I haven't been tracking my iso'ing / stealing hands, but I have a sneaky suspicion I'm way down (but that's just a guess).

In my 53 hours for $1944, I've only got 3 sets on the list of AA/KK / pots I put in $100+ and are called (2 wins, 1 loss, for $395 = 20% of my profit).

GreallyhasnoideawhatthatmeansG

As promised, here is my analysis of September.

A lot of what I do here will be controlling for small sample variance. Some of how I do that is math, some is common sense, and I may not show all my work. But it is all based on a tried and true method I developed for analyzing small samples in online databases; an online player might have 100k hands, but he may have only 38 button calls with JTs or whatever, and this is the sort of thing I'd do to analyze those tiny specific samples.

I was sort of goofing off during September before my dad got sick, and flying back east ended my month at only 88 hours. So at 32 hands an hour, I played roughly 2816 hands. I tracked all of my cold calls and all hands that I got dealt AA and KK. I wound up cold calling 57 times and getting AA and KK a total of 33 times.

AA/KK hands: At 2816 hands, I "expect" to get AA and KK 12.8 times each. What I actually got was 13 AA hands and 19 KK hands, so I ran quite a bit above expected frequency with KK.

My achieved win rate with all of those hands combined was $30.16 per hand. It breaks down to $47 hand with AA and $3.10 a hand with KK. That KK win rate is absurd, of course, and is largely the product of small sample variance. I got KK all in preflop against AJ and lost. Adjusting that hand for my all in equity results in a much more reasonable KK win rate of $14.95 per hand.

But I also ran good with AA; I got it in preflop 4 times in the month, which is unusual, and held all 4 times. So reducing those wins to their all in expectation, and increasing KK to account for the suck out leads me to an equity adjusted win rate for the group of about $23 hand, or 11.5 bb hand.

Online, good players usually had a win rate with this group of between 7 and 9 bb per hand, so me doing 11.5 bb per hand is probably in the neighborhood of fine.

Ordinarily, the next thing I do is break these hands down into positional categories; in this sample, I am skewed by negative variance. What I should see is this:

Blinds: 22%
EP: 22%
MP33%
LP: 22%

What I actually see is:

Blinds: 25%
EP: 34%
MP: 18%
LP: 22%

And such small samples, of course, that it's hard to draw conclusions. For example, I lost 5 of 8 times I got KK in EP, so my WR there is -$3/hand. So I'm not going to draw any significant conclusions on my by position win rates. For now I am going to be reasonably satisfied; I took a small sample of hands in which I was OOP more often than I ought to be, and I did pretty well with it.

With my WR up around $23/hand, I don't expect to see any major leaks. But I looked through my hand histories, and I did find two hands that I think I made significant mistakes in.

In the first hand, I have KK on the CO. UTG limps, the HJ goes $7, I 3 bet to $23, folds back to the HJ who calls. Flop is A86r, check check. Turn 2, he checks, and I bet $15, and he calls. River is another 2, he checks and I check back. I missed a bet there OTR; even if I just stick in another $15, I get a little value. I knew I had the best hand, I just didn't figure he'd call a second bet. So I am calling that a $20 loss of value, more or less arbitrarily.

In the second hand, I snatched defeat from the jaws of victory. I had AA UTG. I raised to 8 and got called by the button and the BB. The BB was waiting for 2/5 table, and was donking into pre flop raisers like it was his job when he flopped top pair or a good draw. The flop came down QT9r. The BB checked, which I assumed meant he had nothing or had flopped the nuts. So I checked thinking I was giving him a chance to catch something and avoiding a check raise. The button checked too, which wasnt a surprise, as he had a big calling range. The turn was a 7 that also put a flush draw out. BB checked again, I bet 17, button folded and BB called. The river was an 8 that also completed the flush, and it went check, check, and I lost to 96o. So if, as I should have, I bet/folded the flop, I most likely take it down there and win $14; instead, I lost 25. So that was, at least, a $39 bone headed check on the flop.

Those two hands add up to $59, which is close enough to $2/hand to go ahead and say that I have a 1bb/hand leak doing boneheaded stuff with my premiums. Not that big of a deal, but it's something to work on in a spot where we don't really think that much about our play.

One other use I make of this data is to adjust my monthly win rate with precision. For example, I was at $18/hr when my month came to an end. But I know that I got "too many" AA and KK hands; 6.4 too many. So if I am calling my WR $23/hand, and I know I got 6.4 too many, then I also know that 23 x 6.4 = 147 of those profits were attributable to run good, which reduces my actual win rate to $16.33/hour. As I go through the other ways in which I ran good or bad, I will make the appropriate adjustments upward or downward to my WR. For example, I ran $1300 below all in expectation, mainly by losing three huge pots when I had excellent equity, which adds another $14/hour to my WR, so I actually was at an EV and variance adjusted WR of $31/hr when the month abruptly ended. Would have been a nice heater had I held in those all in spots, instead of a meh month.

OK, this is long enough that I'll go ahead and put it up and do cold calls in a separate post.

Partly, depending on the table dynamic.

When there's an EP raise and a couple of calls, I'll tend to call with small PP or broadways, where I might otherwise have raised myself and c-bet/barreled on a tighter table with fewer people involved. In the first case you can get paid really well when you hit, sure. But you have to see enough of them to get that win. If it's the first or second time you play that spot in the night, you're golden. If it's the 15th your shorter stack cripples you. Although, this is probably more of an argument as to why you should top off more often.

At a table where there are fewer callers postflop I feel like I can find more spots to get value or bluff players off even when I miss, leading to more winning sessions even if I'm *not* hitting that huge hand.

Cold Calling analysis

So, again, I played 2816 hands in September. I cold called 57 times, or, about 2% of the hands I played.

Breaking them down by position:

Blinds: 21/57
MP: 20/57
LP: 16/57

(as an aside: If I played 2816 hands, that means I played something like 341 orbits [assuming the table was usually, but not always, full]. So that means I posted 682 blinds totaling roughly $1000. This will be useful to help us see how much we're losing from the blinds without actually tracking it. It also means that I only called 21 raises out of 682 hands played from the blinds; so, even given that not every hand is raised by any stretch of the imagination, it is still pretty clear that I am pretty snug in the blinds. But we'll have to wait a couple months to find out exactly how snug.)

In terms of my starting hand selection, I am heavily weighted toward pocket pairs.

Pocket Pairs: 18/57
Broadway (JTo-AKs): 24/57
Suited Connectors: 4/57
Suited aces (A2s-A9s): 4/57
Anything else: 4/57

For the record: I consider my cold calling range to be ONLY pocket pairs and broadway hands. If I call with anything else, I am playing the situation--maybe he stuck in an abnormally small raise that I can call with suited connectors, or maybe I am playing the player.

Here are my actual results with those groups:

Pocket pairs: $15.60/hand
Broadway: $9.91
Suited connectors: $9 hand
Suited aces: -$56/hand (lol, I got super coolered top boat < quads)
Other: $9.25

Adjusting for variance:

Pocket pairs: I flopped three sets, rather than the two I expect in 18 hands. I think from previous study* that my WR with a flopped set is around $45/hand, so I give myself $135 in winnings with those sets rather than the $383 I actually made, and that reduces my winrate to about $1.50 per pocket pair.

That tells me pretty clearly that I am losing money when I don't flop a set. That's not that surprising. The good news is that if my total winnings were $281, and my winnings with sets were $383, then my losses with pocket pairs that don't flop a set were $102; in 15 hands, that is a loss rate of $6.80 per pocket air that doesn't flop a set. That is significantly smaller than the average size of the pre flop raise I am calling, so that means that I am winning back a bit of each raise I call even when I don't flop a set. What we will look for in our hand history review is ways to do a bit better.

(*I'm really soft on this number; I derived it from one month in which I felt like I ran bad getting paid with my sets. So I'm not convinced it is actually my true win rate with a flopped set. But it's the best info I have, so I use it with an asterisk. Shrug.)

But, for right now, we're going to call my variance adjusted win rate with pocket pairs $1.50* per hand, and just leave the asterisk there to remind us that this is pretty much guaranteed to be a low ball estimate.

Broadway hands: I played 24, and flopped trips twice. That's run good, but I'm not going to adjust for it, because neither of those pots wound up being giant contributors to my WR, and because I flopped anything a bit less than I should have. I won 10 of the 24, and I had air in three of the wins. So, all in all, I think variance is a wash here, and I am going to go forward assuming that my WR when cold calling with broadway is somewhere in the neighborhood of $9/hand.

(Note what we have now learned about our blind play: We make about $1.50 when we call with a pocket pair, and about $10 when we call with broadway, for a weighted average of about $7 per hand. We did that 21 times, so we reduced our blind losses from $1000 to $853 by cold calling. We also expect 6.2 premiums in this sample, each of which is worth $23, or a total of $142, further reducing our blind losses to $711. Hands where we 3 bet or are the preflop raiser out of the blinds should further reduce these losses. Let's just give ourselves another $100 on those hands, and then another $100 in profits when the pot is limped and we win it, and we're down around $500, and we're winning back roughly half the money we post in the blinds. That's a good, solid performance, but it's not great, and we definitely have room to improve it. So we're going to make tracking our blind play a high priority. It's pretty safe to assume that we have identified a potential leak in the blinds, just as a side effect of looking at our cold calls and premiums.)

Suited connectors, aces and other: I am going to group these together or the purpose of analysis, because all of them have this in common: when I play them, I am playing the situation, not the hand. So they tell me how well I am reading a marginal situation as exploitable.

My achieved WR with these three together is -$16 per hand. But, as mentioned, one of the 13 is a super cooler that won't even happen once a year. So that raises an interesting question of how to control for it. I could take it out, pretend it never happened, and look at the remainder of the hands. I could give myself a fraction of the loss; if I think I am going to get super coolered once a year, then I could amortize the $256 cooler over those 12 months, counting only $21 of it in this sample. Or I could replace it with an average sized loss (this is what I would do in a slightly bigger sample). In this case, I am going to amortize the loss, and credit myself with losing $21 on the hand.

That gives me variance adjusted winnings of $27 in 13 hands, or $2 hand. That sounds about right for spots where I am speculating with trash I would normally fold because I see a rare spot where calling is +EV.

Summing up: We have adjusted steeply downward from $15.60 to $1.50* my win rate with pocket pairs. We think $10/hand is representative for the broadways, and we have adjusted our ash and trash win rate upward from -$16 to $2. That leaves us with a variance adjusted win rate of about $3 per cold call. That seems a bit low to me, so I suspect a leak.

Turning to the hand histories, the leak is immediately apparent, and exists with the pocket pairs. I won't bore you by reciting the hand histories of the hands where the leak manifests. Suffice to say that there were three hands in which I hero called with underpairs either one or two streets where I should simply have given up, and cost myself an extra $55 I didn't need to lose. This is pretty key, as simply folding to those bets saves me $3 per hand, and increases my win rate with pocket pairs from $1.50* to $4.50*.

I caught this leak midway through the month, and, after noting that I had made two bad calls in the first three or five sessions, I only made one more for the remainder of the month.

So that's about it. Summing up both posts, we identified a 1bb/hand leak with premiums by missing value, we identified a 1.5bb leak with pocket pairs by making dumb calls, and we identified a potential leak in the blinds that we have put on the agenda for future study.

If I had to sum up the process it goes like this:

1. Record the hands
2. Calculate your achieved win rate
3. Calculate your EV Adjusted WR
4. Calculate your variance adjusted WR
5. Look at the variance adjusted win rate to determine whether it indicates a potential leak (obviously, any negative win rate would so indicate, but, as we saw here, so also does a small positive number where you think "hmm, I feel like I should be winning more in this spot).
6. Look for patterns in your positional winrates, allowing for small sample variance.
7. Look for the hands that were misplayed.
8. Look for patterns of mistakes, such as my hero calling post flop bets with underpairs in bad spots.
9. Tally the mistakes you made and divide by the number of hands to get the leak/hand value.
10. Plug the leak by focusing on it in game.

Hope this helped. And, as I said, in the OP, if you have questions about the hands you recorded, feel free to ask them.

Thanks for sharing your routine mpethy. I just got my iPod touch all set up for recording poker hands today and will be out of town most of the weekend-hopefully I can squeeze a session in tomorrow.

Do you think you could have come to the same conclusions simply by looking through the hand histories and trying to spot leaks without first trying to determine a hypothetical ev adjusted, variance adjusted, win rate? It seems like you're doing a lot of detailed analysis and making many assumptions because of the small sample size when you really could just go through the groups of HHs with a critical eye and spot the leaks on your own?

Is it that the numbers just make leaks more apparent?

I guess what I'm really asking is: how much am I missing out on if I only do steps 1, 7, 8, and 10?

Those are the gist. You can spot mistakes by analyzing hand histories. You can spot leaks by organizing your hand histories and then analyzing them.

If you skip the WR calcs, you're losing valuable, but not essential information. Chiefly, if you do the WR calcs, you'll know how much money you've cost yourself, how much you stand to gain by plugging the leak, and the baseline WR so you can be confident that you plugged the leak (or not) when you compare later results to current results.

Like all poker study, making the easy rapid gains pays big dividends, and then you have to go to more in depth study to get increasingly marginal gains.

At this stage, most people will get the most benefit from just recording their hands, organizing them and then analyzing them. That's the big, easy improvement. Afterwatd, if you get into empirical analysis of your game in a big way, you'll know how to do the detailed analysis.

Isn't this way too much results oriented thinking? I mean, we could have also shoved preflop and this would have been far more +EV than our result; does that mean it would be a better play?

Betting/checking the flop for me here is much more dependent on (a) how comfortable we are folding to a raise versus these opponents and (b) stack sizes (i.e. SPR / commitment issues).

GcluelessNLnoobG

I'm assuming you had good reason to continue with your non-set hands, so even though you lost a total of $55 in these 3 instances, isn't it the case that you actually expected to profit long term with those calls?

GcluelessNLnoobG

In both hands, the issue is not the result, but the fact that subsequent analysis shows that the thinking behind the play was flawed.

In the AA hand, I made two mistakes in my thinking: I failed to adequately consider the button's presence and I failed to think about what it meant to let them catch something. There are way too many turn cards that beat me or kill my action for me to be letting two people draw.

In the case of the small pocket pairs that I am hero calling with, I am only counting it as a leak when it is clear that it was a mistake to continue. I have subsequent calls in my hand histories that lost that I didn't count as mistakes. In these cases, it's all pretty simple ranging errors. I gave villains a bluffing frequency that, given the situation, was just unreasonably high, and, had I ranged appropriately, I would have folded.

You can't just say it wasn't a mistake because I thought it was profitable at the time. That thinking eliminates the possibility that you can improve, because you never make a mistake. What you have to do is challenge the reasonableness of the in game conclusion that the play was profitable.

The way you do that is with data. If you make 5 calls that have to win 33% of the time to be profitable, and you go 0 for 5, with all 5 villains showing up with a hand you didn't really consider as in their range, or even if all 5 are just in the top of their range, both logic and math would indicate not that this is just bad luck to run into the top of their range, but that in actuality, the top of the range you assigned is their entire range. That is, that you assigned an unreasonable range.

^^^^

Regarding continuing with your non-sets, that's fair enough.

I don't think the AA hand is as clear-cut as there could easily be other factors, such as the ones I already mentioned, plus lots of others (is it more profitable with our image to attempt to get paid off versus setting up a bluff catcher if one of these guys is known to be aggro, throwing in some checks here for overall balance ain't horrible, etc.). I mean, sure, checking here certainly isn't the default play but I don't think it's necessarily the mistake your are making it out to be either.

GimoG

Ok. It's a judgment call. I'm fine if we disagree on what was best. In game, I certainly thought the check was justified. Afterward, it looked like a mistake. Given the BB's exact holding, it was clearly correct. Given the board and his range, I'm not so sure.

I'm assuming online winrates (unless you are talking uber penny stakes?) are <<<< live winrates? If so, should we expect a big difference in live winrates with regards to AA/KK?

In my 53 booked hours so far I've had 9 AA (above expectation) and 3 KK (below expectation), for about the total number of times I'd expect them.

21.5 bb/hand per exact results
31.3 bb/hand per equity adjustment
18.1 bb/hand minus the time my AA coolered KK and AK

ETA: 40% of my profits for these 53 hours, fwiw. With 24% of my profits coming on the single AA vs KK vs AK cooler (wow, what a stat, imo).

GcluelessAA/KKwinratesnoobG

Yes, we should expect a significant difference between live and online win rates. But the difference will be smallest with respect to AA and KK, simply because they are so absolutely strong relative to other hands that even playing them online you can realize most of their inherent value (assuming there is such a thing).

Where you will see a big difference in win rates is in a lot of spots that are marginal online but are fat live. For example, my WR cold calling with broadway was almost 5bb per hand, and looks reasonable. Online, my win rate cold calling those hands was maybe 1.5bb/hand (SWAGging an adjustment for differences in how it was calculated on my tracker versus how we're doing it). Similarly, my win rate online cold calling pocket pairs was about .5bb/hand; here, it is 1.5bb/hand.

In these marginal spots, I have always guessed that live win rates would be between 3x and 5x online win rates, and that is turning out to be pretty much the case.

I'm not sure what it will look like for fat value stuff like AA and KK. But it pretty much has to be smaller, as it would send live win rates through the roof if the live WR with AA was 50bb/hand, 5x a good online WR, and 35bb with KK.

^^^^ makes sense

M, what would your opinion be on how much of our overall winrate AA/KK makes up? It's almost disheartening to look at my lol 53 hour sample and realize the majority of my profit is simply waiting for these 2 hands.

Gcourse,Iguessalotofitdependsonhowhorriblyweplayal lnon-AA/KKhandsG

Online, about half the money people won was AA and KK. based on the above, it should be less live, but it's still going to be a substantial fraction.

And remember the blinds. You're losing a good fraction of your blinds every orbit. If you're doing well there, you're losing $2 an orbit, call it $7 per hour, $56 per session. That's a lot of drag to overcome. Online, where a good WR was 2 or 3 bb per 100, I used to do the math to show people that if they played every other hand just with an eye toward paying their blinds, that would leave them with AA and KK as pure profit, and they could make a decent win rate.

So, yeah, you're not wrong in thinking about it the way you are. But it's not a bad thing. What you have to remember is that the very fact that you're paying your blinds (and the rake) with everything else is what makes you a winning player.

If it makes you feel better, just reverse the thinking. Tell yourself that AA and KK are designated toward paying the blinds, and you'll pull your profit from everything else.

Think instead I'm just going to practice on coolering KK and AK with my AA, that seems to be a solid strategy too.

GcluelessNLnoobG

Booked my first session since committing to myself to record hands. Tonight I recorded almost 30 hands which is far more than I have in any other session. I started off trying to record everything and I was nearly able to. I will wait until I have more than a 6 hour sample to do much analysis, but already this is proving to be a valuable learning experience.

I'm beginning to think one of my leaks might be opening a bit too wide after 0/1 limpers or 2 tight limpers in LP with very weak hands (hoping to take down the dead money now or with a flop cbet). I'm curious as to whether this really is profitable for me. So I'm going to start tracking the results of these (and only the results instead of the whole HH) just to give me an idea of what's going on. Last night I was -$73 in 3 hands in this regard, but obviously need a lot bigger sample size to get the bigger picture.

Having said that, I realize LP steals ain't *just* about the bottom line, but also has some sorta meta game / image benefits. If I completely removed this from arsenal I'm sure I'd be even more face up than I already consider myself. If I end up being anywhere near breakeven I'm pretty sure it's a good investment. But right now I'm guessing I might be quite a ways from breakeven with these plays, and hopefully tracking the results will tell me.

Lol, only recorded 2 hands in a 6.5 hour session last night: AA and KK. Ran my AA into KK and pretty much made the minimum due to both of us playing our hand weirdly preflop, sigh.

GcluelessLPstealingnoobG

Holy ****....this COTM is incredible. Much thanks to Spike and Mpethy for putting it all together, not to mention the great posts from everyone else.

I haven't played poker with any intentional regularity for quite a while, but I'm thinking about taking a serious look at it again. I can't tell if this COTM makes we want to play more, or not.

Good stuff! I quit playing for a few years and just got back to the tables a few months back ago. I can't imagine recording every hand with my methods, it'd kill me. Session record keeping is a big chore already, and I'm hoping to streamline it. A paper form is used to jot down session records, notable hands and players, and I also make an audio recording on my phone while on breaks, or on the drive home. From there, it gets ugly.

Afterwards, I put the data in (up to) four different files on the PC:
Session spreadsheet- records time, date, location and $ data
Journal- records the hand, action, my performance etc.
Player spreadsheet- thumbnail info on regulars
Opponent journal- detailed notes on key regulars

The player files don't need updating after every session, but when they do, it takes a chunk of time. I'm interested in how others are managing things, and looking for a way to reduce paperwork.

A results tracker app will take care of the math immediately.

I use voice recordings extensively because it's faster than paper/phone and I can download to comp later for storage without having to transcribe every word.

Skimmed through it all, looks pretty good. I think its important for intermediate players to understand how to use stove to proof their hands with math. Ive found that stove (cruncher for iphone) has immensely improved my game.
Doesn't look like you included any examples of how to do that here, perhaps for another time

Thanks, I started my methods way back before I had a smartphone, and am just starting to make good use of it. I tried using the note function, dictating when away from the table, and also typing at the table after key hands. I look just like another guy busy texting- perfect.

Used it for a 10 hour session yesterday, and it made the post-session records a breeze. My next minor challenge is carrying an extra phone charger.

I've been tracking every VPIP hand for the past few months, have over 150 tracked hours (I've played more but didn't take notes every time). Modest by any standard but I'm just a weekend warrior. Some of results are ldo, others more intriguing, but I still have sample size issues. So far I've broken the hands into categories: monsters, premiums, PP; by action (PFR, CC, limps); and all hands over 50BB swing. Still need to separate more by hand (such as Axs, SC), but not a ton of data yet.

The most interesting thing was that virtually 100% of my winnings come from hands when more than 100BB went into the middle. Smaller pots, so far I'm just passing chips back and forth (though as GG said above, it's still important for metagame). When I have more time I should look at EV as best I can, and see how much run good I'm on vs actually playing well.

If there's anyone who wants to compare notes, send me a PM.

+