Pre-flop Theory - want many or few opponents?

[bm259]

For years I’ve been playing live poker basing my Texas Hold’em pre-flop decision generally off Sklansky’s hand groupings. I now find myself playing online with Poker Tracker providing me a wealth of additional information to help make this decision. The more I’ve thought about how to use this additional information the more I have realized how little I really understand about the pre-flop decision process.

For instance, there are many hands like TT and KJs where I’m not really sure if I want to play against many or few players. This hand can win without improving against few opponents but it can also draw to a big hand that can beat many opponents out of a big pot. I find myself in the situation where I have a much better chance of predicting the number of opponents I will play against after I call or raise but not even understanding which situation I prefer. For example, I have TT in early position and my competitive intelligence is telling me that if I call I will get many callers but if I raise I will get very few (let’s also just assume for now that there is little chance of a re-raise). What is the correct play?

There are plenty of hands, such as 44 and 87s, where I know I want to play against many opponents. There are also hands like AJ and QQ where I think I want to play against few opponents. It is all the marginal ones in between that I’m curious about.

One theory I am bouncing around is that these “in between” hands play well against very few opponents and they play well against many opponents but do not play well against a medium number of opponents. On the face of it, a large probability of winning a small pot and a small probability of winning a large pot do not seem fundamentally any better than a medium probability of winning a medium pot.

That said, perhaps the difference lies in the fact that poker is not played “hot and cold” but rather there are bets and raises that I will have to call on the way to the showdown. In the situation where you are playing one of these hands I am describing as an “in between” hand against a medium number of opponents you may not have a medium probability of winning because these hands are easily “scared.” My TT against a medium number of opponents still has the chance of drawing to set (but for a smaller implied pot) but really doesn’t have much chance of winning without improving. The reason for this is that there are likely to be over-cards and almost certainly going to be bets from the medium number of opponents that may scare me into folding the best hand.

So if I am right about the “in between” hands not liking medium numbers of opponents then is my correct strategy to accentuate the situation that I am presented with? That is, if I’m in middle position and everyone folds to me should I raise these hands so I can play against fewer opponents? And, if everyone calls to me should I just call so there are more calls behind me?

If I am on the right track here, how do I know which hands fall into this ‘already made’ or (but not “and”) ‘drawing’ category? Or is it the case that all the hands where it isn’t clear if I should raise or call are situations where the optimal choice is only marginally better than the second best choice?

Finally, does anyone understand how the artificial intelligence community goes about making this pre-flop decision for their bots?

[sthief09]

I used to think playing preflop was easy. After all, if you have a chart, you can just do exactly what it says! Wrong. So much depends on your opponents. If a weak, predictable player who plays any 2 cards limps and you have A9o with tight blinds, this is a great spot to raise. However, if a strong, tricky, tight player open-limped, and the blinds are loose, then you should muck the hand. As simple as preflop strategy sounds, there are so many different situations that you need to be able to recognize the strength of a hand. But I think you realize all this.

KJs is a big "depends" hand for me. Depending on the game, I can play it completely differently. If you can be more specific, I can give you my take on it. For me, TT is a raise from any position after any amount of limpers, unless I'm in a tough game (and I've never actually been in a tough game). It wins more than its fair share. If there are something like 3 limpers to you and you have the button, I think it's pretty neutral whether to raise or call. I usually go with the raise, because I enjoy raising.

In general, I don't think it's necessarily all that important to consider whether you want more or less people when raising from EP. The way I think of it now is "are people going to call me with worse hands?" If the answer is yes, then I'll raise. If people want to cold-call my TT raise with A7o or 65s, that's good for me. The same goes with JJ from the BB. Sure, you won't chase anyone out, but your hand is so much better than everyone else's that you want to push the big edge that you have. I do consider how many people I want when I'm raising in LP. If I have something like A9o in the CO and it's folded to me, I don't want callers, so I'll raise.

As for drawing hands, I tend to play pairs and suited aces anywhere. If you are trying to improve, playing a lot of suited aces will really help your post-flop game. They put you in many difficult situations since you likely have a sub-par kicker. I tend to stear clear of suited connectors. Having them out of position often puts you in a bad situation. In LP, they are nice to have with a lot of people in, but in general I think they tend to leak chips, at least for me (maybe I'm just playing them wrong).

So, if you want, be more specific and I'll try to better answer the question. I'm no expert, but I am pretty good a regurgitating what I've read by people who are.

Good luck.

[miamikid]

[ QUOTE ]
As for drawing hands, I tend to play pairs and suited aces anywhere. If you are trying to improve, playing a lot of suited aces will really help your post-flop game. They put you in many difficult situations since you likely have a sub-par kicker.

[/ QUOTE ]

Hi Thief,
Playing Axs from any position is a bad idea, particularly A8s-A2s. It will not flop hard enough often enough to show a profit. The game conditions would have to be extremely right to play this hand from early position,,, extremely loose passive. You definetely want a lot of callers with these hands. If you routinely play it from early position, you will be losing money, save your bet. There are too many people to act behind you, and you definetely don't want to be isolated with one or two others going into the flop.

Summary: Fold medium to low Axs from early position!

miamikid

[Spyder]

Interesting....

I Play AXs from any position and will even cold-call a raise with it. So far I am making money overall with AXs:

Overall: +0.21 BB/hand
Best Hands: A9, A3 & A8
Worst Hands: A5, A7 & A4

First 4 Positions: +0.12 BB/hand
Best Hands: A9, A8 & A3
Worst Hands: A7, A2 & A5

Last 4 Positions: +0.26 BB/hand
Best Hands: A3, A6 & A2
Worst Hands: A5, A4 & A7

Blinds: +0.60 BB/hand
Best Hands: A2, A8 & A4
Worst Hands: A9, A6 & A3

I am slightly negative (-$0.34 BB/hand) when cold-calling a pre-flop raise with AXs.

The above results were in the $0.50/$1 & $2/$4 tables over about 15,000 hands. When I limp with AXs, I want limpers and I WANT a late raise behind me (I'd rather it not be 3-bet, though, unless a lot of players are along for the ride)...that way I KNOW that I have someone to check-raise when I hit.

It's paying off for me now /images/graemlins/smile.gif

Spyder

[astroglide]

you can't lump them all together because they're different hands. -.034bb/hd is HUGE, and a 15k database is tiny.

[sthief09]

I don't think cold-calling raises with these hands is a good idea in a non-blind without a LOT of people seeing the flop.

And you definitely do NOT want someone raising your limp.

[sthief09]

Playing Axs from any position is a bad idea, particularly A8s-A2s.

no it's not

It will not flop hard enough often enough to show a profit.

yes it will

The game conditions would have to be extremely right to play this hand from early position,,, extremely loose passive.

like Party Poker games? and I contend that looseness doesn't matter very much. aggressiveness matters. I definitely don't mind seeing an unraised flop 2-4 ways with a suited ace, but if there's a good chance it's going to get raised behind me, I won't play.

You definetely want a lot of callers with these hands.

so you wouldn't want a suited ace headsup?

If you routinely play it from early position, you will be losing money, save your bet.

if you aren't routinely playing a hand like A8s from EP in online games, then you're losing money

There are too many people to act behind you, and you definetely don't want to be isolated with one or two others going into the flop.

that's ok, because the games I play are passive. I don't play 15/30 where there's a good chance it'll get raised behind me.

[TXTiger]

"Summary: Fold medium to low Axs from early position!"

I agree with this. Unless you are playing very low limit where there are consistently 4-6 people in for one bet.

[Spyder]

Well, after looking at those results, I've decided to stop cold calling raises with AXs /images/graemlins/smile.gif

Spyder

[Spyder]

[ QUOTE ]

and a 15k database is tiny.


[/ QUOTE ]

You know, it really bugs me when people say things like this. I'm a degreed person in Computer Science with a minor in Math...I KNOW how significant the numbers are in the DataBase. Just because the sample is small doesn't mean you can't derive useful information from it. This kind of comment sums up to be a simple "I'm better than you and you need to get over it." comment. It doesn't help anyone, and it pisses some of us off. Most of us don't like to be talked down to.

Now, on to your constructive comment:

[ QUOTE ]

-.034bb/hd is HUGE


[/ QUOTE ]

As I replied above, after looking this information up, I've decided to quit cold-calling raises with these hands...it might save a buck or two /images/graemlins/wink.gif

Spyder

[sthief09]

[ QUOTE ]
This kind of comment sums up to be a simple "I'm better than you and you need to get over it." comment. It doesn't help anyone, and it pisses some of us off. Most of us don't like to be talked down to.

[/ QUOTE ]


FWIW, I don't think he was talking down to you. He's also right that you shouldn't take stats 15k hands to mean much. How many times have you gotten suited aces? You'd probably need to have them a total of 20,000 times in a certain situation for any conclusion to be statistically significant. For example, you'd have to limp with A6s-A8s in EP something like 20,000 times at a given level, with 10 people in the hand, for this to really prove a point.

[astroglide]

[ QUOTE ]
This kind of comment sums up to be a simple "I'm better than you and you need to get over it."

[/ QUOTE ]

you mean a comment like "I'm a degreed person in Computer Science with a minor in Math...I KNOW how significant the numbers are in the DataBase." ??

15,000 is absolutely nothing for specific suited hands

[Spyder]

Well...I apologize for the rant. Being talked down to is one of my hot-buttons.

If you check through your Sadistices texts you'll find that you only 29 samples from ANY SIZE POPULATION to produce useable data. Now, given, your SD will be larger than you like, but, the data is still useable.

That said, I don't have 29 or more hands of each type in my DB. However, by looking at trends you can see whether you are generally losing or winning from certain positions with certain groups of hands because, collectively, you DO have 29 or more. For instance, when you look at my numbers you see that there is absolutely no correlation between 9 or 2 as the X. This tells us we don't have enough of each hand to make those kinds of determinations (which is why I didn't). However, if you look at them as a whole, or, in larger chunks (say, 9-7 or 2-4) then you can begin to derive information for those groups as your totals reach the magic number of 29.

In the games I'm playing right now, AXs is profitable from any position. That might change when I move up to 3/6 or before then if the 2/4 tables tighten up.

Sorry if I offended folks /images/graemlins/frown.gif

Spyder

[sthief09]

that's not right. we are dealing with a hypothesis here, which requires a confidence interval, which is inversely correlated to the standard deviation. if a player has a winrate (mean) of 2 and a SD of 17, the he'll need a huge number of hands for the winrate to fit in a 95% confidence interval.

you didn't offend anyone... I just wanted to explain to you that astroglide wasn't talking down to you at all. sometimes someone will read something and take it the wrong way, and I think that's what happened.

[Spyder]

Right, but with, say, a 66% to 75% confidence interval, we can begin to see trends and gain information with which we might be able to make changes to our play.

For instance, lets say that we have 20 hands of A3s. 10 hands in early position, 10 in late. If the (small sample of) data showed that we won 9 of 10 in late, and lost 9 of 10 in early, we could begin to assume that playing A3s early might be a bad wager and start watching it more closely, or even change our behavior. However, if, we had additional data, say 20 hands of A4s where we won 9 of 10 early and lost 9 of 10 late, then we'd know that both sets of data were anomalies and were to be disregarded. However, if the A4s hands demonstrated, say win 9 of 10 late and lose 8 of ten early, then the similar data would tend to corroborate the A3s data and we could begin to draw a (guarded) conclusion that A3s (and maybe A4s) were poor wagers early.

At any rate, true data is rarely useless. You just have to understand when you have information and when you have data. A small source of data is rarely information, but, multiple small, related sources of data almost always is.

I apologize, Astroglide...sometimes things just catch me wrong.

Spyder

[Freudian]

I'm also showing profit on Axs, even when coldcalling (+0.8 BB/hand), but I am pretty selective when playing them. I don't like to limp in with them in EP (with the exception of A9s and A8s in a very loose game). A2s is the only one of them I am not winning with.

One thing I noticed though was that I am losing money with them in SB, where I probably play them a bit too often.

[Saborion]

I'm no math expert, far from it. But I do know one thing, and that is that short term luck in poker exists. You say you can begin to see a trend with only 10 hands from a given position. Maybe that is true, in some weird math-way. But I feel that it is NOT true. I mean, after 30.7k hands at the 2/4 limit at Party, I've been dealt AA 133 times, and won 77.44 % (I'm likely to have folded a winner or two due to me being weak post), but I wouldn't be surprised at all to loose with AA 7 times in a row, regardless of position. My worst as I know of, is 6 times in a row. So 10 hands I'd say doesn't mean anything. But what do I know, being the newbie that I am. /images/graemlins/grin.gif

[Spyder]

That's not what I said /images/graemlins/smile.gif

I specifically said that 10 was too few to be sure...at that point it's only data...but, with other, RELATED, groups of small samples we can begin to see information by corroboration, or by grouping. We can also make the determination that we need more data if the sample results are so disparate as to not show any kind of trends. But, this is still information.

Scientifically speaking, luck is a myth. It's all in the probabilities. In all random things with a finite number of different results (ie, a Deck of 52 cards, 6-sided die, etc), over time, each event will occur a predictible number of times. Statistical theory allows us to define that in terms of probabilities, but, you can never predict short-term results because each random event is a brand new test and can fall anywhere along the predicted curve.

At any rate, again, there's no argument that more (good) data is better. However, you CAN make use of small amounts of data if you understand its shortcomings and how to use it.

That's my piece /images/graemlins/smile.gif

Spyder

[t_petrosian]

Exactly...you're talking about the Central Limit Theorem...with 29 hands, you might find you have a expected payoff of $0.30 per hand, with a SD of $0.45 and a confidence level of 45%. That's not useful data. 15,000 hands may or may not speak accurately about certain hands within the overall field...I doubt you've gotten Ax 15,000 times...you may have gotten that 175 times from that database, and each one of those about 10-15 each. That is not nearly enough data to make any decisions about it for your longterm efficacy as a poker player.