Game theory. People don’t get it. I don’t really get it, but I probably
get it more than 90%+ of poker pros. Part of becoming good at something is
being humble enough to realize you suck and don’t know anything. The
inspiration for this post can be found here:
where almost nobody gives a reason for their preferred bet size beyond “oh
just bet what he calls duehh”. Let’s try to figure out the most profitable optimal
betting strategy is for a similar situation, that is, a betting strategy that
doesn’t let our opponent improve by deviating from his optimal strategy and
that makes us the most money. If the players are bluffing and
calling with optimal frequencies, the 1-period (as opposed to multiple periods)
“game” of this river situation is in equilibrium and neither player can improve
by changing their play.
There’s not really a need for a hand history, so I’ll just describe the
hypothetical situation:
1.
There is $100 in the pot on the river.
2.
The players both have $200 remaining in their
stacks.
3.
Villain is first to act and checks. Villain’s hand
range is entirely made up of “bluff catchers”, or hands that beat all of hero’s
bluffs and lose to all of hero’s value bets. Villain only ever calls or folds
on the river.
4.
Hero gets to the river with 40 made hands that
beat all of villain’s hands, and 60 missed draws that lose to all of villain’s
hands. He needs to choose a bluff % and a bet amount.
Any equilibrium strategy for hero will result in villain being indifferent
between calling and folding, meaning villain will call 50% of the time and fold
50% of the time. Let’s look at what that means from villain’s perspective.
The most villain ever has to win to call a bet is 50% of the time. If
someone bets $100 into a $0 pot, you must have the best hand 50% of the time to
call and break even. If someone bets pot, you must have the best hand 33% of
the time to call and break even (.33*200 + .66*-100 = 0). If someone bets half
pot, you need the best hand a quarter of the time, etc.
Now we’ll examine the EV of different bet sizes. Let’s start with a
half pot bet. If we decide that half pot ($50) will be our bet size, we need to
have a betting range containing 75% value hands and 25% bluffs to make villain
indifferent between calling and folding. Since only 40% of our
getting-to-the-river range is made up of value hands, now we figure out what 40
is 75% of and fill the rest in with bluffs. 40/.75 = 53.333. Round that to 53…53-40
= 13 bluffs. If we bet half pot with 40 value hands and 13 bluffs, player B
makes the same amount of money calling as he does folding: $0. With the other
47 bluffs, we have to just check back and lose, or else villain can call every
time.
Half pot bet
Villain EV for calling = .25*150
+ .75*-50 = $0 = villain EV for folding
.75*-50 is the probability that we lose the pot times the amount we
lose. .25*150 is the probability that we win the pot times the amount we win.
It’s slightly more complicated to calculate hero’s EV for playing this
strategy. Four things can happen:
1.
Bet, get called, win
2.
Bet, get called, lose
3.
Bet, villain folds
4.
Check and lose
We’re checking and losing with 47 hands (also 47% of our range) so that
part is easy: .47*0 = 0. When we bet, which we do 53% of the time, we know that
villain folds half of the time: .5*.53*100 = +26.5. The other half of the time
we bet, he calls. 75% of the time he calls, we win, and 25% of the time he
calls, we lose. .5*.53*(.75*150 + .25*-50) = +26.5.
Hero EV for this strategy =
.47*0 (check back) + .5*.53*100 (bet, fold) + .5*.53*(.75*150 + .25*-50) (bet,
call) = +$53
Just as a demonstration, let’s see what happens if we decide to bluff
more and villain changes his strategy to calling every time. Let’s say we
decide to value bet 40 hands, bluff 40 hands, and check back 20 hands.
Villain EV for calling every time vs. new strategy = .5*150 + .5*-50 =
+$50.
Hero EV for new half pot strategy that lets villain call every time = .2*0
(check back) + .8*(.5*150 + .5*-50) (bet, call) = +$40.
So when we deviate from optimal bluffing frequency and start bluffing
too much against a perfect villain, villain’s EV goes from $0 to +$50, and ours
goes from +$53 to +$40. Our reaction to him calling every time would be to
bluff less and less until we end up back in the first situation.
Let’s move on to different bet sizes. For a full pot bet, villain is
indifferent between calling and folding when he has the best hand 33% of the
time, so player A will be bluffing 33% of the time. 40/.66 = 60, so our betting
range will be made up of all 40 value hands as well as 20 bluffs. We check back
40 missed draws.
Full pot bet
Villain EV = .33*200 + .66*-100
= $0
Hero EV = .4*0 (check back) +
.5*.6*100 (bet, fold) + .5*.6*(.66*200 + .33*-100) (bet, call) = +$60
Looks like against a perfect player, in this specific bluff catching
scenario, a full pot betting strategy wins more than a half pot strategy. Let’s
look at all in for 2x pot now. Villain needs to put $200 in to win the $300
sitting in the middle, so he needs to win 40% of the time to break even on a
call. 40/.6 = 67, so hero will be value betting 40 hands, bluffing 27 hands,
and checking back 33 hands.
2x pot bet
Villain EV = .4*300 + .6*-200 =
$0
Hero EV = .33*0 (check back) +
.5*.67*100 (bet, fold) + .5*.67*(.6*300 + .4*-200) (bet, call) = +$67
Betting as large as possible yields the biggest profit against a perfect
opponent in this situation. When I figure out GTO (game theory optimal) solutions like this, I like
to use them as starting points in my thought process in real situations, and
adjust according to mistakes I believe my opponent will make. If I’m playing
against someone who will call a large bet too often in this spot because lots
of draws missed, I will still bet as big as possible but will check back all of
my bluffs. If I’m playing against a very good player who may call and may fold,
I simply revert back to GTO strategy and become unexploitable. If I’m bluffing
an optimal frequency for my chosen bet size, the very good player can’t do
anything to win money against me. I don’t care how much he calls or folds, he’s
still not beating me for more money by adjusting his play.
Sometimes in a real game, you can’t make such rock solid assumptions.
If I’m betting $1,000 into a $100 pot it could go from optimal to a massive
losing play if there’s any chance my opponent slowplayed a monster hand (one
that beats some or all of my value betting hands). Most situations aren't as clear cut as in this example. Sometimes the villain check/raises, sometimes villain has monsters, air, and mediocre hands all in his range, etc.
That’s it for now. Any feedback is appreciated.
"If I’m betting $1,000 into a $100 pot it could go from optimal to a massive losing play if there’s any chance my opponent slowplayed a monster hand (one that beats some or all of my value betting hands)."
ReplyDeleteDon't have the math in front of me but if he calls correctly he needs to be doing this greater than ~0.5% of the time before 10p is the wrong betsize.
"Any equilibrium strategy for hero will result in villain being indifferent between calling and folding, meaning villain will call 50% of the time and fold 50% of the time."
This is wrong, villain will call an amount of the time which makes hero indifferent between bluffing and not bluffing. If hero bets pot that would be 50% of the time, if hero bets 2p it would be 33.3% of the time, if hero bets 0.5p it would be 66.7% of the time, etc.
You're also missing that creating a nuts/nothing betting range means a perfect opponent gets to check/raise against your smaller-sized value bets. Not that that necessarily makes it a bad play, but it does lower the equity of your lines overall against a perfect opponent compared to what you've calculated.
good call on the second point, looks like I did miss something big. luckily it doesn't change the conclusion for hero in this spot but ty for pointing that out!
ReplyDelete