Baccarat card counting

I just read the wizards site on baccarat and noticed the part relating to the effect of card removal. It is stated that card counting is impossible in baccarat because the situation come up too rarely. I was thinking about the effect of card removal and thought if the shoe was strongly biased one way or the other (in favour of player or banker) then that implies that it is more likely that the cards favourable to that side have been played already leaving more cards favourable to the other side left in the deck. I thought I would test this hypothesis by analyzing the wixard's 1000 baccarat shoes available on his site, using a cutoff of more than 54% with 10 hands remaining. However using this data only gave me a sample size of 1200 hands which is much too small to be useful.

Has anyone else investigated this idea before? Is there any possibility it may work? (I don't think it will but I wanted to test anyway)
Does anyone know of a baccarat simulator i can use to cook up some more shoes and potentially do the analysis? (It took me 12 hours to compile all the data from the wizards 1000 shoes and I cant be bothered spending considerable more time on what may be a wild goose chase)

Quote: allin127

Has anyone else investigated this idea before? Is there any possibility it may work?

Yes to the first question. No to the second. I recently published Card Counting in Baccarat on the subject.

I've also examined card counting commission-free variants including EZ-Baccarat and 6-push. There's nothing there.

Baccarat side bets are a whole different story.

In relation to comission free variants, my local casino in perth, australia offers a form of baccarat where the banker pays even money except when it wins on six in wich case it pays half. I thought that it might be possible to devise a counting system to do with reducing the frequency at which the bank wins on 6. This basically revolves around betting on the bank when there are less 6's in the pack as this is the card most associated with the bank getting 6. However after writing a simple computer programme to analyse this I found that the positive effect of making 6 less was roughly cancelled out by the reduced frequency of the bank winning when a six is removed from the deck.
At my casino the only side bets offered are the pairs (which i found to be uncountable) and the dragon bonus. I havent investigated counting the dragon bonus. Is that feasible?

I wonder if some system of betting when the deck is low in 6 7 8 9 (the cards that favour making naturals and stand hands) could favour the dragon bonus?

Quote: allin127

In relation to commission free variants, my local casino in Perth, Australia offers a form of baccarat where the banker pays even money except when it wins on six in which case it pays half.

I examined this game, as I mentioned above, computed the EOR's and modeled a counting system against it. The point of attack is to determine when the Banker is unlikely to get a 6, then bet on Banker. I could find no non-trivial edge for the player against this variation. I am primarily interested in these sorts of things from a game-protection perspective, but certainly understand and appreciate there are two equally valid points of view.

what are your thoughts on the dragon bonus bet? (a bet that pays a different amount depending on the difference between the player and banker totals after 3 cards)

Quote: allin127

what are your thoughts on the dragon bonus bet?

I haven't looked at it.

Quote: allin127

Has anyone else investigated this idea before? Is there any possibility it may work? (I don't think it will but I wanted to test anyway)
Does anyone know of a baccarat simulator i can use to cook up some more shoes and potentially do the analysis? (It took me 12 hours to compile all the data from the wizards 1000 shoes and I cant be bothered spending considerable more time on what may be a wild goose chase)

Well done for your creative thinking and hard work. That will pay off for you in the end. But, no, there isn't a plausible way to get an edge on the bank/player bets.

To make money at baccarat from counting requires several things:

1) Very deep penetration.
2) Almost computer-perfect analysis of the last few hands.
3) A huge bet spread.

The vast majority of the profit comes from the tie, not the main wagers.

Doing this is very difficult. Where I disagree with Griffin, Wizard et al is that it is not impossible. Difficult and impossible are as night and day to an advantage player. Much better conditions exist and have existed than is generally understood.

There are several factors which make counting more practical that are never discussed publicly and most gambling experts are not aware of:

1) What happens in the early part of the shoe indicates to some extent what will happen in the latter. If a lot of tens are dealt out early, it becomes significantly less likely that you'll get a good bet at the end of the shoe. In such situations, it is a good to go find a fresh pack, which significantly increases your hourly returns.
2) Rebates are pretty common for high-rollers at baccarat. Rebates don't normally provide much of an edge, but counting combined with rebate-exploitation is a whole different animal. There's more EV in this than almost anything because of the huge potential absolute returns due to the high maximums.

Quote: allin127

what are your thoughts on the dragon bonus bet? (a bet that pays a different amount depending on the difference between the player and banker totals after 3 cards)

edit:
the Dragon Bonus (by ShuffleMaster) IS listed over at Discount Gambling.

Rapid Baccarat Dragon Bet (on Player Hand)

Quote: allin127

what are your thoughts on the dragon bonus bet? (a bet that pays a different amount depending on the difference between the player and banker totals after 3 cards)

That bet IS countable

Baccarat is an interesting example of a game whose rules have changed as a direct result of countability. When they first started offering Baccarat in Vegas, there wasn't a Tie side bet, but there was a Banker Natural Nine side bet instead (pre-Vegas Baccarat games only offered Player and Banker bets with no side bet).

Ed Thorp wrote a paper on the exploitable nature of the obsolete side bet, published in 1966 as "A favorable side bet in Nevada baccarat" (Journal of the American Statistical Association, Vol. 61, No. 314; June 1966). Wizard's page on the Dragon Bonus side bet suggests that there are casinos that are offering Banker Natural Nine as the Dragon Bonus, reintroducing this countable side bet that was deliberately removed from the game years ago.

There is a (very) old Sports Illustrated article about Thorp's experiences counting cards at Baccarat. He apparently assembled a team that was ~$5000 in profit after a week (that being the non-inflation-adjusted figure from half a century ago), at which point the casino slipped a mickey in his drink and he went all rubbery.

Unless you can find a place offering that exact side bet, though, there have been lengthy academic papers that have concluded that Baccarat isn't countable enough to make a significant difference to your EV (it may be slightly countable towards the end of the shoe, but the modulo 10 valuation still renders it too unpredictable to be of any real utility). I've always thought that something more interesting to explore might be shuffle tracking of the unusual shuffle at big table Baccarat. It is the longest, slowest, and most visible shuffle of any game in any casino, as far as I'm aware. It takes ten minutes and you see the cards coming out of the packs so you have at least a vague idea of what's where, if you have a good familiarity with decks of cards (like maybe if you're a magician or something). Just something that I've thought about but never seen discussed.

I should probably note, though, that the taxpayers of Canada have recently funded a paper entitled "Three-person baccarat" (by Sherry Judah and William T. Ziemba) which makes the suggestion that "It is optimal for the players to collude and place indentical bets in which case the game becomes a Bernoulli trial with banker's edge of 0.82%." I'm not sure that I follow the logic, but it seems like a significant reduction in house edge. It does require team play, which makes it seem more questionable to me.

There is a (very) old Sports Illustrated article about Thorp's experiences counting cards at Baccarat. He apparently assembled a team that was ~$5000 in profit after a week (that being the non-inflation-adjusted figure from half a century ago), at which point the casino slipped a mickey in his drink and he went all rubbery.

Thankyou for being one of a very few people online capable of citing this reference correctly.

Unless you can find a place offering that exact side bet, though, there have been lengthy academic papers that have concluded that Baccarat isn't countable enough to make a significant difference to your EV (it may be slightly countable towards the end of the shoe, but the modulo 10 valuation still renders it too unpredictable to be of any real utility).

Yes and no. You can't design a simple linear point-count system to beat baccarat. That's what those papers prove, using much the same approach every time. Bear in mind also that a) computers were very primitive when most of the seminal articles when written b) the relationship between variance and EV was not fully understood c) blackjack games were of vastly higher quality than nowadays.

However, you can note down the exact composition of the cards in games with deep penetration. It is possible, through extensive practice, to recognize advantage subsets. With the larger advantages it is usually very obvious when you have a great subset (>50% edge). You need excellent penetration to make this work, penetration which theorists generally don't know exists. I have played in 100% pen games for example.

This post goes into more detail:

http://forumserver.twoplustwo.com/31/other-gambling-games/favourable-baccarat-subsets-commentary-long-236618/

Unless you can find a place offering that exact side bet, though, there have been lengthy academic papers that have concluded that Baccarat isn't countable enough to make a significant difference to your EV (it may be slightly countable towards the end of the shoe, but the modulo 10 valuation still renders it too unpredictable to be of any real utility). I've always thought that something more interesting to explore might be shuffle tracking of the unusual shuffle at big table Baccarat. It is the longest, slowest, and most visible shuffle of any game in any casino, as far as I'm aware. It takes ten minutes and you see the cards coming out of the packs so you have at least a vague idea of what's where, if you have a good familiarity with decks of cards (like maybe if you're a magician or something). Just something that I've thought about but never seen discussed.

Not quite sure what you mean here. The wash or csm or whatever usually eliminates most of the non-randomness in a big table shuffle, though there are exceptions. So classic ST or card location, as used at blackjack doesn't work.

I think what you are talking about here is something else. You are not trying to model the shuffle, you are trying to locate cards exposed in the shuffle process? That's not shuffle-tracking as such.
I wrote about it this in my book "Baccarat for the clueless". There's a difficult with using imperfect information like this at baccarat. No single card is that important. Stick an ace at random in a BJ hand and it massive impacts the player's advantage. There's no card like that at baccarat. The information would be most useful when situational and would require some knowledge of the card composition.

Quote: GBV

There is a (very) old Sports Illustrated article about Thorp's experiences counting cards at Baccarat. He apparently assembled a team that was ~$5000 in profit after a week (that being the non-inflation-adjusted figure from half a century ago), at which point the casino slipped a mickey in his drink and he went all rubbery.

Thankyou for being one of a very few people online capable of citing this reference correctly.

My apologies. Should have been Life magazine, March 27, 1964. My ageing mind had conflated it with an article from SI's January 13, 1964 issue ("Bye Bye Blackjack"), which simply reported on the winnings of Thorp's Baccarat team. But, hey, I went to the trouble of looking the academic papers up before I posted.

Also, I should probably note that I never intended to say that shuffle tracking could be potentially profitable; only wondered why I'd never seen it proposed (especially with people claiming to shuffle-track CSMs and whatnot).

I'm kind of interested in what you're saying about deep penetration. I had always been under the impression that even the slight countability that might occur towards the end of a shoe was negated by the cut card.

I haven't seen it mentioned yet, but the two bets on EZ Baccarat are very easily countable (Panda 8 and Dragon 7). In fact, CET shut down ALL bets one night after they discovered a card counting team in Vegas tearing up the tables on the Dragon 7. EVERY CASINO in the company couldn't offer the bet until they could figure out a solution. The answer: max bet $25. This policy seems to be in place a lot of places now, though I've still seen the side bet offered at $100 as well.

Panda 8 Bet

Dragon 7 Bet

My apologies. Should have been Life magazine, March 27, 1964. My ageing mind had conflated it with an article from SI's January 13, 1964 issue ("Bye Bye Blackjack"), which simply reported on the winnings of Thorp's Baccarat team. But, hey, I went to the trouble of looking the academic papers up before I posted.

You misunderstood me. I was actually complementing you on getting the primary source (the Thorp/Walden JASA paper) correct, and describing it properly.

It amazes me how many people never bothered to read that paper given you can just go to a library and look it up. For example: There's a Q&A on the Wizard Of Odds page where someone asks the Wizard about the paper, and the possibility of a library visit seems to elude him, so he asks his friend Don Schlesinger, who tries to remember back fifty years and gets the details completely wrong.

This comedy of errors actually pales in comparison to James Grosjean's work on baccarat which repeatedly cites the paper in various writings, as "proving" that the baccarat tie bet cannot be beaten. As you know, the tie bet didn't exist when the paper was written.

There are numerous other instances of misquotation and misattribution, so I was enormously impressed you managed to get it right.


Also, I should probably note that I never intended to say that shuffle tracking could be potentially profitable; only wondered why I'd never seen it proposed (especially with people claiming to shuffle-track CSMs and whatnot).

It's a strong idea actually, I just thought you'd be interested to know it wasn't a completely unknown concept.

I'm kind of interested in what you're saying about deep penetration. I had always been under the impression that even the slight countability that might occur towards the end of a shoe was negated by the cut card.

It often is, but as with blackjack game conditions vary considerably and there are various ways to make this fly.

I am 100% certain that Thorp's "A favorable side bet in Nevada baccarat" (Journal of the American Statistical Association, Vol. 61, No. 314; June 1966) is available in its entirety for free online somewhere. That's where I learned about the Tie bet replacing Banker Natural Nine, as Thorp had personally observed it changed between two of his team's visits (and had never seen a Tie bet prior to the trip where they had changed it), suggesting that he had personally led to the death of the side bet in a very short amount of time. But you're right about libraries. There are university libraries available to the public that have journals that aren't available in any online databases (at least yet).

But, talking about that paper naturally brings us back to my Dragon Bonus question. Wizard's page on the Dragon Bonus has odds for Banker Natural Nine as one of the possibilities for the Dragon Bonus. I don't know if he was going for completeness and just did odds for everything the Dragon Bonus could possibly be, or if the table represents hands that have been actually been offered or are being offered as the Dragon Bonus somewhere. If it's the latter, and some place does offer Banker Natural Nine as the Dragon Bonus, the Thorp paper would be back in play.

Your reference to penetration made me wonder whether, if I were doing the whole elaborate Macau scoring system, I could get away with asking them to play past the cut card for superstitious reasons. Then I remembered that I've made this exact request at actual casinos and never been allowed one or two more hands out of the shoe. (But it's been years since I've tried.)

I didn't realize that Grosjean was doing Baccarat research. He was hole carding at carnival games and going through elaborate legal nightmares last I was aware. Glad to hear that he's found something more fun to do with himself.