HU Theory [0,1] Game - Splitting your range

Posted by

Posted by posted in Mid Stakes

HU Theory [0,1] Game - Splitting your range

Lets assume a game of [0,1] with symmetrical ranges from 0-100 percentile.

Both players antes 1$ and oop can only check, bet or call. No check-raising is allowed. Only potsize bets are allowed.

There will be 3 streets of betting. (Similiar to HUNL with restricted rules)

As the button:

For the first round of betting, would you rather:

1) Bet the top x% of your hands and check back and fold the bottom y%.

2) Bet the top x% and bottom y% and check some z% in the middle and have a call/bet range on the 2. round.

3) Some more complex split, which likely will bet/check hands in all of x/y/z %


If anyone want to make an attempt at assigning actual numbers to each x/y/z etc. or would like to discuss splitting ranges with 3 rounds of betting left, then that would be awesome too.

Please give some reasoning for the choice you make though.

If you look at the flop in isolation, you would need a hand in the top 25% to valuebet, as oop will fold bottom 50% to a potsized bet.


Edit: A forum dedicated to poker theory would be great too.

Edit: OOP is forced to check the first round, button is only allowed to bet, check or call on any round. OOP can bet 2. round if button checks first round. OOP can bet 3. round if button checks second round or both of the first two rounds.


17 Comments

Loading 17 Comments...

Sauce123 11 years, 8 months ago

Well put question Teddy.  I think you'll need to provided some sketches at answering it yourself though before people are going to volunteer this type of info. 

Juan Copani 11 years, 8 months ago

I have my own sketches of how i would split my range in some type of textures. Id show you here if i see any commitment on the answers, and people could give me any usefull information.. ! But im not sure if this gonna happend.. 

For now i can say that i usually split my range in order to my opponents weakness. Would be nice if i could chose always the option 2), bet all my air getting profit, but against some thinking regs i will have to resign some part of my range in some board textures, having to chose option 1).

Would be nice could start to design some starting points of splitted ranges, and then start to deviate of them in order to our opponents weakness.

GameTheory 11 years, 8 months ago

Lets start the discussion by giving the solution to the one street version of this game:

Pot P=2, there is one pot sized bet of size 2 left. 

Both players can check, bet, call and fold. Lowest number wins.


OOP will:

Bet: [0, 0.17] 

Check,call: [0.17, 0.50]

Check,fold: [0.50, 0.92]

Bet: [0.92, 1]

Thus the OOP player will bluff his ~8% worst hands, bet his strongest hands and check three quarters of his range.

BTN will:

Call: [0, 0.50] when facing a bet

Fold: [0.50, 1] when facing a bet

Bet: [0, 0.33] when facing a check

Check: [0.33, 0.83] when facing a check

Bet: [0.83, 1] when facing a check.

Hence, the button bets a polarized range. The cutoff value betting hand 0.33 is the mean of the check,call range [0.17, 0.50]. And the ratio between value bet and bluffs is 2 to 1, which offers exactly the right pot odds.

Note that all numbers are actually fractions of the form n/12, but are shown as percentiles.


If you look at the flop in isolation, you would need a hand in the top 25% to valuebet, as oop will fold bottom 50% to a potsized bet.

In this statement bot the premise and the conclusion are incorrect. Firstly, for OOP to be folding the bottom 50% of his range, he needs to make it a losing or break even play for the BTN to bluff with his dead hands, but there is not reason to believe that BTN will not bluff his worst hands. This was exactly what he was doing in the one street example, and what he should be doing. When you check down your worst hands, the expectation to win the pot is very low. And when they bluff successfully they win the entire pot. Therefore the worst hands gain the most by bluffing.

Secondly, in a multistreet game it is no longer enough that you have 50% against the continuing range of your opponent. Since by betting you blow up the pot by a factor of 3, and you will still need to defend the bottom of your range on future streets. Calling bets on future streets after betting is much more expensive.

If you have 51% equity on the flop and you bet flop, check turn and call a bet on the river, you put in 2+6=8 (flop+river). While you only tried to win a bet of 2 on the flop. And if your opponent starts betting turn and river this becomes 26 to 2. The 1% equity advantage is most likely not enough to warrant this.

Optimally protecting all your ranges can be very difficult, even in a simple static game like this.


Juan Copani 11 years, 8 months ago
OOP will:Bet: [0, 0.17] 
Check,call: [0.17, 0.50]
Check,fold: [0.50, 0.92]Bet: [0.92, 1]
Thus the OOP player will bluff his ~8% worst hands, bet his strongest hands and check three quarters of his range.
BTN will:
Call: [0, 0.50] when facing a bet
Fold: [0.50, 1] when facing a bet
Bet: [0, 0.33] when facing a check
Check: [0.33, 0.83] when facing a check
Bet: [0.83, 1] when facing a check.

I don´t understand how you get to all this numbers...

Teddy 11 years, 8 months ago

Thanks for correcting me GT and pushing the discussion in the direction I was hoping for. I didn't realize you had replied before I edited my post the second time, so everyone should feel free to ignore or include that.

eq.fest 11 years, 8 months ago
OOP will:Bet: [0, 0.17] 
Check,call: [0.17, 0.50]
Check,fold: [0.50, 0.92]Bet: [0.92, 1]
Thus the OOP player will bluff his ~8% worst hands, bet his strongest hands and check three quarters of his range.

GameTheory I'm interested in how you came to having OOP value bet the top 17% of hands. I assume at that point the EV of checking that hand must be equal to the EV of betting that hand. I tried some calculations of my own to get that number but was getting some garbage answers. I took some shortcuts and assumed that whenever we checked we effectively lost the pot as IP would bet with a balanced range. Is this right?


GameTheory 11 years, 8 months ago

If you hold 1/6 (=0.17) and you decide to check and call your EV is:

Lose 3 (=1+2) versus [0, 1/6], win 3 versus [1/6, 1/3] + [5/6, 1], win 1 versus [1/3, 5/6]

EV = (1/6)*(-3) + (2/6)*(+3) + (3/6)*(+1) = 1.


And when betting:

Lose 3 versus [0, 1/6], win 3 versus [1/6, 3/6], win 1 versus [3/6, 1]

EV = (1/6)*(-3) + (2/6)*(+3) +(3/6)*(1) = 1.


Hence, at exactly 1/6 you become indifferent between betting and checkcalling. That is why 1/6 is the cutoff between the intervals at which you bet and checkcall.

Teddy 11 years, 7 months ago

"OOP will:

Bet: [0, 0.17] 

Check,call: [0.17, 0.50]

Check,fold: [0.50, 0.92]

Bet: [0.92, 1]

Thus the OOP player will bluff his ~8% worst hands, bet his strongest hands and check three quarters of his range.

BTN will:

Call: [0, 0.50] when facing a bet

Fold: [0.50, 1] when facing a bet

Bet: [0, 0.33] when facing a check

Check: [0.33, 0.83] when facing a check

Bet: [0.83, 1] when facing a check."


I was curious why BT doesn't bet more when OOP only calls (.50-.17)/(.92-.50+.50-.17)= 44%

Fx. EV check .82 = lose 1 vs [.17-.82] win 1 vs [.82] win 2 vs [.82-92] = 2(.92-.82)+(.82)-(.82-.17)= .37EV 

Ev bet .82 = lose 2 vs 44%, win 2 vs 56% = .56*2-.44*2 =.24EV 

Ev check .83 = lose 1 vs [.17-.83] win 1 vs [.83] win 2 vs [.83-92] = 2(.92-.83)+(.83)-(.83-.17)= .37EV

Ev bet .83 = lose 2 vs 44%, win 2 vs 56% = .56*2-.44*2 = .35 EV

Did I make some error when calculating [.83] since the EVs of betting and checking aren't equal or is it just rounding ?


acidboy 11 years, 8 months ago

I see a problem with your solution, GT.

When you bet [0, 0.17]  of your range and get called by [0, 0.50], you are assuming OOP is valuebetting and winning always the pot. So, if you have 2/3 valuebets, 1/3 bluffs, EV equation for IP player when calling, assuming PSB, is:

EV=%OOPvaluebets * (Eq*Pot-Pot)+%OOPbluffs * (Pot)


So you are doing

 0.66(-2)+0.33(4)=0

But the thing is, when [0, 0.17]  bets against a calling range of [0, 0.50], hes gonna win around 2.5/3 of the times (so it has 83.33% equity). Thats because 2/3 of the times the calling range has hands between [0, 0.50], and 1/3 of the time they are going to show up with the same range. So, the EV equation for IP player should be in this example:

0.66(0.166*2-2)+0.33(4)=-1.11111+1.33333=0.22222.

This number obv should be 0.



GameTheory 11 years, 8 months ago
When you bet [0, 0.17]  of your range and get called by [0, 0.50], you are assuming OOP is valuebetting and winning always the pot.

No I don't think so. Why should I make this assumption? I am assuming that OOP maximizes his expectation against the BTN strategy.

OOP bets a polarized range, any hand between 0.17 and 0.92 that the BTN can hold will be indifferent between calling and folding. And any hand from the OOP valuerange will not be indifferent since it will beat some value hands, hence they must call!

If you want to show that my solution is wrong, all you have to to is to pick an interval from either player, suggest a different strategy than mine with this interval. And then you must improve the expectation of the hands in this interval.


For instance when OOP holds [0.92, 1] currently he gets folds half the time and calls half the time. Every time he gets called he loses, so the expectation of this interval is exactly zero.

If OOP were to check and fold with this interval he will never win, since all worse hands that the BTN could hold will bluff. Hence checking and folding loses the ante of 1 in expectation, checking and calling is obviously worse. Hence betting is optimal for OOP against the BTN.

If you can show one interval like that where the strategy from my solution is not optimal, then my solution must be wrong!

GameTheory 11 years, 7 months ago
I was curious why BT doesn't bet more when OOP only calls (.50-.17)/(.92-.50+.50-.17)= 44%Fx. EV check .82 = lose 1 vs [.17-.82] win 1 vs [.82] win 2 vs [.82-92] = 2(.92-.82)+(.82)-(.82-.17)= .37EV Ev bet .82 = lose 2 vs 44%, win 2 vs 56% = .56*2-.44*2 =.24EV Ev check .83 = lose 1 vs [.17-.83] win 1 vs [.83] win 2 vs [.83-92] = 2(.92-.83)+(.83)-(.83-.17)= .37EVEv bet .83 = lose 2 vs 44%, win 2 vs 56% = .56*2-.44*2 = .35 EVDid I make some error when calculating [.83] since the EVs of betting and checking aren't equal or is it just rounding ?


The ante is 1, so by checking you either win or lose 1, never 2.

And the bet is 2, so by betting you either get a fold and win 1, or get a call and win or lose 3, never 2.

You could consider the ante of 1 sunk cost, but then you have to be consistent.


Teddy 11 years, 7 months ago

Thanks! I believe I found my error and this is correct:

As BTN:

If you hold 5/6 (=0.83) and you decide to check your EV is:

Lose 1  versus [2/12, 10/12], win 1 versus [10/12, 11/12]  (9 intervals)

EV = (8/9)*(-1) + (1/9)*(+1) = -7/9

And when betting:

Lose 3 versus [2/12, 6/12], win 1 versus [6/12, 11/12] (9 intervals)

EV = (4/9)*(-3) + (5/9)*(+1)  = -7/9


Be the first to add a comment

Runitonce.com uses cookies to give you the best experience. Learn more about our Cookie Policy