PokerTracker and Spotting Leaks..

by Excession

OK this topic was voted the ‘most useful’ for the next instalment so here we go..

It’s time to move the focus away from your opponents and back to yourself. Let’s try to find our leaks. For this purpose I’m defining a leak as any sort of play or conduct that lowers our overall profit (or increases our loss).

OK what do you think your biggest leak is? Overplaying AQo. Not spotting sets when they flop for your opponents? Folding too much at the river?

Actually it unlikely that these technical leaks are your worst (useful though it is to spot them).

Your biggest leak is lack of discipline!

I’m going to divide leaks into 'discipline' and 'technical'.

1. Discipline leaks
A) Playing too many hands (which once you have seen a list of recommended starting hands is usually a discipline more than a technical issue).
B) Going on Tilt (includes all sorts of failures of discipline).
C) Not selecting tables properly (which in the GT+ era includes not pre-observing tables for at least 20 minutes before sitting down if you are playing at Party or UB).
D) Not properly reviewing sessions you have played.
E) Playing outside your bankroll.

2. Technical Leaks
General
A) Too little or too much post-flop aggression.
B) Calling too much.
Specific
A) Positional leaks (including defence of blinds).
B) Misplay of particular starting hands.

This is a huge topic – this article is 4,000 words long and deals with 1A only!

However I guarantee you that by reading it and following its suggestions it you will find several problems in your game and be able to identify and lessen the ‘playing too many hands’ leak that just about every player has.

By playing ‘too many’ hands I don’t necessarily mean more than a given ‘magic’ number in a session. Whilst I would strongly recommend an overall Vp$iP for NL 50c tables of 20-30%, there may be tables or circumstances (such as a run of good or bad cards) where 12% or 40% are perfectly appropriate. However it is almost never appropriate to play starting hands that consistently lose us money and yet, for whatever reason we all do.

1A – Starting Hands

A wide variety of Vp$iP’s are consistent with playing winning poker at the lower NL stakes. Different types of aggressive post-flop players with Vp$iP from 15-35% all show an average profit in my 25c and 50c BB NL database (which rates over 12,000 players). The optimum seems to be 25-30% of hands at these stakes. However at higher levels I am reliably informed that this drops to 15-20%

Most of us have mental or actual list of ‘core’ starting hands. These are the core hands we will limp with in most positions on most tables (you might occasionally fold one from EP on an aggressive raising table but broadly they are your ‘always playable’ hands). When we start we usually get these from a book or a website. The problem with books and other people’s advice is that they aren’t metagame* specific – often they aren’t even NL Ring specific. If you play the Party $25’s mainly on a weekend you are going to find a much wider selection of starting hands with a Positive Expected Value than in you play the Monday Morning $200 tables at UB. If you select the highest average $/hand tables (usually the ones with the most maniacs and pre-flop raises) as opposed to the lowest (usually more passive) you will again find that this will affect what hands can be played for a profit. Similarly your particular style will affect EV’s for various hands.

* Metagame is a term imported from the Magic the Gathering card game. It means adapting your play to precisely fit the playing environment around you. This is particularly important in MtG as each player makes his or her own deck up (don’t ask!). It’s unlikely to be a co-incidence that 2 of the top 9 in this years WSOP were MtG players and that Annie Duke got busted out early when she tried an all-in bluff against an internet player which ‘he couldn’t possibly call’. PT allows you to be metagame specific because the data it contains is, by its nature, about the very environment that you play in. Trust me when I say that this is a very good thing.

So the best way to build a list of +EV hands for you, once you’ve got say 10,000+ hands in your DB, is to actually examine your starting hand stats from the PT ‘General’ page. If a hand shows a statistically significant +EV then you should think about adding it to your core list. It is of course possible that you are losing money with hands that would be profitable if you played them better or only from certain positions. We will look at this in 5.2(D) later on. For the moment we are taking the simple approach to plugging starting hand leaks – if you win good money with it wherever you play it from, put it on your ‘core’ list. If you win a little money with it or a fair bit but have to pick your spots, put it on your ‘marginal’ list. If you lose money with it, don’t play it.

I know that is can be argued that by laying marginally losing hands you get more action with your winning ones (the reason maniacs get paid off so well when then do have AA) and so that logically it may be worth playing marginal losing hands if this increases your EV overall, but I hope you can see what a slippery slope to losing focus and discipline this argument is.

It’s a good excuse for playing losing hands in theory. In practice, for all but the very best players, it’s just an excuse for breaking rules you have imposed for your own good and it will cost you money. Trying to get such marginal edges as a ‘looser image’ may provide isn’t usually worth it at low stake tables. So for now we will keep it simple – play your winning hands, ditch your losing ones. If that isn’t good leak plumbing then I don’t know what is.

One of the great benefits of examining your starting hands to compile a core and marginal list is that it flags ‘problem’ hands for you straight away. We will see several of mine as we work through the process. When we come across them we will add them to a ‘watch’ list and will look at them in more detail in Part5.2D.

The other point is that you can still be winning with a strong starting hand, but winning less than you ought to, which is just as much a leak as playing hands that make a loss. This is where objectivity is difficult as there is no table of ‘what this hand should win on a Party $25 table’ available. The best you can usually do is to compare its performance to similar starting hands in your list although if you are absolutely stuck you could use the PokerRoom EV stats. I am trying to address this lacuna in the data and there is more on this at the end of this article.

OK as an example let’s build a ‘core playable hands’ list from my PT database of my last 10,000 or so hands (1st December 2004 – 22nd January 2005).

• Fire up PT Ring Game Stats.
• Edit out any game-types that aren’t relevant from the ‘Prefs.’ Page first (you can edit by dates, sites, limits, NL,PL or Limit short-handed or full ring for example – you shouldn’t mix your 6-handed stats with your full ring ones for obvious reasons).
• Then go to ‘General Info’ page. Click on the ‘filter’ button.
• Exclude the Blinds and hands where less than 8 players at table (assuming you are looking at full ring games). Click ‘OK’.
• Now look at your known starting hands. Click ‘p’ and make sure ‘preview zoom’ is set at 100%. You can then scroll down the list.

Here are the first few hands of mine (click the graphic to enlarge). We’ll look at what on earth is going wrong with AKs and AQs in 5.2(D) J (this means we add them to the watch list). Looking at this screen grab I am immediately embarrassed by how much A-junk I play, but at least I can claim to be getting away with this on the numbers shown. You’’ll see that having excluded blinds we are down to 7,000 hands or so already. You can do this exercise with less than a starting DB of 10,000 hands if you like, but I wouldn’t put a lot of faith in the results for obvious reasons.

OK so now to compile your ‘core’ and ‘marginal’ lists.

First of all you have to look at hands in groups. You can see that AKs and AQs are both losing hands. Should I stop playing them? No - for two reasons. First, suited (non-pair) hands are obviously three times as uncommon as their non-suited equivalents (being 4 of the 16 permutations of that hand) and so, with small databases such as we are using here, results can be very misleading – one bad beat and it looks like a losing hand (more on this in 5.2(D). Secondly, and more importantly, AKs is self-evidently better than AKo (with which I have won $157) – so overall AK is up $145 and I am playing both versions.

What you do is look for the boundary – if KQo and KJo are both +EV but KTo, K9o and K8o are –ve, then if K7o is shown as +ve, I’m still not adding it to the group. The EV boundary for Kxo would be set at KJo/KTo.

A similar point can be made about the pairs. Filter for ‘pairs’ and bring up the list. You can see mine in the next graphic (click to enlarge). Notice no evidence at all of the alleged Party Poker KK v AA ‘fixes’ to increase their rake. What a surprise! Also notice what a rough time I’ve been having with QQ and TT and what a good time with JJ lately We will check in 5.2D that I’m not overplaying QQ and TT. I do notice that the only 2 pairs I’m losing with are 55 and 33 and those are the two I’ve had the most of, but maybe I’m just getting paranoid now :-)

It looks as though 55 and down taken as a whole aren’t profitable, so one way of looking at these stats would be to just include AA-66 in the core playable hands, and take 66/55 as the boundary.. However with 99-66 are so consistently profitable and with the logic of ‘set or fold’ seemingly applying the same to 33 and 55 as to 66, for the moment I’m willing to put 55-22 into the core set (though I will examine them in more detail later). So add QQ,TT, 55-22 to the watch list.

One important stat to watch for on these screens is Vp$iP. As these are all non-blind hands, to see a flop I need to put money in. The lower the Vp$iP for a hand, the more its figures have benefited from me ‘picking my spots’ with it. When we are looking for core ‘always playable’ hands we need to be suspicious of hands which don’t have a very high Vp$iP (say 85% or more) even if they show a +EV. The ones with good profits but less than 85% Vp$iP are more likely to be the marginal hands.

Here you can see that I pretty much always play any pocket pair. My rule of thumb to call raises is Iceman’s ‘both stacks have at least 10x the calling cost’.

Next let’s filter by any ace with a suited kicker (Axs) (use the ‘select specific hands’ button). Again, click the graphic to see what I'm talking about here. Oh dear – maybe I shouldn’t bother playing them at all? :-( Well, writing off AKs and AQs as aberrations (or the results of overvaluing these hands when I’m playing them – definitely flagged for checking later) Axs seems to be OK down to A6s, so I’ll add AKs-A9s to the core and A8s-A6s to marginals (see Vp$iP stats) Again I am playing these most of the time, although my rule of thumb is not to see a standard or better raise with Axs below about AJs unless it’s real multi-way pot and I’m just in it for the flush or two-pair etc (if an ace falls, Axs is likely to be dominated by an ace with better kicker too easily for my liking in those circumstances).

Next some of the Kxs, Qxs and Jxs hands (click the graphic to see). What a mess. The Vp$iP for some of these is very high. Perhaps I need to be more selective.
Looks like the boundary for Kxs is K9s/K8s, Qxs doesn’t look worth playing at all (not even QJs) but JTs and J9s are ok (J9s very marginally so). Strangely all the x9s are profitable for me. I wonder if that’s because so many people play any two paint cards that a paired 9 on the board is more likely to be good than a paired J or T. It seems silly to leave QJs off my core list if JTs is going on it , but by a huge margin here it’s my worst hand. This could be just the small sample size and a couple of really bad beats, it could be because I’m overaggressive with it - note the higher PFR% with it - or it could be because it really is a losing hand on these tables. For the moment I will add KQs-KJs to the core list and KTs-K8s, QJs-Q9s and JTs-J9s to the marginal one.
QJs is straight onto the watch list for further analysis in 5.2D.

OK now unsuited paint cards (click on the graphic). 3 months ago AKo was a losing hand for me, so I’m very happy with the way I’ve turned it around. You can immediately see problems with KQo (looks like I’m raising it far too much). ATo, QJo and KTo are marginal winners. QTo and JTo are marginal losers. As KQs is such a big winner, KQ is profitable overall and because the obviously weaker KJo and KTo are also +EV, I will keep KQo on the core playable list. Core list AKo-AJo, and KQo. KJo, ATo, KTo marginal due to low profit and/or lower Vp$iP. QJo marginal. KQo onto the watch list.



Medium and small suited 1 and 2 gap connectors. Well T9s is marginal, T8s seems ok and then the really low suited connectors 54s,43s and 32s also perform better than expected (perhaps because they have surprise value or because they can make the wheel and bust aces with big kickers, I’m not sure – they have low Vp$iP so maybe I’m just limping in from LP to good effect). Suited connectors are really LP hands and yet I’m up to 92% Vp$iP with some of these. I think we can see a leak right there and that overall we can safely leave these to be marginal at best. I think the whole shooting match can go on the marginal list, except for 42s which can be classed as pure junk. Putting 98s and 87s, 76s and 65s on the watch list to see if they are more profitable from LP.

Finally some paint cards with medium kickers (usually known as ‘junk’). This surprised me a bit. It looks like I’m picking my spots quite well with junk like A6o and K8o (look at the low Vp$iP – these are just marginal hands timed well), but I’m playing A8o 60% of the time out of the blinds and managing to make a profit with it, and K9o 33% of the time and doing even better. I didn’t expect A8o to play that well to be honest. Still 4 out of 10 times I’m ditching it without a bet pre-flop and that smells more of a marginal rather than core hand to me. Core list – nothing. Marginal A9o, A8o,A7o,A6o K9o, K8o.

OK so now we have the following lists.

Core ‘always play’ hands

• Any pair
• AKs-A9s
• KQs-KJs
• AKo-AJo
• KQo

Marginal hands

• A8s-A6s
• KTs-K8s
• QJs-Q9s
• JTs-J9s
• Any other one or two-gap suited connector (except 42s)
• KJo-K8o
• ATo-A6o
• QJo

Junk

• The rest!

Watch list of problem hands/probable leaks flagged for later hand and position specific analysis

• AKs*
• AQs*
• QJs
• QQ*
• TT*
• 55-22*
• KQo*
• 98s
• 87s
• 76s
• 65s

* these hands are on the core starting hand list so they will have priority when we start our detailed analysis.

OK so the next step is to make two filters (one for the core list and one for the marginal list) and save them so they can be accessed in one click from the PT General info page at any time.

Do this as follows..

Go to the PT general page, click 'filters' [check that ‘not a blind’ and ‘8-10 players at table’ are still selected], go to 'specific hands' then check the boxes for your core hands. Type ‘core hands’ into the filter description box. Click OK. Your core hand filter is now available as a dropdown whenever you open the filter page (don’t alter it and click ok unless you want to change it). Exit using the cancel button or better yet create another filter if you want to play around with them.. You can see mine by clicking the graphic to enlarge. You will note the total profit out of the blinds is $926 – that is the theoretical amount that I would have won (outside the blinds) had I just stuck to my core hands and dropped everything else.

Next the Marginals (click the image below to enlarge). Quite a nice $140 added to total profit by these marginal hands.


If I had just played these and the core hands I would have made $926 +$140 =$1066 (excluding blinds).

OK let’s make make a ‘junk’ filter now for those hands that aren’t on either list.



My actual profit (excluding blinds) was $896 – so over the past 2 months I have thrown away $171 by playing total junk hands (I don’t mean from the blinds). That’s 16% of what I would have won had I just stuck to playing my pretty loose starting hand selection.

So now you do the same exercise – how much are you throwing away? If you are registered at this site you can simply post the size of your junk hands leak here And in this context guys, bigger is most definitely NOT better :-) (oh and no cheating by adding losing junk into your marginal hands just to make your % look more respectable :-)


Positional Filtering of Core and Marginal Lists

I will look further at general positional play and hand-specific positional play in later instalments of Part 5, but just wanted to briefly mention the possibility of using positional filtering of especially marginal and core hands.

Given how we have defined marginal hands you would expect them to play better from late positions than early ones. There is another leak we may be easily able to spot here.
Filter your marginal hands by the positional dropdown filter (the one under the ‘number of players’ filter). We are ignoring the blinds so we have 8 positions free on a full ring. Filter for EP (more than 3 off the button) and LP (less than 4 off the button) (there will be a little overlap in the MP positions when we have 8 or 9 players at the table but let’s not get too worried about that). In my case I see that playing my marginal hands from EP lost me $22 and that I would have been $169 up (not $141) from my marginal hands if had stuck to just playing them in LP (the overlap means that some MP hands are counted in both which is why marginal LP win minus marginal EP loss doesn’t equal overall marginal hands win amount.

If you see that certain marginal hands are holding up pretty much as well in EP as in LP that may be a sign that these are really core hands that you are playing badly from LP (perhaps overaggressively). If they are winners from LP but losers from EP then they are likely true marginal hands (and you should only play them from LP).

You can also filter your core hands in the same way. If you see that a core hand is an EP loser, you should consider shifting it to your marginal list (assuming it’s a boundary type hand and not just AKo that you are playing badly).

A final word of warning on statistically significant data. By splitting EP/LP hands you are halving the numbers used to draw conclusions from . In a small database this can lead to very misleading results. As a rule of thumb I’d like 50 instances of something to feel I can draw a fairly solid conclusion and at least 20 to start to point toward a trend. Anything less than that is probably too small a sample to rely upon in any way. I bow to any students of stats who have superior knowledge of exactly what margin of error can be expected and if any of them would like to send me suggestions I will update the guide as needed.

Get into the discipline of checking for junk hands played daily

So I think that we are agreed that having gone to the trouble of reviewing your starting hands and coming up with definitive core and marginal lists you should do your best to stick with them. This discipline is easy to acquire. After each day’s play filter PT from the prefs page just for that day’s play and pull up that dreaded ‘junk hands’ list. How disciplined have you been?
This is the easiest way to spot if you have been on tilt. Of course there are other tilts that this won’t spot (such as overagression with playable hands) but this simple 60 second check should be your first port of call after each session.

Review your big hands

Another good post-session tip, seeing as you will have PT fired up in any case, is to pull up you biggest 10 winning and losing hands from that day’s play and replay them. If you find yourself thinking ‘what on earth was I doing’ that is another exceptionally good way to spot tilt. Also, as chess grandmasters know, it is often more important to analyse the games (here hands) you won than the ones you lost. If you lose a big hand you spend a lot of time thinking about what you did wrong. But if you win it, well in the rush of 3-4 tabling you might not even bother to check what the opponent mucked – you may find out that you were much luckier and played it much worse than you thought…in Poker a lesson missed is one that you will have to pay for later…

More Detailed Starting Hand Analysis

We will look at starting hand analysis in more detail in Part 5.2D. There we will follow up on those problem hands on the watch list.

We will also see if our winning hands are winning enough. Unfortunately we immediately hit a problem. The most obvious way to this out is to compare a given starting hand’s stats in PT against an objective ‘target’ figure. This target figure could either be the average (presumably mean) stat for all players or an optimum target (being the average for that stat of the most profitable players). Now for stats that we have pretty solid data on such as Vp$iP we can glean target figures from all the players in our own database. For other figures though this isn’t as easy. Think of starting hand selection. In a DB of 35,000 50c NL full ring-hands there will only be a few players who I have logged over a couple of hundred hands or so. The only hands of my opponents I see are the ones that go to showdown or are voluntarily disclosed. It is impossible to actually draw any conclusions about the average BB/100 of say AJo when played by these players. I can see my own data on AJo’s (and compare it to my data on AQo and ATo) but that isn’t as good as being able to use an objective comparison.

On the BTP forum, I did ask for anonymous starting hand data exported to Excel to be sent to me - if you could send me yours that'd be a great help as I am short on data.

This is the real reason I will look at misplay of starting hands in more detail at the end of Part 5 of these articles. I’m hoping for some help from you guys to enable me to tell you what are average and good win rates at the 50c NL tables with various hands. If you do have 10,000+ 50c NL full-ring hands sitting in your PT database, please help – it should only take 5 minutes of your time and is totally anonymous.

If you want to rush ahead and try to do you own analysis of starting hand performance vs the average, I have found a make-do solution. PokerRoom has a very handy +-EV chart for all its hands in a full ring game (I think this includes NL games but this isn’t expressly stated anywhere so it could just be for Limit). Of course it won’t give you enough information to work out when to call, cold call a raise, raise or re-raise pre-flop from EP, MP, LP and the Blinds, so suggestions for that will have to await better data, but it’s better than nothing if you want to try.

In the next instalment, I will turn my attention to the other leaks caused primarily by lack of discipline, starting with Tilt, how to spot it and how to stop it.

Cheers,

-- excession

23rd January 2005.


You can find excession on our forums.