you able to tell the difference between 1.6 and 1.8 decks of penetration for example?

I'm starting to think I am not as good at it as I thought.

I want to buy some used cards from casinos to practice, but I heard from a pit boss that a lot of casinos stopped selling those hole-punched cards.

getting very good at it, as I'm trying to ST in the future. 1.6 vs 1.8 I can nail in a DD game. Piece o cake.

that is bullcrap. what else is the casino gonna do with the cards? I find it highly unlikely that will be adopted industry wide.

 

[Southpaw]

Deck-estimation is definitely something that I think should be practiced regularly (with real cards), as with all of one's other AP skills.

In calculating the TC, I only use full-deck resolution, although I had always practiced 1/2 deck resolution. (I practiced resolving by half-decks, so that I would be more accurately be able to round to the nearest deck for the TC calculation).

However, now that I have been reading the cookbook and practicing some of the ST'ing drills, I realize that the only limit to deck-resolution is what you set your mind to. Using a shoe of cards that are labeled 1-312 (for a 6-deck shoe), there is no reason one cannot learn to resolve to the exact card +/- 2-3 cards. I am not quite this good yet, though I intend to be after practicing more.

How important do I think higher deck resolution is for simply calculating the TC? Minimal, I'd say.

SP

P.S.--The store I play at sells their decks for $1 a piece.

 

[Syph]

Many have argued that it's unimportant provided you are within certain standards. I believe it's half deck estimation for shoes, and quarter deck for pitch games. While the sims may bear this out to be true, I find that the mindset that glosses over this sort of accuracy is lacking in the successful players I have encountered over the years. On a similar note, I remember, once, someone asked Alienated how many indices he knew:

"All of them."

To him, it made no sense that you would adopt a system and not be able to use it to it's full capacity. Certainly the time and effort to apply, and be capable of using all the indices, was out of proportion to their gain (which, admittedly, is quite negligible). Nonetheless, it goes back to your relation to the game. The mindset that learns the entire matrix, that can cut within three cards of her mark, is the same mindset that will lead you away from straight counting, to NRS approaches, to slug location, and more advanced techniques. And perhaps will even lead you away from blackjack altogether.

Of course, it depends on the individual.

Some are quite comfortable with a limited approach, and they will play the game for years comfortable in their haphazard deck estimations, limited index matrix, Wonging, and deep penetration. They've basically stopped after the second week of learning the game. They've stopped progressing as advantage players. And when they do attempt to move on to more advanced techniques, they bring these same habits to the table with them, to little, if any, success. I would encourage you not to follow in their footsteps. It is a toxic mindset. And far too common.

So, let's be a little uncommon:

1) Learn all your indices.
2) Improve your deck estimation skills.
3) Learn your shuffles. Stepladders, R&Rs, plugged, not plugged, etc ..
4) Learn the vulnerabilities of each shuffle.
5) Read this post by Alienated:

(Dead link: http://games.groups.yahoo.com/group/blackjackcardcounterscafe/message/22118)

6) Work through the numbers, with your shuffles, and put the puzzle together. Combine this with the basics you learned when you started playing the game. You know have a skill set that will eclipse 99.9% of all counters out there. And you will be all but invisible to the camera.

And you've only just begun.

All the best,
Syph

 

[assume_R]

Very interesting post, Syph. Seems like more of a mindset than anything, and I will indeed try to take that to heart. Kind of like, why do it only half-hearted. Just because you might only get a few tenths of percentage points, it will compound in this game.

I wanted to run a sim to see if accurate deck estimation is useful. So I ran the following controlled study:

Methods
Both players:

• Penetration: 7/8 decks
• Rules: AC (H17, nRSA, nLS, DAS)
• Speed: 100 rounds / hour
• System: Zen Count w/ Full Indices
• Played: Playall
• Betting: Optimal betting
• Spread: $10-$150 spread
• Bankroll: $10k
• Seat position: Random
• Simulation: 1 billion rounds

Player 1 (perfect deck estimation):

• Resolution: Exact (essentially 1/52 deck resolution)
• Divisor: Knows exactly how many cards have been dealt.
• TC Calculation: Recomputes the TC before every decision
• Deck Estimation: Round to nearest

Player 2 (poor deck estimation):

• Resolution: Only full-deck resolution
• Divisor: Just looks at the discard tray
• TC Calculation: Only recomputes the TC before betting
• Deck Estimation: Round down

Results:
Player 1: (perfect deck estimation)

Player 2: (poor deck estimation)

Conclusions:
Player 1's more accurate deck estimation skills allowed him to be more granular in his bets. The count frequencies changed, because he was able to more accurately determine the difference between a TC of +4 and a TC of +5, and because he rounded to the nearest deck, instead of just truncating his deck estimations.

Hence, he had a lower RoR, a higher WinRate, a higher SCORE, and a lower N0. Player 2's poor deck estimation required his large bets to be placed a bit earlier, and hence his standard deviation went up, and overall results weren't as good as could be.

The question of whether this extra effort is worth it for the minor differences shown can only be answered on an individual basis. I know that I, for one, will be improving my deck estimation if only for my own personal satisfaction. The most important conclusion, however, is that one must run a simulation for his own skills of deck estimation. It makes no sense to run a simulation for perfect deck estimation, if you only have full-deck estimation capabilities, because your bets will no longer be optimal.

 

[blackjack avenger]

Return vs Effort

Good work assume_r
I draw a different/similar conclusion for counting:
The difference between exact and 1 deck estimation is not much. Now, what is considered what a person can do? Half, quarter, eigth or sixteenth deck estimation? The difference between a sixteenth of a deck and deck will be much closer then your sims show. A nickel difference in SCORE? I would venture one would be better off to play more or get a job to build bank vs working on exact TC resolution.

We know being off by 1 on indices does not add much value. So being off on estimation is not that important in regard to hand play.

Our advantage at any given point is just an estimation. Whether you place that correct bet at a 1% or 1.2% advantage does not matter. Many players limit their spread which hurts SCORE more then "sloppy" limited importance deck estimation. If you don't spread your bets then estimation means very little. Do you use betting camo? So that also cuts down on the importance of perfection.

In straight counting deck estimation is not a big deal. With other things then it becomes more important.

Wonging vs ST:
It depends on the game variables:
A 1 table casino with an easy shuffle or NMSE ST should dominate.
A multi table casino with a complex shuffle or ASM (not counting cracking the ASM) wonging should dominate. It depends on your circumstances.

Some other points. I have read some state that using an easier count makes tracking easier, so they start behind what a wonger with a strong count can do and are counting on ST to catch up. Also, ST definately makes play more complicated and you better get it right! Finally, a wonger plays 100 hands faster then an ST because an ST'er has to sit through the shuffle.

In this endeavor, the law of diminishing returns holds true.
With indices one can strive for perfection:
All of them
known to decimal point
hand composition
A side count
shoe depth
EOR
rule differences
I will venture to say no one does all the above.

:joker::whip:

good cards

 

[FLASH1296]

Kudos to Assume. Fine work. Thought-provoking, although I assume that
your readers will conclude that their sloppiness need not be corrected.

For me, always striving to play "better" is the only thing that keeps the game
interesting. After 1/8 million ± x hands "there is nothing new under the sun."

In fact, I focus on Spanish21 because of the greater complexity of the game.

 

[zengrifter]

Many novices get confused between TC 'deck resolution' and 'deck estimate'.

I use a graduated-rounding approach >>

1DTC (ie, 1D resolution)
Remaining Decks \ deck estimate fraction
Over..3 \ 1D
Under 3 \ 1/2D
Under 1.5 \ 1/4D

-----------
Example -

RC = 8
remaining decks / TC
--------
8 / 1
7 / 1
6 / 1
5 / 1.5
4 / 2
3 / 2.5
--------
2.5 / 3
2 / 4
1.5 / 5
--------
1.25 / 6.5
1 / 8
.75 / 12
.50 / 16​

 

[blackjack avenger]

Ummmmm

Don't many of us use counts that require no TC conversions? Don't they have an error rate?
:joker::whip:
good cards

 

[zengrifter]

Don't many of us use counts that require no TC conversions? Don't they have an error rate?

I don't know the error rate, but UB'd RC systems are not as accurate as TC except at the pivot. zg

 

[sagefr0g]

to thine own self be true

i'd just say, yeah, knowing all the indices, being able to recognize how many cards have been played to the nth degree, errhh yup that stuff should be performed as best as one can.
just me maybe, well i can't hack it.
so i'm one of those hacks, Syph mentions, i guess, but i work darn hard at it, lol.
so for me, one deck resolution, estimated, I-18, and errh well, not always sure what the count is, but i got an idea, an estimation if you will of the hi-lo richness, poorness of the pack to be dealt. errhh, my simulations, i have no doubt are very accurate, lol. i know what i should be making if i did do stuff as well as my computer can.
oh yeah, and i practice my sloppiness as well. errhh, i do try and at least know what i should be doin .
i believe if i tryed to do everything perfect, i'd, 1.) go nuts and 2.) just plain screw it up.:whip:

 

[blackjack avenger]

Ability VS Sloppiness

Sagefrog touched on it, sloppiness might not be the correct word. Perhaps one's ability to use more indices, better deck estimation etc. should be the issue. Of course practice would improve things up to a point, but if it takes away from time playing it may not be worth it.

 

[Automatic Monkey]

Precision vs. accuracy, you know the drill...

Very interesting post, Syph. Seems like more of a mindset than anything, and I will indeed try to take that to heart. Kind of like, why do it only half-hearted. Just because you might only get a few tenths of percentage points, it will compound in this game.

I wanted to run a sim to see if accurate deck estimation is useful. So I ran the following controlled study:

Methods
Both players:

• Penetration: 7/8 decks
• Rules: AC (H17, nRSA, nLS, DAS)
• Speed: 100 rounds / hour
• System: Zen Count w/ Full Indices
• Played: Playall
• Betting: Optimal betting
• Spread: $10-$150 spread
• Bankroll: $10k
• Seat position: Random
• Simulation: 1 billion rounds

Player 1 (perfect deck estimation):

• Resolution: Exact (essentially 1/52 deck resolution)
• Divisor: Knows exactly how many cards have been dealt.
• TC Calculation: Recomputes the TC before every decision
• Deck Estimation: Round to nearest

Player 2 (poor deck estimation):

• Resolution: Only full-deck resolution
• Divisor: Just looks at the discard tray
• TC Calculation: Only recomputes the TC before betting
• Deck Estimation: Round down

Results:
Player 1: (perfect deck estimation)

Player 2: (poor deck estimation)

Conclusions:
Player 1's more accurate deck estimation skills allowed him to be more granular in his bets. The count frequencies changed, because he was able to more accurately determine the difference between a TC of +4 and a TC of +5, and because he rounded to the nearest deck, instead of just truncating his deck estimations.

Hence, he had a lower RoR, a higher WinRate, a higher SCORE, and a lower N0. Player 2's poor deck estimation required his large bets to be placed a bit earlier, and hence his standard deviation went up, and overall results weren't as good as could be.

The question of whether this extra effort is worth it for the minor differences shown can only be answered on an individual basis. I know that I, for one, will be improving my deck estimation if only for my own personal satisfaction. The most important conclusion, however, is that one must run a simulation for his own skills of deck estimation. It makes no sense to run a simulation for perfect deck estimation, if you only have full-deck estimation capabilities, because your bets will no longer be optimal.

Something to point out- your two players differed in not only deck estimation, but 1) recalculating/not recalculating the TC before making play decisions and 2) rounding/truncating the TC. Because of this, they are differing in not only how precise their data is but in how they are using that data and you end up trying to solve for 3 variables with 1 equation. If you try it again where they differ only on the 1 variable of deck resolution, it might look a little different. As a cross-check you can try two players both with exact deck estimation but differing in rounding and in recalculation interval and see what that does.

Being we are rounding TC to whole numbers (assuming we are indeed doing that) getting too involved with fractions is going to have limited value, for obvious and provable mathematical reasons. One place it will make some difference is down at the end of a shoe, where when the RC is +6, having 2 or 1.6 decks left is going to make some kind of difference in what TC we are declaring, while towards the beginning of the shoe having 6 or 6.4 decks left will not.

Another consideration is the choice of counts- using higher level counts automatically increase the effective TC resolution relative to lower level counts. Being we mostly agree that using High-Low as opposed to a level 2 system is nothing to wring hands over, neither should be deck estimation or TC resolution.

One place I think deck estimation down to the minutia does have a place is in walking around a pit and picking which game to play for good penetration, and in this case 1.6 vs. 1.8 really does make a difference. But that's a relative measurement, and if I showed you a pack of 1.5 decks and another of 1.6 decks you would be able to tell the difference, even if you could tell me to 0.1 decks accuracy how many were there.

Here's the method I use for judging the size of the cutoff: look at it relative to the whole shoe. Picture the shoe cut in half, that's 4 decks, half of that is 2 decks, and half of that is 1 deck. Reckon where the cut card is between that imaginary 2 deck and 1 deck point, with 1/4 deck being around the thickness of a chip. Or just hold 4-8 chips together and look at them along with the cutoff. That should tell you if a dealer is worth playing.

 

[assume_R]

Something to point out- your two players differed in not only deck estimation, but 1) recalculating/not recalculating the TC before making play decisions and 2) rounding/truncating the TC. Because of this, they are differing in not only how precise their data is but in how they are using that data and you end up trying to solve for 3 variables with 1 equation. If you try it again where they differ only on the 1 variable of deck resolution, it might look a little different. As a cross-check you can try two players both with exact deck estimation but differing in rounding and in recalculation interval and see what that does.

Indeed, that is a limitation of the study. I could have just changed each of those 3 parameters individually, and run different sims, to determine exactly which of those variables is most important. But I figured they'd all be somewhat important, and I wanted to compound the effect of a "poor TC calculation player". I will leave it up to the interested reader to run their own study on each of the variables individually to determine their relative importance

 

[blackjack avenger]

Give a Monkey a Bananna, He Pops Your Kneecap

One place it will make some difference is down at the end of a shoe, where when the RC is +6, having 2 or 1.6 decks left is going to make some kind of difference in what TC we are declaring, while towards the beginning of the shoe having 6 or 6.4 decks left will not.

I would think not much difference in indices and in betting only if 1 is spreading their bets, which is not always the case.

Another consideration is the choice of counts- using higher level counts automatically increase the effective TC resolution relative to lower level counts. Being we mostly agree that using High-Low as opposed to a level 2 system is nothing to wring hands over, neither should be deck estimation or TC resolution.

I am glad you did not say level 3 counts or I would have had to get on my hobby horse and I am not biased at all.

One place I think deck estimation down to the minutia does have a place is in walking around a pit and picking which game to play for good penetration, and in this case 1.6 vs. 1.8 really does make a difference. But that's a relative measurement, and if I showed you a pack of 1.5 decks and another of 1.6 decks you would be able to tell the difference, even if you could tell me to 0.1 decks accuracy how many were there.

I agree but I tend to look more at number of players and fast dealers which I can see at a distance.

I agree the 2 different deck estimation methods muddy things. I think consistency between the 2 players would give us a better answer, though the difference won't be great. I also think a comparison using one player using deck and the other using quarter or eigth deck would be more indicative of human ability.

 

[SleightOfHand]

Something to point out- your two players differed in not only deck estimation, but 1) recalculating/not recalculating the TC before making play decisions and 2) rounding/truncating the TC. Because of this, they are differing in not only how precise their data is but in how they are using that data and you end up trying to solve for 3 variables with 1 equation. If you try it again where they differ only on the 1 variable of deck resolution, it might look a little different. As a cross-check you can try two players both with exact deck estimation but differing in rounding and in recalculation interval and see what that does.

Being we are rounding TC to whole numbers (assuming we are indeed doing that) getting too involved with fractions is going to have limited value, for obvious and provable mathematical reasons. One place it will make some difference is down at the end of a shoe, where when the RC is +6, having 2 or 1.6 decks left is going to make some kind of difference in what TC we are declaring, while towards the beginning of the shoe having 6 or 6.4 decks left will not.

Another consideration is the choice of counts- using higher level counts automatically increase the effective TC resolution relative to lower level counts. Being we mostly agree that using High-Low as opposed to a level 2 system is nothing to wring hands over, neither should be deck estimation or TC resolution.

One place I think deck estimation down to the minutia does have a place is in walking around a pit and picking which game to play for good penetration, and in this case 1.6 vs. 1.8 really does make a difference. But that's a relative measurement, and if I showed you a pack of 1.5 decks and another of 1.6 decks you would be able to tell the difference, even if you could tell me to 0.1 decks accuracy how many were there.

Here's the method I use for judging the size of the cutoff: look at it relative to the whole shoe. Picture the shoe cut in half, that's 4 decks, half of that is 2 decks, and half of that is 1 deck. Reckon where the cut card is between that imaginary 2 deck and 1 deck point, with 1/4 deck being around the thickness of a chip. Or just hold 4-8 chips together and look at them along with the cutoff. That should tell you if a dealer is worth playing.

No need to run sims, it has all been done for us by Norm

Changes in SCORE due to indeces
http://blackjackincolor.com/penetration4.htm

Changes in SCORE due to deck estimation skills and rounding style
http://blackjackincolor.com/truecount4.htm

Changes in SCORE due to TC calculation frequency
http://blackjackincolor.com/penetration9.htm

 

[LIB]

While I agree with theoretically negligible difference (how do you disagree with math?!), I think the psychological factor is seldom discussed. Lax approach in one area, say deck estimation, spreads to other areas as well. I noticed from myself that when I cut myself some slack with just about any aspect that needs to be practiced, the attitude spreads even to frequency of practice as well. Infrequent practices also have an effect of shifting frame of reference. what I mean is that instead of the 234 cards mark for 4.5 deck, for the sake of demonstration, shifts up by 5 cards. This per se isn't so bad except that if I were at a borderline TC with # of used cards between 4.5D and 4.75D, I could mess up the next bet amount, insurance decision, and the play decision. That to me sounds more like living on the edge instead of playing with an edge.

I've also noticed the misguiding "overall" impression of skill level when practice records aren't kept meticulously. For example, if at the beginning of deck estimate practice for the day I was reliably accurate, miss a lot during the middle of the practice, and nail the last 12 estimates with unusual accuracy, I get an impression that I got quite good when in fact the record I kept during that practice show otherwise.

 

[blackjack avenger]

Don't Hesitate

instead of the 234 cards mark for 4.5 deck, for the sake of demonstration, shifts up by 5 cards. This per se isn't so bad except that if I were at a borderline TC with # of used cards between 4.5D and 4.75D, I could mess up the next bet amount, insurance decision, and the play decision. That to me sounds more like living on the edge instead of playing with an edge.

Think you are overstating this. These decisions are not lines in the sand, they are shades of areas based on deck depletion and actual deck composition which we don't know.

On bets there is a difference in advantage from 1.1 tc to 1.9 tc yet many probably make the same tc1 bet if they round down.

Do you use any bet camo? This throws concerns over precise betting precision out the window.

It's been shown indices are not lines in the sand, there is not much difference being off by 1. Here is an interesting example:
9 v 7 at 3 is EV indice
9 v 7 at 5 is RA indice
So anywhere from 3 to 5 is "correct".
A stict kelly better would like RA. Someone with a very large bank and having trouble getting bets out would use the EV.

In summation:
Don't sweat close calls, don't waste time thinking about close calls.

:joker::whip:
good cards

 

[Pelerus]

No need to run sims, it has all been done for us by Norm

Changes in SCORE due to indeces
http://blackjackincolor.com/penetration4.htm

Changes in SCORE due to deck estimation skills and rounding style
http://blackjackincolor.com/truecount4.htm

Changes in SCORE due to TC calculation frequency
http://blackjackincolor.com/penetration9.htm

Great links, thanks for posting. I needed a refresher course. It is good to have precise terminology to describe the methods of calculation and estimation you are using, both to make sure that you are being consistent in the application of those methods and to be able to communicate effectively about them with others.

 

[Dopple]

Couldn't you use some kind of marker like rotating a chip on the table to count the rounds and get a very precise number of card played.

Back when I was a young lad I used to try and turn a silver dollar on the table to keep the count so there must be some way to count the rounds.

We count the aces with our feet in the Uston school and there must be a way that does not use up too much disk space (thinking) to count the rounds, from rounds played it is a quick multiplier to get deck you could even come up with the simple equation to translate rounds played in a 2,3 or 4 person setting to decks played.

For the purpose of deck estimation I have often wondered if it might be better for an AP to make certain table player sizes his or her forte and thereby be better attuned to equate rounds played with decks played.

I have created tables based on heads up play that produced the multiplier needed to do the rounds to decks remaining multiplier and in fact took it one simple step further to have it produce the number to hit the RC with from the rounds played. I never got this down but would have liked to.

Sorry it was long no choice.