Stealing, Take 2*

Back in early October, I mused in this blog that the "best" steal rate percentage in NL cash games may actually be lower than that which I was currently stealing at the time. Specifically, I was around 40% back then, but after reviewing a million hands of villain data, I saw that the top players from that data set were stealing, on average, at around 35%.  I concluded from this that 35% was probably the "best" steal percentage target to shoot for.

Well, opinions change.

In my last post, I mentioned a poker writer/player named Matt Flynn, whose podcast I greatly enjoyed. In his writings and podcast, Flynn advocates a high percentage stealing approach to the game. He argues (convincingly) that whenever you're not stealing, you're passing up potential dead money that someone else is going to claim if you don't. Successfully stealing just a few extra times per hundred hands can have a significant effect on your overall hourly win rate. In other words, Flynn argues, if you're not stealing a lot, you're simply passing up a lot of easy profit that could be yours.

After a bunch of reading, surfing, and errant thinking on the subject of stealing, I'm revising my original conclusions on steal percentages. In fact, I now think that the "correct" steal rate at my NL cash stakes is probably somewhere north of 65%. With that in mind, I've been experimenting with this idea of upping my steal percentage during these first 12 days of November. During this (albeit short) period, I've increased my steal percentage from 38% to a sky-high 75%. In fact, the only time I'm not stealing when presented with the opportunity is when the situation is grossly unfavorable; i.e., when the blinds' stats tell me I shouldn't even think about trying it (more on this in a moment). Anyway, here are my results for the month thus far:

As you can see, the numbers have been pretty good*. Of course, I've only got a couple thousand hands of data, so I can't read too much into the graph, but it is very interesting to me that my red line, which is normally in the negative, is now the bulk of my profit. Looking further at the data shows me that this is due in large part to steal attempts that paid off. Here's another look at the data, where I sorted by my steal attempts, minus all pairs and big aces (i.e., true "steal with any two cards" situations):

 

Holy crap. Nearly 45% of my profit has been from stealing with rags. Wow. How is this possible? Is it sustainable? Dunno, but I'm going to keep up this approach until the data says otherwise.

In the meantime, here are some miscellaneous thoughts on two things I've figured out thus far about stealing: namely, that the trick to making crime profitable seems to be a combination of bet sizing and reads.

Bet Sizing.  We generally bet for two reasons in poker**: 1) Get a worse hand to call; and/or 2) Get a better hand to fold. When we're trying to achieve the former (i.e., value betting), our goal should be to bet the largest amount of money our opponent will call***. For the latter, we want to bet the minimum amount that will get the opp to fold****. In other words, we want to optimize our risk-to-reward ratio.

Said another way: when stealing, we want to bet the minimum amount that gets the blinds to fold. I.e., if we get RR'd, and we're holding rags, we can fold with a minimal loss. How important is this? Well consider that the difference between stealing with a 2.5xBB raise and a 3xBB raise is half a big bet. In a typical Rush game, you might get the opportunity to steal ~30 times per hour. This means you're risking 30 x 0.5 x BB = 15BB more per hour of your profit for the same potential reward.

Now of course the trick to all this is you have to bet enough that it puts pressure on the opp to fold their middling hands and not defend. Min-betting for instance, may improve the numerator in the risk/reward fraction, but it's still a lower EV play. Said in plain English: betting too little will cause the opp to defend too much. The secret is to find the stealing size sweet spot for the stakes you play at.

When I was at the $5NL and $10NL tables, getting the villains to fold their blinds for anything less than 3xBB was a waste of time. A common beginner's leak is to defend the blinds too lightly, so you end up getting called a ton if you bet too small. Getting the blinds to fold at the micro-stakes requires a relatively big hammer.

Now that I'm playing at the $25NL and $50NL tables, however, I see that the beginner's leak of defending is less prevalent. Yes, some players defend more than others, but generally speaking, you can steal with a 2.5xBB bet just as easily as with a larger bet at my current stakes.

Villain Stats. The other key factor for me in stealing is deciding whether it will work or not against the villains in the blinds. This, of course, means that you have to have info on the opp, which means PT3/HUD stats.

For me, the most important stat is the FTS, or Fold to Steal number of the players in the blinds. This is so important to me that I've color coded it (red if it's less than 33%, yellow if it's 33-66%, and green for go if it's greater than 66%). Further, I've increased the font size and moved the stat to the top line of the HUD so that I can see this number quickly and correctly when I'm multi-tabling in Rush.

The next stat that I care about is the number of hands of data I have on the opp, for obvious reasons. If the villain has a 100% FTS number, but I only have two hands of data on him, it's kinda meaningless.

The third thing I look at is any notes I have on the player. I'm still experimenting with this, but I'll probably come up with a color code for players who defend a lot.

Next up is a quick look at VPIP and PFR. Tight, nitty players are much more willing to give up their blinds than regs and LAGs. Yes, this is a generalization, but I think it's fairly accurate.

I'm also looking at stack sizes before pulling the trigger on a steal. It's been my experience that short stacks tend to defend (often by re-shoving) than big stacks. Dunno exactly why this is, but it seems to be very common at the $25 and $50NL stakes. Maybe it's just that the bad players have the short stacks, or maybe it's a strategy of the short-stack specialists. Again, dunno.

Finally is the ATS, or Attempt to Steal stat. This gives me verification that the player understands (or doesn't) the concept of stealing.


Well, that's all the time I have today on this long-winded post.
All-in for now...
-Bug
*The results are good... except for the awful, awful poker I was playing at the very beginning of the month when I was traveling and trying to play at 3am local times and in strange hotel beds. Just look at that see-saw variance in the first 500 hands of data . Note to self: don't play poker when on biz trips...

**There are other, secondary reasons for betting (such as pricing out draws, gaining information, blocking, semi-bluffs, etc), but these two biggies cover the vast majority of reasons to bet in poker.

***Actually, this isn't entirely accurate. We want to bet the amount that maximizes our profit. This means we want to pick the bet size that gives us the maximum expected value return on our outlay. If our opponent only calls, say, 10% of the time to a huge over-shove, but that 10% multiplied by the expected profit is greater than other bet sizes multiplied by %'s, then this is the "best" bet size for the situation.

**** The same argument can be applied here; we want to maximize our EV.