Can the count be too good?

Over quite a few years of counting I've noticed what appears to be a trend when the count becomes very high. My win rate actually seems to drop off. It seems as if the deck becomes so rich that a high percentage of hands push, and anything less than a 20 is an automatic loser. I know the dealer is supposed to bust more often at higher counts, but it seems I'm usually the one who ends up busting because I have to act first. I've looked at my recent stats in CVBJ, and over about 5000 hands (I know that's not a big sample) my win rate charted against the bet amount (proxy for the count) is a bell curve. I have a slight loss at 1 & 2 units, a moderate win at 5 units, and a healthy win at 10 units. Then at 15 units my win rate drops off significantly, and I actually show a loss at 20 units. I know this can just be variance, and that in 5000 hands very few will be at these very high counts. But it got me thinking that most of what I've read about effects of removal and the true count focus on TC in the range of -5 to +5. So is it possible different dynamics come into play at, say TC +10 or more? Just curious if anyone has run (or seen) a simulation of very high counts.

I'm using REKO, and making count dependent adjustments to BS (Illus 18) if that matters.

The higher the count, the more pushes you'll get, because you'll get many more 20 vs 20's. Have you had a decent sample size in the TC 10+ range? With few hands, you can obviously have a wide range of results. Just because you have a 2-3% edge (or whatever it is at +10)... doesn't mean you'll have a 2-3% WinRate at that count range over a few hundred hands.

I don't know how many hands in each range you need to get a good sample size, but I'm thinking with an overall sample of 5k hands, you're probably nowhere near the sample size needed for that high TC range.

No, you're right. The sample size at a very high count is too low to mean anything. And common sense tells me what I'm seeing is an anomaly (aka, variance). I'd think that with all the research done on BJ any change in the curve at very high counts would be well known and documented. It's just an interesting trend I've seen both in live play and in practice, and I don't recall reading anything that talked about extremely high (or low) counts specifically. But it's also been about 15 years since I've done any serious reading on BJ.

I wish I had a better feel for how often the count gets that high. I play shoe games, and the count can swing pretty wildly at times. I've had shoes that start very positive, swing negative mid-shoe, and then back to strong positive before the cut card. I've also had the shoe end after a long run with a strong count, to the point where I'm thinking there must be nothing but 10s left behind the cut. Bottom line is I know big swings and streaks aren't just possible, they're common. FYI, I'm using REKO which uses an unbalanced, running count, so it's a bit difficult to correlate TC.

Just to lay out the scenario that prompted this question . . . I was practicing on CVBJ last night and got into a very strong count, +34 RC. For those unfamiliar with REKO and as a point of reference, I place my max bet at +24. I drew stiffs and busted over the next 3 hands, and the count rose to +38. Then I pushed a BJ, and pushed 2 consecutive 20s. Then I busted a stiff, as did the dealer. Then another 20 push. Then my 20 lost to dealer BJ. End of shoe. So, betting 15 units per hand I had 5 losses (for a total of 75 units), 4 pushes, and no wins. Glad it was just practice.

Quote: Rio481

...I know the dealer is supposed to bust more often at higher counts...

This is where rookies (not saying you are) get somewhat misled. While true that they bust more often, it's ONLY when faced with a stiff which happens LESS often.

Quote: Ibeatyouraces

This is where rookies (not saying you are) get somewhat misled. While true that they bust more often, it's ONLY when faced with a stiff which happens LESS often.

Yep, absolutely right. The dealer won't get a stiff hand as often, but is far more likely to bust when it does happen.

What I'm wondering is if there's some loss of advantage at very high counts due to the fact that the player has to act first when both player and dealer have a stiff. While acting first is always the case, it seems it might be more "deadly" at high positive counts.

I understand that stiff vs stiff shouldn't happen often at high counts, and that in roughly half those cases the dealer will show a stiff card allowing the player to stand. But my experiences at the table and in practice got me wondering if there's something else at play with these extreme counts. As per my last post, I suspect it's just low sample size and variance . . .

Also if the count is high enough ,you will be surrendering or standing on some of those hands.

Quote: Hunterhill

Also if the count is high enough ,you will be surrendering or standing on some of those hands.

Most of the games I play (low mins) no longer offer surrender, but good point.

This is a great thread that really highlights why a HUGE COUNT doesn't mean "instant winnings" that a lot of newbies that come to counting think. They see the movie 21 and go "Oh, it's a huge count, here comes my BJ and/or the dealer bust!" The movies don't show that the dealer can get 20/21 all the same as you... but counters leverage 4 rules for the money, not the count:

1) You get paid 150% on BJ's, the dealer does not.
2) You can double, putting more money on the table in good situations. The dealer can not.
3) You can split, putting more money on the table in good situations. The dealer can not.
4) The dealer will bust slightly more often, but as stated, when presented with a "bustable" situation (which will be less).

Quote: Romes

. . . They see the movie 21 and go "Oh, it's a huge count, here comes my BJ and/or the dealer bust!" The movies don't show that the dealer can get 20/21 all the same as you . . .

Interesting you brought this up. I recently watched (for maybe the 10th time) the documentary on the MIT team - I think it's a really interesting story. At one point Semyon talks about winning something like $75K on a single hand. But he goes on to say his winnings for the trip were about $150K. Don't know if those numbers are right, but they're in the ballpark. So half his winnings for a 3-4 day trip came on 1 hand. Toward the end of the show they talk about the team's fall and the mounting losses. Some attribute that to the players being burned out and getting sloppy, but it could have also been simply the ebbs and tides of the game. Remember when they were winning early they were well above EV.

I think there are a few lessons here:
1) Even with the best counters playing as a team, the advantage is very small and only occurs in a minority of hands.
2) To achieve a statistical advantage you need to maximize every opportunity (per your 4 rules).
3) Even when the count is strong and you're playing perfectly, luck (variance) plays a role.
4) To win, you need to rely on the law of large (actually HUGE) numbers. No one (even the MIT team or the house for that matter) wins consistently.

Quote: Rio481

Interesting you brought this up. I recently watched (for maybe the 10th time) the documentary on the MIT team - I think it's a really interesting story. At one point Semyon talks about winning something like $75K on a single hand. But he goes on to say his winnings for the trip were about $150K. Don't know if those numbers are right, but they're in the ballpark. So half his winnings for a 3-4 day trip came on 1 hand. Toward the end of the show they talk about the team's fall and the mounting losses. Some attribute that to the players being burned out and getting sloppy, but it could have also been simply the ebbs and tides of the game. Remember when they were winning early they were well above EV.

I think there are a few lessons here:
1) Even with the best counters playing as a team, the advantage is very small and only occurs in a minority of hands.
2) To achieve a statistical advantage you need to maximize every opportunity (per your 4 rules).
3) Even when the count is strong and you're playing perfectly, luck (variance) plays a role.
4) To win, you need to rely on the law of large (actually HUGE) numbers. No one (even the MIT team or the house for that matter) wins consistently.

Do you know what the documentary is called, I would be interested in watching that.

https://youtu.be/QflVqavHHM0

Quote: PokerGrinder

Do you know what the documentary is called, I would be interested in watching that.

Breaking Vegas: The True Story of the MIT Blackjack Team

I'm having trouble pasting a link, but you can find it on YouTube if you search "MIT Blackjack Team".

Quote: Romes

1) You get paid 150% on BJ's, the dealer does not.
2) You can double, putting more money on the table in good situations. The dealer can not.
3) You can split, putting more money on the table in good situations. The dealer can not.
4) The dealer will bust slightly more often, but as stated, when presented with a "bustable" situation (which will be less).

Actually Romes, this brings up another thought. If the count is extremely high I would think there would be far fewer opportunities to double and split. Fewer player hands that would permit doubling/splitting, and more likely the dealer shows a card you don't want to double/split against (A, 10). So does that possibly eat away at the player edge? Or is it a wash with play adjustments such as doubling A,8 v 6 and 10 v 10, and splitting 10s (which I don't do - too much of a tell that I'm counting). And more frequent BJs help as well.

Before anyone starts accusing me of being a math atheist, I'm not in any way suggesting anyone change the way they play and I'm not changing my play. I have every confidence in the statistical probabilities and variances, and recognize what I'm seeing is very likely to be just variance. I just like to understand the logic behind blackjack. I think when I was first learning BS it helped significantly to understand WHY to make certain plays, not just memorize the tables. And the dynamics of deck/shoe composition are interesting to me. (OK, so I'm a geek!)

Quote: Rio481

Breaking Vegas: The True Story of the MIT Blackjack Team

I'm having trouble pasting a link, but you can find it on YouTube if you search "MIT Blackjack Team".

Thank you Rio481

Quote: Rio481

Actually Romes, this brings up another thought. If the count is extremely high I would think there would be far fewer opportunities to double and split. Fewer player hands that would permit doubling/splitting, and more likely the dealer shows a card you don't want to double/split against (A, 10). So does that possibly eat away at the player edge? Or is it a wash with play adjustments such as doubling A,8 v 6 and 10 v 10, and splitting 10s (which I don't do - too much of a tell that I'm counting). And more frequent BJs help as well...

Let's take a look at what it really means to have a true count. If you have a TC of +10 (monster, by your account and mine) what does that mean? Well, let's break it down to just a single deck of cards. With a TC of +10 that just means there are 10 bigger cards than little cards in the deck. What does that concentration look like?

Single Deck (ORIGINAL):
20 little (2-6) cards
12 medium (7-9) cards
20 high (10-A) cards

Single Deck (TC +10)
10 little
12 medium
20 high

So there are still 22 cards that are "little" or "medium" as opposed to the 20 "high" cards remaining. This means slightly more than 50% of the cards are still NOT "high" cards. Every other card, on average, would be a little/medium card.

Now let's apply that math to a regular 6D game...
6 Deck (TC + 10)
110 little
72 medium
120 high

So when you get a hand next hand, there's 302 cards in the shoe, of which 120 are high.. This means 60% of the cards are still NOT "high" cards (here's another example of how multiple decks slightly hurt the player). You're not going to see only 10's come out of the deck, not by a long shot. There are still plenty of little cards for splitting and doubling opportunities, and this doesn't even take special cases in to account... Splitting 10's to a 4, 5, or 6... Doubling A-8 or A-9 in a monster count (which would be 1 high card and still a double). So you see, you don't even need little cards to split/double. That doesn't even account for other index plays such as splitting 2-2 v 8, 3-3 v 8, 4-4 v 4, etc that come from index plays.

In general, you might have slightly less opportunities to split/double (I think you're exaggerating the lessened amount), but when you do they'll be worth a little more and to boot you'll get a few more opportunities from your index plays of splits/doubles.

For it to remain a TC of +10, it'd have to be 15-12-25, not 10-12-20. Of course, many other combinations exist, as there does not need to be 12 middle cards per deck, either. It could be 21-0-31. Or 0-42-10. Or anywhere else really.

10-12-20's TC would be 10/(42/52) = 12.38.

This is mentioned in Peter Griffins theory of blackjack and also In Don's Blackjack Attack. If I remember correctly once you have more than 70% of the remaining deck comprised of tens the edge starts to go the other way.I think optimal ten density was in the 40 something % range.

Quote: RS

...10-12-20's TC would be 10/(42/52) = 12.38.

Keep it up RS... Soon enough I'll join Axel in hoping you're stuck $5k chasing a $2k royal as well.

Quote: Romes

So when you get a hand next hand, there's 302 cards in the shoe, of which 120 are high.. This means 60% of the cards are still NOT "high" cards (here's another example of how multiple decks slightly hurt the player). You're not going to see only 10's come out of the deck, not by a long shot. There are still plenty of little cards for splitting and doubling opportunities, and this doesn't even take special cases in to account... Splitting 10's to a 4, 5, or 6... Doubling A-8 or A-9 in a monster count (which would be 1 high card and still a double). So you see, you don't even need little cards to split/double. That doesn't even account for other index plays such as splitting 2-2 v 8, 3-3 v 8, 4-4 v 4, etc that come from index plays.

In general, you might have slightly less opportunities to split/double (I think you're exaggerating the lessened amount), but when you do they'll be worth a little more and to boot you'll get a few more opportunities from your index plays of splits/doubles.

The index play of splitting 2-2 v 8 got me thinking. Isn't this really a matter of minimizing losses? I would think that at a TC of +10 one would have their max bet (or close to it) out. Is it truly worth the risk of doubling your max bet just to shave off a tiny percentage of potential loss? Getting a 2-2 against an 8 with a TC of +10 would be a very bad thing in my mind. Even though the index play is to split I wouldn't really consider this an "opportunity". Perhaps just hitting and getting your max bet out on the next hand would be a wise move. Just thinking out loud here...

I agree with that specific situation Vlads, I was just spitting some splits/doubles off the top of my head to show you will split/double "slightly" more with some indexes.

Quote: Romes

In general, you might have slightly less opportunities to split/double (I think you're exaggerating the lessened amount), but when you do they'll be worth a little more and to boot you'll get a few more opportunities from your index plays of splits/doubles.

Thanks for the analysis Romes.

I did a quick search online and found a relevant article. Unfortunately I can't post the link. As I understand it I don't yet have enough posts in this forum to be allowed to do so - I'll fix that : ) The article is from . It shows the edge for TC -25 to TC +25 based on a simulation of 25 billion hands.

There were several things I found interesting.
1) Edge does increase all the way up to +25. However, at the extremes (-13, +12) the results get very "noisy" (i.e. a lot of variance). I suspect this may be what I'm seeing in my results.
2) Edge actually levels off from about -10 to -16.
3) The effect of using index plays is dramatic once you get into -5 or +5 territory.
4) Due to the frequency with which each count occurs, most of your $ winnings will happen at +4.

I've probably peaked at that article before if it's from where I think it is. Send me the link in a PM and I'll post it if you want.

Quote: Rio481

If the count is extremely high I would think there would be far fewer opportunities to double and split.

You will get lots more opportunities to split. Not that you should.

Quote: DRich

You will get lots more opportunities to split. Not that you should.

Ha! Good point. Maybe I should have said GOOD opportunities to split. And no, I'm not splitting 10s, especially against a 10 or A.

Assuming you stick to the normal strategy (I know it's a false assumption but it keeps the sim simple) then these are the figures (for UK rules) based on 492M hands - in this table 0 means between -0.29 and +0.20, etc. I've left out the end tails (-28.4 to -17.5) and (21 to +29.4) as the figures start to be inconsistent!

Count	Hands	EV
-17.0	0.002%	- 0.121 867
-16.5	0.002%	- 0.140 705
-16.0	0.003%	- 0.123 667
-15.5	0.005%	- 0.128 209
-15.0	0.007%	- 0.106 820
-14.5	0.007%	- 0.113 165
-14.0	0.011%	- 0.100 872
-13.5	0.015%	- 0.099 522
-13.0	0.019%	- 0.089 479
-12.5	0.024%	- 0.084 017
-12.0	0.033%	- 0.082 744
-11.5	0.041%	- 0.082 336
-11.0	0.058%	- 0.072 431
-10.5	0.073%	- 0.070 035
-10.0	0.085%	- 0.064 683
-9.5	0.120%	- 0.063 574
-9.0	0.138%	- 0.055 325
-8.5	0.202%	- 0.051 788
-8.0	0.240%	- 0.048 770
-7.5	0.293%	- 0.045 892
-7.0	0.395%	- 0.042 735
-6.5	0.488%	- 0.040 128
-6.0	0.610%	- 0.035 898
-5.5	0.749%	- 0.032 309
-5.0	1.043%	- 0.030 490
-4.5	1.268%	- 0.027 228
-4.0	1.638%	- 0.024 517
-3.5	2.053%	- 0.021 692
-3.0	2.718%	- 0.019 192
-2.5	3.471%	- 0.016 114
-2.0	4.596%	- 0.013 471
-1.5	6.012%	- 0.011 062
-1.0	8.206%	- 0.008 524
-0.5	10.408%	- 0.006 769
0.0	14.081%	- 0.004 084
0.5	9.807%	- 0.001 593
1.0	7.132%	0.000 874
1.5	5.482%	0.003 317
2.0	4.100%	0.005 728
2.5	3.167%	0.008 428
3.0	2.413%	0.010 601
3.5	1.912%	0.012 534
4.0	1.467%	0.013 825
4.5	1.148%	0.016 369
5.0	0.919%	0.018 083
5.5	0.730%	0.021 369
6.0	0.541%	0.022 317
6.5	0.453%	0.025 036
7.0	0.351%	0.027 138
7.5	0.272%	0.027 641
8.0	0.227%	0.030 025
8.5	0.169%	0.030 939
9.0	0.138%	0.034 252
9.5	0.105%	0.034 535
10.0	0.084%	0.038 440
10.5	0.065%	0.038 165
11.0	0.044%	0.039 368
11.5	0.041%	0.041 406
12.0	0.028%	0.042 966
12.5	0.023%	0.041 616
13.0	0.016%	0.046 306
13.5	0.013%	0.046 003
14.0	0.009%	0.046 282
14.5	0.008%	0.047 294
15.0	0.005%	0.049 982
15.5	0.004%	0.039 309
16.0	0.002%	0.050 033
16.5	0.002%	0.053 153
17.0	0.001%	0.056 162
17.5	0.001%	0.070 581
18.0	0.001%	0.043 091
18.5	0.000%	0.081 920
19.0	0.000%	0.050 075
19.5	0.000%	0.107 579
20.0	0.000%	0.030 933
20.5	0.000%	0.052 496

Here's the article in reference. As thought, it points to qfit =): https://www.qfit.com/cardcountingindexes.htm