Oh, I understand that the actual data recorded will vary, even over a large sample. But it won't vary based on the toughness or stakes of the game you are playing. And when he used the term "often", to me that implies statistical frequency, not actual data.

Here's the problem, why would you assume this,

I agree. But this is not what you've been talking about.

In this thread, you've been talking about playing additional hands that you wouldn't play in a tough, and since you don't play them in a tough game, you're saying that if you did play them (in a tough game) your expectation would be less than zero.

Mason

Actually, recently we've been talking about hand (or hand) that I think are profitable even in a tough game that you think are not profitable in an easy game. And my argument is that if AJo can be a small winner in really tough games then folding it UTG in soft live games is far more meaningful error than you think it is

Okay. Change it to ATo or some other weaker hand. In this thread we were talking about whether it's correct to play some weaker hands up front compared to a tougher game. In Post #70 you wrote:

So the question has been whether it's correct to play these additional hands that you don't play in a tough game. It doesn't matter what the hand is. David and I have always said that if you play in such a way that you beat the tough games, you'll also beat the easy games.

Mason

An additional problem is that each game is slightly different. So even if you are able to extrapolate your online data to live games, you still have the problem of figuring out which live lineups it is profitable in. If KQo breaks even over a huge sample of all 40-80 live games, there are going to be lineups where it is slightly +EV and ones where it is slightly -EV. Ultimately I think it's going to be impossible to prove whether a certain borderline hand is profitable in an exact lineup (unless you know it is profitable in a tougher game through a large online sample. But I wouldn't consider these borderline hands).

The ability to fine tune these kind of things is probably what separates the best players from the merely great ones. Fortunately, in most live games there are people playing so far away from optimal that getting your ranges even approximately right should be very profitable.

It matters when I can look up a individual hand in a database and see how it performed over a massive sample. All online games were not tough (or you have an action player in BB that will defend 100% vs UTG raise), so over the course of a large enough sample you should have plenty of instances of opening KQo, ATo UTG and you can see how it did. Addiotnally, I can look at other players better than me (and a little laggier pre) who opening those hands and see how they did collectively. .

If you think that AJo is profitable in a tough game, then it should be more profitable in a loose game. I never said what you imply.

However, just because AJo is profitable in a tough game in general, it doesn't mean that it is profitable UTG in the tough game. That's a different issue. But if you think it's also profitable UTG in a tough game then it should be more profitable UTG in a soft game. I have no problem with this logic.

The main discussion in this thread occurred because we were trying to decide if playing certain hands UTG , 77, ATo, KJs, which I don't play UTG, can be profitable in soft games.

MM

Okay. When you can do this and you believe that it's a good data base I agree that matters a lot. But make sure you're looking at it by position and/or other factors that may be important to the particular situation. So what's the point?

By the way, and this is for the readers of this thread, the data base should come from good players, not from run of the mill players.

Okay.

This is a good point. And this is what we call Bayesian statistics where you bring other information into play. However, just because they do well overall, it doesn't mean that they did well on this particular hand in this particular situation.

Also, in HPFAP we do mention playing fast which is where some players will play hands that they believe have slightly negative expectation because they also believe it will help them to win more on other hands. So lots of complex stuff can go into making these decisions and I'm a believer in keeping things simple since by keeping things simple you're less likely to make other errors. But I have no problem with this idea in the hands of a knowledgeable expert.

MM

Jon is almost definitely talking about his own db

Hi Crazy:

This is a good point and is something I talk about in my Gambling Theory book in terms of bankroll requirements.

This is another argument for why good statisticians error on the conservative side.

I don't agree. Even if you're correct, the difference in the win rates should still be small and the large short term luck factor would distort things so that you would never know. So why bother.

I agree. And this goes back to the idea that in statistical distributions your minimums and maximums tend to be broad.

Excellent post.

Best wishes,
Mason

Going back to the original post, if you were to fold jj that would leave you with a super narrow 4b range (4 hands I guess?) I think you have to add jj,10-10, AQs, and in resteal/loose raiser situations add hands like 88, 99, AQo, AJs, KQs.

Sure we can know. Like we have already started if we have a giant database that tells us which hands are profitable and we know these hands will be more profitable in a live setting than the short term luck factor is irreverent.

Now that's impossible to do now, but fortunate for those of us that actually did bother to get better and take the time to create and analyze them

Also I don't know what you consider small but I would estimate these hands to win around .15-.2 bb/hand in mistakes live games (only from UTG). i consider that to be significant.

Doug,

Thank you! Best poker compliment I've gotten in a long long time! Mason, you are probably playing too tight in the BB vs a SB open.

90% is too tight?

No. CrazyLord is talking about the standard deviation of a difference between two players, and it's a different formula from the standard deviation of a result. Specifically you square the standard deviation of each player's result, add them together and then take the square root. This produces a larger number which then has to be compared to the difference between the two players for the hand in question, so I doubt that accurate information would be available even if each player had a large database of his own results.

I agree that creating your own database and analyzing it has a lot of value.

The word significant has a specific statistical meaning and I don't know if your data would have a large enough sample size to tell if a .2bb/hand result is really different from zero. And it's even less likely you have enough data to tell if a .2bb/hand difference between two players (whose results are both based on sample data) on some particular hand, which I believe is what CrazyLord was talking about, exists.

Also, I just found this on the Internet if you want to compute your estimated standard deviation:

http://www.calculator.net/standard-d...alculator.html

Take one of your hands where you think your estimate is approximately .2bb/hand and see what your standard deviation is. Then if .2 - 3sd > 0 I'll be convinced that your result is really positive. But this is still different from the standard deviation of the difference between the results of two players.

Mason

You realize were talking about sample sizes in the hundred or even thousands right

Yes I do. And do you realize that you're looking at very small differences when the average result for each individual hand is quite large.

Let me explain what I mean by that. For a hand where your average result is approximately .2 BB/hand, you probably have some hands in your results where you won over 10 big bets or lost over five, yet you're looking at a mean difference of .2 BB/hand. So to reduce the standard deviation to a small enough value to show that your result is actually greater than zero you're going to need a lot of hands.

I suggest the following. For some particular hand where your average result is around .2 BB/hand post all results for each hand you played. Then we can run it through this calculator:

http://www.calculator.net/standard-d...alculator.html

and we'll see what we have. This should be a worthwhile exercise for anyone with a large database.

Here's an example I'll just make up. Suppose you have ten data points and your results are as follows:

2, 4, 3, 5, 7, -2, -4, -5, -4, -4

This is a sample of ten and it adds to +2 or a mean of +0.2BB/hand

Now using the calculator, we get a standard deviation of 4.5 and using the formula of 0.2 -(3)(4.5) we see that this number is no where close to being greater than zero. So a sample size of 10 is way too small and you can easily need a sample size in the hundreds if not larger.

Mason

Yes. Probably.

Can you guys clarify if you are talking about 6 max or 9 handed?

Winner.

Some of us might or might not have done the grind of many tables of mid-stakes FR games in order to achieve milestones on Stars. This may (or may not) have lead to having sadly large FR samples. Was Stars 10 handed, even? I don't recall. Were I to bring up that computer, I might have 10K-20K hands UTG FR. Real SNE grinders might well have 200K-500K samples. I can think of one person who made SNE multiple years on the FR grind, if you really wanted big samples. Those games don't exist any more, but if we're willing to accept that no live game is as tough when it comes to getting correctly 3 bet, these big samples might show evidence that some hand (say AJo) is wildly profitable even in those games and thus should never be folded.

Freteloo had some interesting thoughts along those lines. He set out to intuit his own hand ranges PF, by using data and some common sense. His argument was about using common sense and hands that run close in value to smooth out variations in smaller samples. So you know before the fact that AQo > AJo > ATo (but close to each other). In your 50 hands of each in EP, AQo is a tiny loser and both AJo and ATo are decently profitable. Rather than waiting for 5K samples to start smoothing things out, you either assume (or look for evidence) that you ran slightly bad with AQ and move on to believing all 3 hands are profitable.

With bigger samples, I think you could look at the worst hands you play in each category. If they are slightly unprofitable, you have evidence that you're playing too many hands. If all of your worst hands are still clearly profitable, there are likely nearby hands that you're incorrectly folding. I think you can also use UTG+1 data to help bolster UTG, with some common sense. Again, small samples mean you can't just look at raw data. Applying "common sense" allows bias to creep in. However, the method of actively considering ranges based on data shouldn't require 50K samples of every hand value in every position before drawing any conclusion.

I'll guess that ~1BB/100 winners who always played AJo UTG in "tough" games did fine with the hand. Well, we get a given hand UTG 1/1000 times. So for every 1K sample of a given hand in a given position, we need 1 million hands. With some supporting data of equivalent hands and nearby position, you could live with less... How big a sample do you want?

You said this regarding AQo and AJs facing the 9 handed utg open raise: Yet you claim that you fold these hands, only 3 betting with 99+, AK, AQs. My question is why would you fold if the hand is ev neutral? Seems to me that ev neutral hands can only enhance the profitability of the rest of your range.

Story time:

A few years ago, I found myself focusing too much on winning money and not having as much fun as I did in years past. So I put fun at the top of my list of motivations for playing that year. If I had what I thought was a close decision between raising or folding preflop, I raised. If I had what I thought was a close decision between folding or calling postflop, I called, and so on.

I had more fun than ever that year at the poker table, and I even won a little money.

Now, winning money is back at the top of my list, with having fun being a close second. However, I'll always look back with nothing but good feelings about that year that I had more fun than ever at the poker table.

Mason, I think it is a bad idea to tout "winning in tough Holdem games" very highly. The author's advice was fine, certainly would help the average player, but his databases that are included and show the profitability of hands came from his play on very soft European sports betting sites that happened to have poker games. Of course there were some strong players as well but he was primarily playing in what would be a berry patch at any stakes.

More importantly the same author was later outed and shunned from these forums for multi accounting, short stacking, rat holing, and colluding with his friend in NL Holdem games on poker stars and full tilt.

I just love Limit Hold'em...
Great thread, thx guys !

Yeah even tho this profitability can be small due to the short term luck factor.
But the majority of bad players tend to still play NL with even bigger leaks instead of playing in a game where the edge are way smaller and losing less rapidly.
Pretty depressing...