Quote:
What is the minimum number of hours of live play needed to determine whether a player is a winning/losing player with a reasonable level of confidence?
I’m going to estimate 200 hours (5000 hands if assuming 25/hour) is enough to be reasonably accurate. For a clear winner/loser, perhaps 80 hours (2000 hands) is enough?
When I got back into poker about 4 years ago, I wondered the same. I had been an online grinder back in those magical pre-Black Friday days, and hung it up not long after BF without really giving live poker much of a shot (really dumb, in retrospect.) I wasn't really satisfied with the stock answers I got, though... usually I have read 1000 to determine if you are a winner, 2000 to get a ballpark of your winrate (which, is also flawed, because your player pool and you yourself + your skills and and a thousand other intangibles are always in a state of flux.)
Fortunately, when I began tracking my results, I opted to build a custom spreadsheet in Google Sheets as opposed to using one of the available smartphone apps. This allowed me the ability to graph a running $/hr or BB/hr for a given # of hours; something that is not available in the poker tracking apps, to my knowledge.
First, a simple graph of my BB/hr. Converges to 13.6 BB/hr over 2650 hours. I had some wicked variance in a stretch in the earlier part of that graph, haven't run quite that bad since that time, but definitely have progressively gotten better too which is pretty evident even here.
Ok now, we'll get into the more illuminating graphs that will paint a more complete picture of the variance over this span. These will graph the hourly of the last X # of hours, starting with the first available stretch of that sample. First, we'll do 2000 hours, the typical stock answer to "how many hours do I have to play to know my winrate." For the sake of clarity, note that it will start at hour 2000, so the first data point on the graph is my winrate over the first 2000 hours, and the last data point (at hour ~2650) encompasses my hourly in the most recent 2000 hours (thus excluding the first 650.)
I have certainly improved as I've gained experience, but I also ran hot as hell last year and gradually my particularly bad stretch in the early days is weeded out toward the end, so I would guess this will make its way back to the 13/14 BB/hr range? Obviously gonna depend on a lot of external factors. Let's check out the same graph but for a running BB/hr for the last 1000 hours.
Do I think it is reasonable to expect my winrate to be 18.1BB/hr as it has been for the last 1000 hours? Lolno. I have run very well, and I must say that I have played generally deeper games the later in the results I go... more 5/5 1k cap and 5/10 1000-3000 in the very last 1k hours and mostly 2/5 500 cap in the first, so that is certainly a factor. But realistically, I do expect some rough waters ahead will take that back down to earth some.
Now, let's drill down into some smaller numbers and see just how low a number I will need to put in before I have a breakeven/down stretch over a given span of hours.
It appears my lowest BB/hr for a 500 hour span so far was about 5.6BB/hr, and at the time my total about 8.2BB/hr over ~820 hours. I also spot a local minimum of about 9BB/hr over the last 500 hours at hour ~2070, and at that time my lifetime w/r was somewhere around 12BB/hr. So we can conclude, if nothing else, that it is easy to run 3BB/hr below expectation over a 500 hour sample. Thus, if you are losing less than 3BB/hr over your first 500 hours, and you are studying and improving, you might well be a long-term winner who has been dealing with the bad side of variance. Now, let's cut that back some more check on BB/hr over a running 275 hour sample:
Yahtzee. The largest sample of hours that I have broken even (in BB terms) is 275 hours. Once I put in some study and stopped running so godawful I haven't come close to a breakeven that long, but even there I was winning at a pretty good clip of 8BB/hr, so you can see that 275 hours is a fairly meaningless sample. OK, let's knock it down some more and find some pain!
Now we're talking. Even in the latter part of the graph when running what is ostensibly well above my long term EV, 125 hour breakeven stretches are commonplace. 125 hours is what I figure to be the max a full time grinder can put in per month without eventually going crazy (YMMV.) So yeah, even if you are crushing full time, you will have some breakeven months. Let's dial it down one more time to a weekend warrior / part-timer kind of schedule and see what a running BB/hr of 50 hours looks like:
Yeah, part timers are gonna have a rough time if they expect to win every month, no question about it.
Anyway! I hope some people find this real-world variance analysis useful, I wish I had this kind of information when I was getting into the live poker streets. Before anyone asks, this is about 95% 5BB and above, and about 99% NLHE (there's a tiny bit of 5/5 mix and 2/2 PLO in there, not enough to make a difference.)