OK.
So I finally sat down and dug into some of my session statistics a little.
First some background on the data:
Just under 3500 hours since mid/late 2008, 3218 of which was of $1/2 NLHE (with some varying buy in caps).
About 100 hours of PLO, and an assortment of other games.
Roughly 2000 hours at my local "regular" casino, nearly 1000 hours at the local charity rooms and home games, and the other 500 hours distributed in 2-40 hour blocks at other rooms across the country.
I've mostly been a rec player. Used to put in a Friday evening and a Sunday afternoon session downtown like clockwork. 12ish hours/week, plus a little time on forums and reading books on the side. I honestly wasn't really very good when I first started (who is?) but managed to run up a nice roll quickly and have improved my game a lot since then. Now I play more shorter sessions in local rooms during the week too.
First off, winrates:
The first thing that jumps out from this is that my "other" games have been pretty much a wash so far. Some of that was playing stupid tournaments that I never made money from, but mostly it's been learning PLO. At least I'm not losing money there.
My $1/2 results seem to have a pretty constant slope with a bunch of heaters/downswings that offset the trend. Around 700 hours I started buying in deeper and trying to implement things that I'd read about. -$2500 swing by mis-ranging opponents and pushing in the wrong spots until I got a hang of it. Around 1200 hours that homework and experience paid and I had a nice heater while traveling to Vegas.
After that my results got swingyer but it doesn't seem to show in the session results (below). I think it may be related to a decrease in session length introducing more +-$600 spots.
The nearly $5k downswing near 2000 hours sucked hard. I won't claim it all to be "variance", I handled some of those losing sessions very poorly. Chasing losses, overplaying some hands I should have folded, folding hands I should have played because I was gunshy. There were a bunch of really gnarly $1000+ pots I lost from ahead that hit my BR and psyche hard though.
At 2500 hours I was finishing my graduate degree and my game (and health, and sanity) suffered. After finishing that everything seemed to start clicking again. Really interesting the effect that "life" can have on our games.
The lower figure shows my winrate over the past 100 hours, 500 hours, and overall, only for $1/2 NLHE. This is really interesting. The noise in the 100 hour sample is larger than I would have thought. For a purely rec player that 100 hours might take them 2-3 months to put in. This tells me that even playing against another regular player there's a lot of noise in our perception of them.
The trailing 500 hour rate seems to remove a lot of that noise and show how my game has evolved over time. All of the previously discussed events show up as changes in that value.
Of course the overall win-rate is difficult to move for a sample of this size.
All subsequent plots are $1/2 NLHE only.
Now for some session information:
In the upper left I've plotted session win rate vs session length. No visibly correlation. Although as the session length increases the volatility in the winrate decreases.
In the upper right the session wins/losses also fail to show any visible correlation to session length. The marathon sessions on the far right end up being "normal" wins/losses, not huge wins or losses.
In the lower left the distribution of number of sessions vs session length shows a fairly normal distribution centered around 5 hours, with only a few marathon sessions.
The lower right shows session wins/losses over time. I'd expected to see some shift in volatility as I developed as a player, but it's not really coming through.
Considering session length over time:
I can pretty clearly see my changing life priorities. Shorter sessions more recently and very rarely do I get a full 8+ hour day to go play. Interestingly, having more short sessions recently (when I'm better than I used to be) doesn't show up in Win-rate vs session length, while it did show up in the trailing 500 hour win-rate.
Lastly, distribution of sessions winnings:
With large ($100) bin sizes my sessions seem to be pretty normally distributed, with a little more weight on the winning side, especially at the high end.
Dropping the bin size flattens the peak a little bit and starts showing some interesting peaks on the losing side.
Going all the way to $20/bin shows some surprising features. The spikes occur at -$600, -$400, -$300, -$200, and -$100. My stop loss is $600, and I typically buy in for either $200 or $300 depending on the max buy in. So there are a lot of sessions that look like I lose a buy in or two and quit. There's less banding in the winning sessions though, although there is a bit visible at $300, $400, and $700.
This points at a couple of things. I probably need to top up more aggressively and/or leave before getting felted and running into a hard loss limit. Better to stand up with $35 and go home than push AT and coinflip at best.
I also see where I'd get kind of stuck with an even $100 increment on cashing out. I used to call bets and play hands until I chipped down to an even stack, hoping for a lock before putting in more money. This is obviously a mistake that I've cut out, although I'll still sometimes tip a random amount at cashout rather than take loose chips. (Does keep me informed about juicy games from floorpeople and dealers texting me when a whale walks in.)
I'm not sure how relevant or helpful any of these statistics will be to anyone else, but I found them interesting. I'm open to discussing them. If anyone would like to see some other figures or could suggest any other metrics to look at just let me know.