Tuesday, June 01, 2010

HORSE of a different color

On March 11, 2009, I reported the following:

WARNING! Boredom alert! I talk herein about my poker results, which nobody
cares about.

I have now played 125 single-table online HORSE tournaments. About 2/3 of
those have been $10 entry, most of the rest $5, with a few others ($20, $15,
etc.) scattered in. Most have been on PokerStars, a handful on Full Tilt. I have
tried signing up for them on Bugsy's Club (see here for why), but never got one going. A few have been turbo, most not. All have been eight-seat deals with three places paid.

If the outcomes were totally random, I should expect to cash 38% of the
time, or 47/125 events; I have actually cashed 44% of the time, or 55/125
events. Similarly, if all players were of equal skill and luck as each other,
one would expect to claim first place 13% of the time, or 16/125; I have
actually taken first 16% of the time, or 20/125.

In other words, I have proven to be just slightly better than the average
of my competitors--something like a 15-20% edge, depending on how you figure


All of which means that this is now a large enough number of tournaments that I can be reasonably confident that the results I'm seeing are a real reflection of a difference in how I'm playing, relative to the other players, rather than just random variance. I can have something like 85-90% confidence in that conclusion.

In short, I am Lord of the $10 HORSE.

I have spent $1186.40 on entrance fees, and taken $1312.00 in prizes, for a net profit of $125.60, which represents a return on investment of $1.11 for every dollar put in. However, my guess is that my average time investment is roughly 45 minutes each, so that's coming out to not much more than $1/hour. What's that they say about a hard way to make an easy living?

Over the last several months, I have really slacked off on these HORSE SNGs. But last night Cardgrrl suggested we do two of them together, a $5 and a $10, which seemed like a good idea. As luck would have it, I finished first in the former and third in the latter. It's not rare that I play two of these suckers at a time, but it is rare that I cash in both when I do. I was one busy boy there at the end, playing heads-up in one and three-way in the other. I know that my results in the $10 suffered because I was so concentrating on the two-way action in the $5; my opponent there was hyperaggressive, and it felt like all I could do to watch him and pick my spots well, with little attention left to devote to the other game.

Anyway, when I entered those cashes into my spreadsheet, I noticed that I had just over double the number of games under my belt as the last time I announced my results here, so it seems an opportune moment for another report.

The bottom line: I have gone from Lord of the HORSE to Lord of the Idiots (with apologies to George Costanza). I got worse, not better.

I have played 258 HORSE SNGs, with the following distribution of finishes:

1: 34
2: 34
3: 40
4: 27
5: 23
6: 40
7: 28
8: 32

If the order of finish were determined purely by luck--i.e., if all opponents were of equal skill--I'd expect about 32 in each category. As you can see, the actual standings don't deviate very far from that.

Similarly, I should expect to cash about 97 times, with the actual number being 108. This is, obviously, a little better than chance alone would dictate. It does not quite reach statistical significance, however; the p value for 108 or more cashes by chance alone is 0.08, a little over the 0.05 that is traditionally used to determine real versus fluke outcomes. (Let me once again put in a little plug for the wonderful binomial calculator you can play with to run such numbers here.)

Here's how the same analysis looks if I limit it to the games I've played since my previous report, i.e., just the last 133 of them:

1: 14
2: 16
3: 23
4: 17
5: 8
6: 19
7: 17
8: 19

By chance I would expect about 17 in each slot, and there's not much deviation from that. I should expect about 50 cashes, and got 53.

Bottom line: There is no evidence that I am systematically playing better than my opponents. It stings to admit that--especially in light of the phenomenally, inexplicably abysmal play that one often sees in these things--but it's an inescapable conclusion. How can I not be better than these people? I hang my head in shame. But I always try to tell you the truth here, even when it's not pretty.

It might look like I avoid 4th and 5th place finishes by buttoning down to survive into the money, with those games turning into 3rd places. But I really don't play that way. I know full well that all the money is in first and second, with third giving me just slightly more than my entrance fee back. I'm not interested in that. Much more plausible--and consistent with what it feels like I'm doing--is that once we're into the money, I turn up the aggression, trying to get a dominating stack for going into heads-up play, but overdoing it and instead fizzling out for the min-cash, thus explaining the excess of third-place endings.

What sucks even worse is that I'm not beating the rake. I have slipped over the break-even money line--in the bad direction. I have spent $2260 on buy-ins, and been paid just $2204, for a net loss of $56, and a ROI of a dismal 0.98. It's a trivial sum in the grand scheme of things, but, geez, it would sure be nice to have something to show for all those hours spent.

There is hope, though. I'm reasonably confident that I know what my two biggest leaks are. Plugging them is just a matter of relatively minor discipline, not some huge overall change in game plan.

Will I manage to actually do it?

I'll let you know in another year or so.


genomeboy said...

Contrary to your comments, I actually find your poker results related posts interesting.

However, I'm not sure that even 200+ tournaments is a statistically big enough sample size to make any comment on. The discourse on 2+2 (of course not the be all and end all, but still pretty good) suggests that at least 5 fold more sngs may be required before a true understanding of your skill and ROI.

Also, I'm sure you and I could have a fun discussion on the relative merits of selecting a p value of 0.05, but to remind the faithful, basically, that level suggests that you would get a positive association (in this case that you are a winning player overall) 1/20 when the actual case is that you are not a winning player.

Heffmike said...

tldr... :-)


1) way too small of a sample size

2) structure in these HORSE SnGs suck across the board - way too quick and too much of a crapshoot.

I tried playing them a few times - and decided anything else was a better use of my time.

You might as well try and figure out why you don't win as much as you should on the red/black line on roulette.

HORSE SnGs are fine to blow off steam and donk around. And that's really about it.

astrobel said...

Did you get rakeback at all ?