AKQJ10’s Poker Blog

[I've decided that actually working through this self-coaching in a document viewable by the public probably isn't the best way to go about things.   It's not that I mind sharing the process with you; it's just too confusing to try to document what's going on when each day, I make tiny efforts to discover something about my game that may or may not produce any fruit.   So I think I'm better off doing a long-winded analysis, figuring out how to apply it, and sharing the results with you when I'm finished.

At any rate, I think I've figured out some key things in my game, but I'll post about them later and stop adding to this post.   This one started around January 26, 2010; the very newest changes in red are from somewhere around February 1 or 2, and anything I left in purple is from somewhere between the two.]

In my previous post I mentioned that my LO8 results took a huge swoon in mid December.   My goal now is to figure out what leaks I need to plug to be a maximal winner at stakes around $2/4-$3/6, and obviously a subtask to that reach that goal is figuring out what if anything changed about my game around that time. As an aside, despite the title of this post, I am most willing to pay for coaching if I become convinced it’s necessary.   Even if I go that route, I’m pretty sure that doing the heavy lifting myself of understanding what needs to be improved in my game (or at least understanding why I can’t seem to find patterns after a deep look at the data) will help me get the most out of each our of coaching time.   Self-knowledge is certainly important to apply other more-cost effective training aids, such as the subscription to Deuces Cracked that I just renewed after letting it lapse for several months.   (Make the Jack a Four, a series on short-handed and tight-aggressive LO8 that came out since the last time I subscribed, looks particularly helpful.)    And thinking a lot about my game with the help of quantitative data will help me choose a reasonable and representative subset of hands for posting on the 2+2 or DC forums. So what questions do I want to ask myself?

• How do my basic stats – VP$IP, PFR, metrics of positional awareness, anything else that jumps out – compare before and after this downswing?   When I looked at this before, I didn’t notice any obvious difference, but it’s worth another look.   Answer: The most fundamental change portrayed by the stats is that I’m raising more, but from worse position, and stealing less.  See below (eventually).
• Am I doing better or worse depending on the number of players starting the hand?   When I looked at this before, I was doing best with the “full game” numbers of 9 or 6 players and not doing well with 8, 7, or the smaller numbers, although the sample sizes for heads-up and 3-handed are pretty thin.
• Am I doing better or worse depending on how many take the flop?   This is key, because I can’t shake the feeling that I’m not adapting my aggression to tougher games properly.
• Maybe play around to find a way to estimate if I’m bluffing too much.   Problem is, for example, I can’t just filter for hands where I flop nothing.   Of course I’m going to be loser in situations where I have nothing.   The real question is whether I’m more loser running the occasional bluff, or whether I’m better off just cutting my losses.   Perhaps I can use the filters to help me learn.
• Look for where I’m losing money by position.
• Am I losing money with trips?   JoeTall says this was a leak for him in From a Donk to a Stud 6, so I should probably check that out for myself. Never mind — this is another feature that doesn’t work on that piece of crap called Omaha Manager.
• HEM/Omaha Manager has some interesting stats on river call efficiency; I looked at them briefly, but it might be worth trying to understand them better.

And one more diagnostic strategy that shouldn’t take a lot of time to implement, but does.   Alas, Omaha Manager just isn’t made for O8 (and I didn’t fully evaluate it during the refund period, so I didn’t notice this until too late): It can’t identify relevant O8 starting hand groups.    You can’t even find your results with A2 hands, for crying out loud; A234 would be considered a “small rundown”, I think, because the hand categories are all based on Omaha high.   This issue in UserVoice (link not working for me at the moment, but that should be the right link) must be fixed before I would consider recommending Omaha Manager to an O8 player, be it limit or big-bet O8. I’ve considered hacking my own SQL queries with Postgres, and I think the data would be most helpful.   It’s time consuming enough that I want to try other means of troubleshooting my O8 game first, but I may end up having to do it.

The results, bit by bit

I envision the rest of this post being a “living document” that I update as I do the rest of my analysis.

Basic stats, before and after

I’m sort of revamping this section – I ran the “2 – Leak Buster – Overall” report for before and after, but figured it was way too much data and offloaded it to this page: LO8 Leak Buster before and after. And I moved the rest of the data because the WP editor hosed up the formatting.

Really the only obvious difference there is that my preflop raise % increased from 8.6 to 10%, which is a pretty good increase a variable that should require less sample size to converge.   Is it just the slightly higher mix of stakes?   Apparently not, because my PFR doesn’t show a nice consistent correlation to the stakes; it jumps around, 10.2, 9.1, 8.3, 9.5 as you go down in stakes.  Each is between 3500 and 6700 hands so I’m not sure if those jumps around mean anything.

How do I do in the hands themselves where I raise?

BEFORE: 96.46 BBet/100 in 1165 hands.

AFTER: 85.5 BBet/100 in 875 hands.

Doesn’t seem all that different does it?   And the non-raise numbers are -4.3 and -19.6 BB/100.

So it’s just not a credible interpretation that I’m raising too much.   It does prompt the question of whether I’d do better to just fold any hand I don’t want to raise, with “do better” defined relative to a loss from the blinds.  If we take out the blinds, before I was +4.88 BB/100 and after -1.21 BB/100.   But overall I’m +EV with non-PFR hands out of the blinds, so I don’t think I need to be so radical as to fold everything I wasn’t planning to raise. So far this seems like a wild goose chase.   Let’s move on.

By number of players in the hand

If you can find a pattern above five-handed you’re more creative than I am.   But it does seem to suggest that five-handed or shorter I might really suck, although it is a wee little sample.   Maybe leave games sooner if they get really short?

By number of players taking the flop

The only clear pattern I see in this data (other than my “after” results being a disaster, of course) is that until 2009-12-12 I made more money the more multiway the pot. That makes perfect sense. But that pattern falls apart not only in the small “after” sample, but also in the cumulative results! What is going on? It’s almost like I’ve been playing a different game since then, a higher-variance game, and the randomness of most of the numbers I’ve looked at so far suggests that maybe I really am just running terrible. At any rate, my leaks are well-hidden so far.

Looking for my leaks in position

I don’t know why this approach didn’t occur to me sooner, but it really makes perfect sense.   When I noticed that my steal % had gone down but my PFR hadn’t changed appreciably, it dawned on me that I’m probably raising more in non-steal positions lately.  Obviously building bigger pots without position could be problematic.

So let’s run the numbers and see in what positions I’m actually losing money relative to before my downswing:

This pretty much tells the story – most of my loss is coming from the early positions.  And how has my VP$IP and PFR% changed by position?

Now, it’s important to be realistic about our sample size here; if my original samples are around 13k, then the positional samples are going to be much smaller.   Now, we know that my PFR went up around 1.4% overall, but apparently my raise % is actually increasing the most on the button.   This doesn’t seem very consistent with my earlier diagnosis of less steal %, so what’s going on?   I must be doing more non-open raising, right?