How to randomise split decisions at the poker table


Some hands need to be played several ways to avoid being exploitable. Dara O’Kearney shares some heuristics to keep your opponents guessing.

Dara O’Kearney 

In GTO we not only split our range by betting small, betting big and checking with different hands, we also split the actions we take with the same hand. Sometimes we bet big with the nuts for value, sometimes we bet small with the nuts to protect our small bet range, sometimes we check with the nuts to protect our checking range, and so on.

Often the solvers will split these actions based on a suited blocker card. If, for example, we have KK on a queen high flop with two spades, if our KK does not include a spade we might be more likely to bet big with it because our hand benefits from denying equity to flush draws. If our KK includes a spade we might bet small or even check with it, because the flush draw worries us slightly less because we have a blocker.

Most spots don’t require you to split your actions, but many do. So how do you, in game, decide when to split an action up when the suit of one of your cards does not impact the action? You cannot realistically know at the tables whether a particular action is a bet 57% of the time and a check 43% of the time, nor can you keep accurate track how how frequently you have been doing one action over the other. 

Random Number Generators

Dara O Kearney
The clock is your friend

Many GTO players have a system whereby they use some sort of randomiser to decide. I have a random number generator on my PC. If, for example, I think I should bet big with my hand 60% of the time, bet small 30% of the time and check 10% of the time, I will assign ‘bet big’ numbers 1-60, bet small numbers 61-90, and check numbers 91-100. Then the random number generator will pick a number between 1 and 100 that decides my action.

If I am playing live I will do something similar by looking at the clock on the wall. For example the seconds hand, in the example above I might say if it is pointing at 1-30 seconds I pick bet big, at seconds 31-55 I pick bet small, and at seconds 56-60 I check.

Before I randomise there is another, more practical, way I decide when to split my action, which is when I have a read or a HUD stat on a player. If, for example, my read tells me I am up against a calling station type of player I might choose the bet big option with strong hands and the give up option with my bluffs. If the player is a nit I might choose the bet small option with big hands making a call more likely and bet big with bluffs.

Any system is better than no system

Dara O Kearney
Keep doing the same action and you will be exploitable

This might seem strange, as it is essentially an exploitative strategy for playing GTO. My argument is that if you have to randomise anyway, you may as well use whatever minor help a read will give you to your advantage.

I’ll even do this when I have a very small sample size of HUD stats or a read based on not much yet. I am still essentially randomising over the population of players I am facing, just not this particular player. If my read is the player is quite bad, it will matter the least anyway because they will not be thinking about balance.

Another approach which is quite ad hoc is to recognise when you have been taking one action a lot recently so you do the opposite for a few days. If you have been betting your flush draws a lot recently, make a pledge to check with them for the next two days. Be aware this approach is likely to lean into your own biases as a player, if you are aggressive you will probably take the more aggressive option most of the time and vice versa.

None of these methods are perfect, but you will never know what the right frequencies are in a particular spot when it happens at the table and it would be a waste of mental bandwidth to try. The only real mistake is to take the same action again and again in similar spots. People will notice and you will become exploitable.

Dara O’Kearney’s new book Endgame Poker Strategy: The ICM Book is out now.





Source link

Back to top button