Exact Split EV

[MangoJ]

I'm interested in exact (that is composition-dependent) EV for splitting hands.
I'm currently confident in calculating EVs for Stand,Hit,Double (and Surrender:laugh using a combinatorial drawing-tree with recursive strategy decisions.

However as each Hit needs evaluation of thousands or more (never counted them) different hands, an exact split would need millions (or even billions) different hands (and thats not accounting for re-splits).

Is this the price to pay - or are there simpler methods available ?

 

[nightspirit]

I'm interested in exact (that is composition-dependent) EV for splitting hands.
I'm currently confident in calculating EVs for Stand,Hit,Double (and Surrender:laugh using a combinatorial drawing-tree with recursive strategy decisions.

However as each Hit needs evaluation of thousands or more (never counted them) different hands, an exact split would need millions (or even billions) different hands (and thats not accounting for re-splits).

Is this the price to pay - or are there simpler methods available ?

I only know that there have been some deep discussions over the subject over at the bjmath.com-forums. Unfortunately the link from this site http://www.bjmath.com/forums.htm (Archive copy) does'nt seem to work. (at least not in my browser...)

But you can access the old post by just using Google.

For example, searching for:

site:http://www.bjmath.com (Archive copy) split EV MGP Cacarulo

should give you some results

 

[gronbog]

I'm interested in exact (that is composition-dependent) EV for splitting hands.
I'm currently confident in calculating EVs for Stand,Hit,Double (and Surrender:laugh using a combinatorial drawing-tree with recursive strategy decisions.

However as each Hit needs evaluation of thousands or more (never counted them) different hands, an exact split would need millions (or even billions) different hands (and thats not accounting for re-splits).

Is this the price to pay - or are there simpler methods available ?

I don't know if there are quicker methods available, but to me recursive iteration is the simplest. My own sim uses this method and it converges on the correct play and EV reasonably quickly.

FWIW, most computations arrived at via simulation require billions of hands to converge.