Baccarat Card Counting
I first saw when the Wiz wrote about it, where he treats it more like BJ in the effects of each individual card removal, but here it's presented into subsets that may occur deep into a shoe. An excerpt from twoplustwo.com:
The link to the forum and thread is: http://forumserver.twoplustwo.com/31/other-gambling-games/favourable-baccarat-subsets-commentary-long-236618/Quote:thought I would publish some figures concerning the expectation from the baccarat tie wager with depleted decks. This should be helpful in determining the viability or otherwise of counting baccarat.
Card subset name: 10's + 5's
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 0 0 0 0 32 0 0 0 0
Tie Bet Advantage = 340.30936%
Card subset name: 10's + 6's + 4's
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 0 0 0 32 0 32 0 0 0
Tie Bet Advantage = 152.01169%
Card subset name: 10's + 8's + 9's
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 0 0 0 0 0 0 0 32 32
Tie Bet Advantage = 88.93232%
Card subset name: 10's + 6's + 7's + 8's
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 0 0 0 0 0 32 32 32 0
Tie Bet Advantage = 47.61894%
Card subset name: 10's + 3's + 6's + 9's
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 32 0 32 0 0 32 0 0 32
Tie Bet Advantage = 8.21853%
Card subset name: all-even valued cards
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 0 32 0 32 0 32 0 32 0
Tie Bet advantage =62.02323%
Card subset name: all-even valued cards+5's
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 0 32 0 32 32 32 0 32 0
Tie Bet Advantage = 6.25045%
Viability and Cover
It should be apparent to the perceptive from the above examples that baccarat can be "beaten" as a practical matter, though not neccessarily easily, at least, in games with deep penetration. If, that is, we have a good means of determining when we have an advantage, on which more below.
Simulation data on computer-perfect analysis of the tie wager shows that if you can find a game where the last hand will be dealt from a 10-card subset, you can obtain a 1% advantage with a 45-1 spread, assuming you can detect favourable situations accurately.
We can therefore deduce that the game can be beaten because baccarat, as a general rule, has nothing like the heat associated with blackjack. You can spread up to the table maximum and down again, indeed you can sit out hands, no one will care. As David Sklansky wrote in Getting The Best Of It "There is not yet any paranoia among the casinos regarding counting...Players can bet anywhere from $5 up to $50,000 at any time. " ($50,000 is actually pretty conservative for bac nowadays). Obviously, if you could only make one $50,000 bet a year with a 100% edge, your EV for that year is still $50,000.
Naturally, the combination of large bets, an absence of heat, and huge advantages on individual hands, can make the game very profitable.
The main difficulty, and this is an issue complex enough to require a separate thread, is the relative frequency of advantageous situations. Advantageous subsets like the ones above, occur quite rarely. You need to know exactly when it is worthwhile to quit a game for a fresh deal in order to maximize your opportunities.
Additionally, it should be obvious that baccarat is not desirable for all conceivable player utilities. It would be a poor choice for a player trying to build up a bankroll. For other utilities it can be quite attractive-for example, some high stakes teams who are looking for an unconventional approach find that the absolute returns to be had from baccarat can be quite attractive.
Failure Of The Linear Systems
Note that in every case in the subset data above, count systems, such as the one devised by the Wizard Of Odds (https://wizardofodds.com/baccarat/baccaratapx2.html) , would get a completely incorrect assessment of advantage, often recommending a negative bet when your edge is enormous. Don't take my word for it, test the Wizard's system on these subsets yourself.
The problem with the Wizard's system and all other existing baccarat count systems is due to an assumption that the basic structure of the pack remains intact, an assumption which becomes very inaccurate at deep penetration when multiple ranks are missing.
Note that other analysis of baccarat count systems (eg Thorp's Mathematics Of Gambling, Griffin's Theory Of Blackjack) predated modern computing power and so would have only been able to create non-linear systems with the available technology.
The Relationship Between Ranks Remaining And Advantage
A theoretical comment: note that the primary factor in determining advantage on the tie wager is the number of ranks remaining. The absolute minimum number of ranks remaining, one, would naturally yield an advantage of 800% on the tie, the largest possible advantage.
Looking at the above examples tie advantage decreases with the number of ranks remaining. With only tens and one rank remaining your advantage will be in the hundreds of %, with two still generally over 100%, under 100% with three, and becoming increasingly marginal or non-existent with four or five.
However, the relationship between the ranks remaining is also important, so a system based solely on remaining ranks would likely not be viable by itself. To win consistently requires a highly accurate knowledge of the relationship between ranks in depleted decks. It is not a simple matter to do this and requires a lot of tedious rote memorization.
The last two examples show how the addition or subtraction of one rank can make an enormous difference to your expected value. In the "all-evens" subset case for example, advantage is over 60%. With the restoration of the 5's to the pack, the advantage is reduced to a miserly 6%.
Casino City Times also has some info on it: http://may.casinocitytimes.com/article/card-counting-at-baccarat-1149
Has anyone else looked into it? Anything worth trying? Looks to me that it could be very favorable IF the casino had a very minimal burn off the top (I think the norm is 14), and the enevn hit more frequent than the 10,000 decisions.
-TheArchitect
Quote:Card subset name: all-even valued cards+5's
Card
Value 0 1 2 3 4 5 6 7 8 9
Number 128 0 32 0 32 32 32 0 32 0
is negligible. It's the remotest of possibilities, so if you just sit around and wait for one of those favorable decks you're going to be waiting a very long time. I wholly disagree that "It should be apparent to the perceptive from the above examples that baccarat can be "beaten" as a practical matter". Those examples are so contrived and rare that no practical strategy can be derived through knowing about them.
Let me put it another way. Suppose you did know you had a 10% advantage on the tie bet, but you only got that advantage one out of every million hands. That's knowledge relegates the scenario to an academic curiosity, not a practical advantage play. The reason card counting in blackjack is effective is because the shoe is frequently player-positive. The shoe in baccarat is almost never player-positive.
Quote: MathExtremistSure, but the probability of a depleted deck having even the last distribution he cites,
is negligible. It's the remotest of possibilities, so if you just sit around and wait for one of those favorable decks you're going to be waiting a very long time. I wholly disagree that "It should be apparent to the perceptive from the above examples that baccarat can be "beaten" as a practical matter". Those examples are so contrived and rare that no practical strategy can be derived through knowing about them.
Let me put it another way. Suppose you did know you had a 10% advantage on the tie bet, but you only got that advantage one out of every million hands. That's knowledge relegates the scenario to an academic curiosity, not a practical advantage play. The reason card counting in blackjack is effective is because the shoe is frequently player-positive. The shoe in baccarat is almost never player-positive.
I guess another question is: What is the probability of any +EV subset occuring?
Quote: TheArchitectI guess another question is: What is the probability of any +EV subset occuring?
The Wizard addresses that topic here.
I've seen that appendix MathExtremist, but isn't it for linear card counts (like Black Jack). I agree that baccarat cannot be counted for any intents and purposed that way, but it seems that a very high +EV bet may arise very rarely, but perhaps with enough probability and payoff to make the game beatable.Quote: MathExtremistThe Wizard addresses that topic here.
I concede that I do not have the skills in finding such probability, and every attempt I make in trying fails horribly. How would I go about trying to find such probability in this case? Identify all subsets at card remaining levels that would produce +EV situations, Identify the card removal advantage or disadvantage for each subset, apply a count to such cards.
-TheArchitect
However, based on my simulations, it wasn't a practical use of time, at least my time. Maybe you would have more patience than me to sit at the tables for hours, waiting to make one bet.
Quote: WizardLet me give you a tip. If you must study beating baccarat by card counting, look at the pair bets. If you know two ranks are completely eliminated you would have an advantage. Maybe one rank, depending on how the others were distributed.
However, based on my simulations, it wasn't a practical use of time, at least my time. Maybe you would have more patience than me to sit at the tables for hours, waiting to make one bet.
I agree Wiz that this would be both a mind and butt numbing way to gain a +EV. However, so is video poker (and that's also eye numbing and lifeless). If a high +EV bet were to arise once every few thousand decisions, then it would certainly be worth it. If it were to arise once every few thousand shoes, then of course it's not doable in any real life scenario.
I'll look into the pairs bet as requested, but judging as you don't think it's worth the time, I will take any findings I come across with a grain of salt. I value your word and expertise.
I'm also going to go out on a limb and venture those dream subsets John May came up with (as much as a guaranteed tie win), as occurring so infrequent that they may never come up in a lifetime.
-TheArchitect
Quote: TheArchitectIf a high +EV bet were to arise once every few thousand decisions, then it would certainly be worth it.
Really, though? Suppose I offered a hypothetical game with a rate of 60 hands/hour, 1% house edge, but an identifiable 1/1000 chance of having the next bet be +15% player favorable. That's an *enormous* player edge.
So once every 16 2/3 hours, you have a +15% wager. How is that "worth it" by any practical measure? You're not just going to sit there and make your one bet once/day, are you? A $5000 bet has an EV of +750, which is a great daily win rate, but are you really able to bet $5000 once a day, either (a) within the limits of your bankroll and variance, and (b) without the casino telling you to make bets more often than once/day or vacate the chair?
Quote: MathExtremistReally, though? Suppose I offered a hypothetical game with a rate of 60 hands/hour, 1% house edge, but an identifiable 1/1000 chance of having the next bet be +15% player favorable. That's an *enormous* player edge.
So once every 16 2/3 hours, you have a +15% wager. How is that "worth it" by any practical measure? You're not just going to sit there and make your one bet once/day, are you? A $5000 bet has an EV of +750, which is a great daily win rate, but are you really able to bet $5000 once a day, either (a) within the limits of your bankroll and variance, and (b) without the casino telling you to make bets more often than once/day or vacate the chair?
I understand where the both of you are coming from, and the short answer is no. But if there's a will there's a way. team play could certainly speed things up by spreading multiple tables and signaling when such an event arises.
With your hypothetical game, you wouldn't flat bet table minimum and lose $0.60 an hour waiting for your +EV payoff?
Not trying to start a war, or even argue my position. Just seeing if it is profitable or not, bottom line.
-TheArchitect
Now it's not going to happen immediately, but the casino will eventually catch on that you're just biding your time until some rare event. Nobody - I mean nobody - gets down a $5000 bet without surveillance and floor management being notified. So now you're on the radar, and if you make your 5k bet three times a week and start winning, the casino is going to get curious. Then, perhaps after some consultation with someone like me (sorry...) they'll figure out what's going on. At that point, the ball is entirely in the casino's court. Except in certain jurisdictions (e.g. New Jersey), the casino can simply toss you out on your ear. They can discriminatorily flat-bet you. They could even institute a non-discriminatory rule that says "players cannot increase their bets by more than 50x between hands", a rule which would even fly in NJ. That'd be a non-issue for normal players but would absolutely kill your advantage.
So yes, a game like that can be profitable, but not without some serious ongoing negligence by the casino.
