I'm pretty sure the chart was supposed to be from LIMIT holdem data where slop that catches some piece of the board can peek at a turn relatively inexpensively compared to NL.

I would expect a real LLSNL graph of this nature to hold the same general shape but with more exaggerated magnitude.

Mpethy your analysis of leaky positions should be quite the eye opener.

The chart came from pokerroom.com s database of the ev of hands...

It's worthwhile to mote that the chart is pre pt4 so BB is not the same as bb... Also, it included limit and block games.


Sent from my PC36100 using 2+2 Forums

Why would knowing the ev of the Sb from a bunch of different players playing differently help someone's game in anyway?

All of this would be pointless

So THIS post (derrived from data on "a bunch of different layers") couldn't help your game in any way?

We have tons of data from online. At Live Low Stakes however we encounter a unique situation frequently where there are multiple limpers with wide ranges presenting us with tempting pot odds. My belief is that it's normally more profitable to ignore the temptation to complete our junk lacking some specific reason.

I understand but there are a million different variables, people play different pre and post. If you figure the ev of the Sb from all these players what exactly will this tell us that will improve our game, besides giving us an arbitrary #?

I suspect because of LOL sample size it would all be somewhat anecdotal. But it would be interesting to look at some basic trends of the aggregate. ie is there a relationship at the aggregate level between tightness/looseness and EV.

I understand your suggestion that player skill and game conditions may be so varied to render any data meaningless.

I can tell you this... from the few sessions I have logged my SB play it has really got me thinking hard about my decision making here. mpethy's comments above reinforce my suspicion that significant gains in winrate can be made by plugging leaks here.

I agree it's a spot where we could plug leaks, but compiling data (as opposed to just analyzing specific hands) is generally going to be useless

<shrug> Whatever the reason, it's way off in terms of NLHE online, even if you look at it as big bets rather than big blinds.

The graph turns out to have been an accidental derail. We just ignore it and carry on from here.

It tells you a TON of things, if you open your eyes to the data.

By comparing your performance to the average, you can figure out whether you are outperforming or underperforming the average. This tells you whether you have leaks to find and plug.

If you compile a big enough sample, you can look at the profitability of individual plays with individual starting hands. Calling a button raise with K8s may not be profitable from the SB, but 3 betting it might be.

In any event, you can construct a calling range that consists only of hands that lose less than folding, on average.

You can do the same thing for your completion range.

You can compare the performance of the top of your completion range to the bottom of your raising range to determine whether hands need to be moved from one category to another.

You can select with precision the hands you should be semi-bluff raising preflop, and the hands you should be 3 betting preflop.

It's best to have a pretty big sample to do this sort of thing. That makes cammando's undertaking in this thread ambitious and difficult, but not off base. He's definitely on the right track in terms of how to compile data and use it.

I've spent several years devising variance control procedures such that I can do this sort of analysis with very small samples of hands, so I'm confident that I could make sense of even a small sample of hands, and draw some valid conclusions. But, of course, the bigger the sample, the better.

This argument is wrong, period.

The whole point here is to extrapolate an average, and then deduce from our relationship to the average the leaks we may or may not have.

For example, suppose that we identify A8o as the worst ragged ace that showed a profit completing in the SB. You say to yourself, "self, you're pretty new to this whole NLHE thing, you're barely above break even. If the average 2+2er can barely play this hand for a profit, perhaps you're better off folding it."

Or you can go the other way: say we identify all suited kings as profitable completing hands. You might say to yourself "self, you play in a pretty fishy game down here in Florida. If K2s is profitable for an average winner in an average game, there's a very good chance that you can be completing even worse suited hands. So you should start completing some of the marginal queens you're currently folding."

What sort of data would you say is relevant for a task like this?

We would need what cammando asked for--abbreviated hand histories for every hand we play out of the small blind. We don't need the details; just "K2s, completed, -1bb;" or "A8o, completed, +22bb."

ETA, including folds, so we'd also want "72o, folded" or "KJo, fold to ep raise"

Soooooo IN.

This thread is awesome. Forget the derail. Do this.

Can we do it for all positions, and create a massive live database? Or just for BB, SB, BTN, CO, as those are the most important per M's rant?

Regardless, I'm in for SB data. Will start this week (I only play ~8hours/week live), so I will be contributing comparatively less than others. Thanks in advance. This could be amazing.

In other words, we are strictly recording hand ranges to develop a fold/call/raise/3-bet chart?

Do you think such chart is useful in live setting?

I can see it being useful in a tough online game where players are more balanced, but in a live setting where game dynamic can vary greatly, how will a "fixed" chart help our game?

It's not a fixed chart. It's an average, from which we deduce how we might need to modify our game to optimize for our skill level/game conditions.

But we're only recording hand and its return in each occurrence.

How can we use such information to apply toward varying game conditions?

Like for example we are somehow able to collect 10,000 samples of A6o and A7o and have come to the conclusion that when we complete SB:

A6o = -0.5bb
A7o = +0.1bb

How will this information help us in different live game conditions?

It's a baseline to start from with using logic to make deductions. We'd know that on average, A7 is a profitable call. So, are we likely above average in skill? If so, everything else being equal, we'd know it was a profitable hand, and that A6o would be for us as well.

Or we'd say, "well, I'd say my edge is smaller than average, so I should definitely be folding A6, but it looks like I can play A7."

But without the data, you're just guessing at what is profitable and what is not, and you're very likely to wrong at the margins (because everybody is wrong about something at the margins. Everybody).

M, any idea what a minimum meaningful sample would be?

But if the data is collecting from all over, can quantity overwrites various biases that are inherently built in?

If so, why not simply use a reliable online database that has already been collected for these adjustments?

If you would be willing to have me analyze the data, I could probably squeak by with 400 hands.

To put it in perspective, If 10 volunteer players recorded 10 4 hour sessions at 3 sb/ hr that would be 1200 SB hands.

That'd be more than enough, but I'd need more detail than what I previously suggested.

If we have a small sample, then we're going to need some information designed to control for variance. That would basically be hand strength information.

So a usable HH would include our starting hand, our action preflop, our hand strength on the flop and at the end of the hand (whether shown down or not).

So something like this:

KJs, called 3bb raise, c/r K hi gutshot, all fold, +10bb

Or: 22, complete, flop set, +12bb

Or: A8o, flop top pair, turn 2 pair, +23bb

For big pots, it'd be useful to know our equity:

AA, raise, 2 callers, all in on flop v. FD, -113bb.

Done. (Requirements detailed).

10 people, 120 SB hands each. With details as M listed. For me this will take until 5/31, because I can't play this week, but should play next week. For the LLSNL, this should be done in a day or two. gogogogo

Friday night's sample

1) K3o, fold to a raise and a reraise
2) Qd2d, fold to a raise
3) 24o, fold to a raise
4) 2h4h, fold to a raise
5) 65o, fold to raise
6) 6To, fold (no raise)
7) Qh4h, fold to a raise
8) 69o, fold (no raise)
9) 28o, fold (no raise)
10) 63o, fold to a raise
11) T2o, fold (no raise)

A little more than 3 hours and never played a hand from the SB.

lol @ 400 hands telling us anything