Bk Rvw: The Complete Book of H 'em by Gary Carson

Jim,


I haven't read about Carson's book. I am not really commenting on that. I want to clear some things up about your statements here on simulators. (i don't know what carson uses).


Gary makes the same mistake a number of other intelligent guys with computer simulation experience make. They merely look at their own chances of making the hand they want and set aside what other players figure to have given the particulars of the situation. It is like they assume thier opponents hold random cards which is flat-out wrong.


This is incorrect. Using TTHE, the sim profiles will play whatever hand they have been programmed to play. This includes preprogrammed profiles, and user created profiles. One can also program the flop, the turn, and the river. Properly run simulations (and these are by no means easy to do without experience), are, essentially EV equations.


For example, you are concerned here with flush over flush. This would be fully accounted for in a properly done sim. The profiles would hold the whatever hands they have been programmed to play. Or, if you wish to test a special circumstance, you can give them individual hands of your choosing. To help ensure accuracy, you can test certain boards, and see if the profiles are playing properly. You can simulate, if you choose, say, five different situations, find the results, then weight them mathematically as you would in an EV equation. There are lots of advanced, accurate ways to use simulations.


I am speaking from experience, as I have run many of these simulations. I can, and sometimes do, simulate hands that have been discussed on this forum as the basis for some of my opinions. This option is superior to EV equations in my opinion, as it takes much less time and is not subject to simple math errors.


To be specific, to test whether flush draws can be profitably raised in these situations, I can set it up with any assumptions you choose. What hands do you want your oppoents to have? What board? How many bets in? Now, these things can't easily be programmed, and take some serious profile manipulation, but it can be done. At the very least, one can get good baselines. These things are not easy to do with TTHE, and it takes some trickery to get this them work.


Quite honestly, proper use of simulations is the only way many hold'em questions can be answered, other than extremely tedious mathematical models. These sims are one of the most powerful tools available today.


For example, there is a post on JJ in the small blind on the medium stakes forum. Sklansky, in his book, reckoned its better to call with this holding with 8 limpers, rather than raise.

Simulations of various types show this to be incorrect thinking, under most almost all circumstances. (I haven't bothered with the thread because I prefer that it be widely held that calling is the better play. Only a handful will read this thread.) You can change the opponents hands to whatever you want, change the board to whatever you want. have the JJ checkraise the flop, whatever. Only extreme circumstances make calling better. So what does this mean? It means that you should look closer at the "math and logic" that Sklansky uses. You will find that it hinges on some assumptions that are different. So, in this case, at the very least, you will educate yourself in the matter, and determine what is more correct: The assumptions the sim used, or the assumptions the expert used? Right or wrong, as long as you are thinking and not following blindly, thats gotta be good. Only dimwits would not benefit from such examination.


Back to the flush draw. If the sims show (the advanced sims I am talking about....not just the garden variety out of the box ones), that raising is more profitable than calling, than maybe we need to look at the math closer. Maybe, for example [], there is benefits to raising that have something to do with one of the advanced (fairly unknown) inferences of Morton's theorem. Or maybe not.


I think this post is long enough. Sorry for the long post, you probably aren't really that interested in a simulator debate.


Take care,


Regards.

one of your very best...you have a cadre of loyal lurkers...keep em coming...gl

"Quite honestly, proper use of simulations is the only way many hold'em questions can be answered, other than extremely tedious mathematical models. These sims are one of the most powerful tools available today.


For example, there is a post on JJ in the small blind on the medium stakes forum. Sklansky, in his book, reckoned its better to call with this holding with 8 limpers, rather than raise.

Simulations of various types show this to be incorrect thinking, under most almost all circumstances"


We have gone over this in detail before and I don't wish to get into another debate, but I strongly disagree. You can only be correct if the TTH sims do a good job of representing real hold 'em. They don't. This must be the case because their playing strategy is so poor and this is easily seen because the program is extremely easy to beat and some of the advisor advice is pathetic. In our opinion it overvalues big pairs and high cards and misses many extra bets the other hands are able to get, especially when they are played from late position.


Since we have gone through this at such detail before this will be my only post on the matter. While I view the program as a good tool for a beginner who is just trying to get a feel for how poker is played, to use it as a serious tool to answer questions should prove to be very costly in the long run.

Thank you for your thoughtful reply about simulations and in all fairness I have never run a simulation so I apologize for unintentionally "maligning" them or any one who runs them. However, I have a few questions and observations to make:


1. In his Chapter on "Playing a Draw", Gary states:


"Five players had called the blind by the time the action got to our here with the 9s-7s. Should he call, fold, or raise? If you are taking an odds perspective to the game, then you need to think about a raise in this situation. There is a popular misconception, repeated in many poker books, that you should only raise if your hand figures to be the best hand. That is not how to get the money in holdem. You should raise whenever your chances of winning are greater than the odds you are getting on the raise. Based on Turbo Texas Hold'Em simulations, 9s-7s from late position against a field of five to seven players, will win over 20% of the time. That makes a hand about a 4-to-1 underdog. With five players already having called, the two blinds yet to act, you will get between 5-to-1 and 7-to-1 on a raise. If you know the blinds will call a raise, then there is profit in a raise and you should take that profit."


I have a couple of questions for you:


1. Where does the 20% estimate of winning chances come from? Does it assume your opponents hold random cards? If someone has a hand like Ks-Js they will limp in 100% of the time. If they have a hand like 8c-2h they will virtually never limp in.


2. Does the 20% assume you go all the way to the river? Suppose you flop a backdoor flush-draw and get bet out of the hand? If you flop a straight draw, does it assume you go all the way to the river to maximize your chances of winning regardless of the board and the betting?


Based on Gary's narrative I sense that the 20% is based on you getting to take 9s-7s all the way to the river regardless of what the flop or turn brings and regardless of the betting action. If so, I question the realism of these assumptions.


It is interesting that a person can test whatever assumptions they want but the results you get are a reflection of the realism of those assumptions. Without knowing the assumptions Gary used, I have to question the validity of his results.


Gary thinks calling is okay but raising might be better. But suppose raising drives out the blinds who would call? Suppose raising results in someone deciding to reraise? What happens when you have to pay two or three bets to take a flop instead of one bet? What is the impact on your implied odds? When you make a pot very large like this, it may be harder to get your hand to hold up when you hit your flush since players will go for runner-runner and so forth. A guy with a lone As might well decide to take off a card even if only one spade is flopped since he has a backdoor and the pot is large. In my opinion, raising is a high variance play that adds very little to your expectation if anything. It may even hurt your EV instead of helping it. I think calling is better.


I also had some serious problems with Gary's flop and turn analyses of this particular problem. He seems to think that adding more players makes this a raising hand when it flops a draw. But why would I want to eliminate players in a large field when I am drawing? I will not eliminate anyone with a better hand or a better draw (a better flush-draw that is). I will eliminate players who may have been contributing dead money when I hit. His method of counting "outs" is bizarre. In poker, there is no reward for simply improving. "Outs" than don't win and don't get you out simply burn up more money. How is this considered in a simulation?

"Quite honestly, proper use of simulations is the only way many hold'em questions can be answered, other than extremely tedious mathematical models. These sims are one of the most powerful tools available today."


I'm not sure that the models always have to be extremely tedious (but they can be and often are). Computers can also help with the math models too. With that said I suspect that many use neither models nor sims.

Regarding the issue of being on the button with a pocket pair against eight limpers. Sklansky says to raise with aces and kings because of the strength of your hand. He says to raise with tens or nines because you are getting the right odds to flop a set. He recommends just limping in with queens and jacks.


If you raise with queens or jacks and undercards flop, it may get checked around to you. Many of these guys like to "check to the raiser" since they see other people play this way. You bet with your overpair and now everyone is getting great odds to call. But suppose you just limp with queens and jacks and undercards flop? Now one of your numerous opponents is likely to bet. This allows you the chance to raise thereby forcing everyone else at the table to call two bets cold which may fold some of them. I wonder how the merits of this play could ever be correctly simulated.

We have gone over this in detail before and I don't wish to get into another debate, but I strongly disagree.


I don't either. But don't sell things like that JJ conclusions are based on "math and logic" when the real issues aren't the math and logic but what you are assuming about the opponents, how they play, etc.


You can only be correct if the TTH sims do a good job of representing real hold 'em. They don't.


This is roughly equivalent to stating, "EV equations can only be correct if they represent real hold'em. The don't."


This must be the case because their playing strategy is so poor and this is easily seen because the program is extremely easy to beat and some of the advisor advice is pathetic.


True. This is the source of much of the ignorance on this matter.

I have alluded to some of the reasons for this in my above post. It is true that many simulations are not being run properly, and the basic unmodified sims bring about false conclusions on occassion. but it is incorrect to say they are wrong. What they are doing is simulating bad play. This is no different than assuming such a thing in an EV equation. They are not wrong. The assumptions are wrong. The only way they can be wrong if there is an internal software error, or the operator did something wrong in the setup.


In our opinion it overvalues big pairs and high cards and misses many extra bets the other hands are able to get, especially when they are played from late position.


As it stands with the unmodified sims, I believe you have landed on the right answer. I might at a later date if I am so inclined detail why this is so and how to account for it.


When Sklansky posted his UTG EV estimate, mine was almost identical. My estimates were based mainly on modified simulators.


This debate will end soon enough, as many of us are collecting and processing online documentation.


Since we have gone through this at such detail before this will be my only post on the matter.


Yes, it is tedious. I don't blame you.


While I view the program as a good tool for a beginner who is just trying to get a feel for how poker is played


It has limited use in that regard.


to use it as a serious tool to answer questions should prove to be very costly in the long run.


Quite the opposite. Those who use it, and understand it, will have a growing edge in the years to come.

Jim,


It would be inappropriate to attempt to tell you what Gary Carson did. But I will try to answer with that caveat. It is unfortunate that he doesn't post here. (why not post this on RGP?)


I have a couple of questions for you:


1. Where does the 20% estimate of winning chances come from? Does it assume your opponents hold random cards? If someone has a hand like Ks-Js they will limp in 100% of the time. If they have a hand like 8c-2h they will virtually never limp in.


He also, from your post, uses the term, "from 5 to 7 players". Suffice to say, that there's a big difference between five and seven players. I just ran a quick showdown sim that has 97s winning 18% of the time against random hands. (7 handed)


Again, it would be wrong to comment on what Carson did, and what context those statements are in without knowing more.


I don't like challenging him (or anyone else) unfairly knowing he can't respond here.


2. Does the 20% assume you go all the way to the river?


If its a showdown sim, it would just run the cards to the end without betting.


TTHE has both showdown sims and full action sims. The full action sims play out the hand according to their programming. This includes betting/raising/folding and the whole gamut of poker moves that can occur. Most of the generic sims don't play that well. this is the cause of some of the naysayers' incorrect conclusions.


It is interesting that a person can test whatever assumptions they want but the results you get are a reflection of the realism of those assumptions


You understand this well. Its exactly what the assumptions are. The other day you posted a hand (JTs vs Q9s), and you had many responders. Coilean, for example, used some Ev equations for part of the analysis. Does that mean that analysis is infallible? Certainly not. IF the assumptions were wrong (like how often you would be raised on the turn, or how often the turn would get checked behind with a better hand, for starters, this could change things). But those EV equations provide solid baselines. So, too, can TTHE sims.


Now, don't let me leave you with the idea that you can test all your assumptions easily. Many you can't, with the present design of the simulator. Often we are forced to set up it in ways that are quite creative and time consuming to get the desired results. I just wanted to express that just because in some situations the sim profiles play badly, doesn't mean they play badly in controlled situations. It also doesn't mean we can't intelligently account for this mathematically.


In my opinion, raising is a high variance play that adds very little to your expectation if anything. It may even hurt your EV instead of helping it. I think calling is better.


To be honest, I am not quite sure. I am not prepared to give an opinion on this without doing the whole series of controlled sims that I would be comfortable drawing conclusions from. Online hand data can't answer this question because you almost never get that type of limping online (nor live, really).


But why would I want to eliminate players in a large field when I am drawing? I will not eliminate anyone with a better hand or a better draw (a better flush-draw that is). I will eliminate players who may have been contributing dead money when I hit.


I haven't read the book, as I said. I don't want to be unfair to Carson or to the truth of the matter.


There may be some value in eliminate dependant higher flush draws(I.e. backdoor aces/kings) (there was a great obscure post about similar issues by "small caps scott" ages ago). This may have some application here.


How is this considered in a simulation?


In the showdown sim (no betting) it wouldn't exactly, although higher flush draws would be accounted for in the %. I agree that there is problems with this.


But in a full action simulations the players have regular hands suited to what their position is. It doesn't simulate perfectly or even close to perfectly what real play is (without controlling the sim), but it provides useful baselines.


One problem i am having here is that it took me over a year to figure this out. I often would run sims as I was playing online, so it was not costing me time. Over this time, I learned how to work the sims to maximize their value, not just accept the dogma of the naysayers.


Hey, I hope you realize I got into a big kuffuffle (sp?) about defending a weak unsuited king in the big blind, because of an article by a certain long time twoplustwo poster/writer.


Regards.