Saturday, May 9, 2015

Continuation Bet. A Lot.

Being the nerd I am, I have actively collected a lot of poker data over the years. A whole lot. In the past ten years alone, I’ve played well over a million hands combined of online Texas Hold’em, Stud, and Omaha. The majority of these hands and/or sessions were recorded either via automatic tracking software or manually in a long series of spreadsheets. I have also added other players’ hands to this database, actually going as far as purchasing blocks of data from various online re-sellers. All-in-all, I have data on over three million real poker hands that I have used to help evaluate and test theories, and prove (or disprove) specific strategies and tactics.

I’ve used this data for everything from helping put together preflop starting hand recommendations, to quantifying the power of position, to determining optimal bet sizing in different situations, and even to help me figure out the best times of the day and week to play that are the most profitable for my own specific LAggy style. As the old saw goes, you can’t improve what you can’t measure, and all this measured data I’ve collected has served to greatly improve my own game.

Lately I used these data to analyze flop continuation betting scenarios. This analysis resulted in an honest-to-goodness formula a beginning player can use to determine whether they should c-bet in any specific situation or not. The method itself is quite simple: you calculate your fold equity based on a point count system, then do the same for your pot equity, and finally just multiply the two numbers together and look at the result. Depending upon the number you get, you then either fold, c-bet, or slow-play your hand…

….ah, but all of this can really be boiled down to a much, much simpler let’s-cut-to-the-chase recommendation:

When in doubt, you should probably continuation bet

The simple fact is this: in small stakes games, if you were the preflop aggressor, you should be firing a continuation bet on the flop at a relatively high frequency. Depending on the number and type of villains in the hand and the texture of the board, this number can range from 60% to higher. Heck, against weak-tight nits on super-dry boards, you should probably be firing a flop continuation bet nearly 100% of the time.

This empirically-based conclusion of mine can be proved analytically by doing some simple algebra with a basic expected value equation. Setting EV equal to zero and ignoring any showdown equity you might have, one can easily see that Fold Equity = Bet Size / [Pot Size + Bet Size]. Here’s that relationship plotted out in X-Y fashion:

What this chart demonstrates is that you only need a villain to fold 33% of the time if you fire a half-pot-sized continuation bet on the flop. Given the fact that the villain is going to miss the flop around 35% of the time, this means a c-bet generally has a high percentage chance of working.

Now, the bad news is that actual (i.e, not required) fold equity will vary as a function of Hero's c-bet size, which is something we’ve ignored in this simple analysis. In other words, a small bet size may only need a small villain fold frequency to be break-even, but villains will fold significantly less frequently if they’re being offered good pot odds, which a small bet causes. At first blush, this argues in favor of c-betting less frequently on the flop, but the other thing we’ve left out of the analysis is that we have ignored showdown equity for all those cases when the c-bet doesn’t result in a fold. These two factors (more or less) cancel each other out, which means that roughly speaking, you should in fact be continuation betting a relatively high percentage of the time on the flop.

How high, you ask? My three million hands of empirical data suggests a continuation bet frequency around 75%. Said another way, you should be flop continuation betting as a bluff an average of three out of four times in small stakes games. Further, I suggest that your c-bet size be around 60-65% of the pot; this seems to be the optimal risk:reward sweet spot that applies enough pressure to the villain, while risking the least amount of the hero's stack if the villain calls.

Bottom line: Three million hands of data don't lie: continuation bet most of the time in small stakes games, and make your bet sizing to be around 60-65% of the pot.


  1. Does your conclusion differentiate between cases where there is only one villain, and cases where there are two or three or more?

  2. Dave, my "bet 75% of the time with a sizing of 60% of pot" rule is a very generalized recommendation. The more villains you face, the less likely you should bluff c-bet (primarily because it's much likelier that someone connected with the board).

    On the other hand, Bart Hanson says that in live $5/$10 games, he generally gets more folds when he faces two villains, vs. one. The supposition is that if you're betting into two villains, you probably have a made hand, so they give you more respect, but betting into one can easily be a standard bluff c-bet. He also says that 3 or more villains in the pot mean you should rarely bluff cbet...

  3. Hi Bug, coupla comments : -
    = is this valid only for cbetting in position or is it valid even when OOP?
    = Regarding the reference to small stakes, does that hint at the player proficiency / skill expected at this level?
    = ok and as a bonus 3rd, what online stakes do you class as small? <$1/$2??

    Thanks Bug.

  4. Rob,
    1) In general you should be cbetting somewhat less when OOP than when IP. It's not a huge effect, however. Maybe 5% or so less frequent.
    2) Yes, absolutely. Lower stakes generally mean lower skill, which will affect your cbet frequency.
    3) Online, I classify micro-stakes as $0.05/$0.10 and under. Small stakes is $0.25/$0.50 and lower. A method to convert this to live play is to multiply by 10x to get the equivalent live stakes. E.g., low stakes live would be $2/$5 live. Higher than this (e.g. $5/$10 live, or $0.50/$1.0 online) is getting into mid-stakes, which means better players.

  5. This comment has been removed by a blog administrator.