I’ve seen in a few places some version of the formula

live cube takepoint = dead cube takepoint * cash point

(e.g., http://www.bgonline.org/forums/webbb...es;read=161954)

This seems reasonable to me, and gives the right answer for money and a few match scores I tested. However, I can’t find a source for this.

If we define the live cube takepoint as the cubeless win probability at which our cubeful win probability is equal to the dead cube takepoint, then the formula above makes sense *if* the following is also true:

cubeful win probability = (?) cubeless win probability / cash point

This also seems reasonable, but I can’t rigorously back this up either. For example, in a money game where our cash point is 80%, if we have 75% cubeless win probability, it seems reasonable that our cubeful game winning probability would be around 75/80 or 94%. However, this seems like it would depend on the position, volatility, gammons, etc so I’m not sure whether it’s true in general.

Any insight on where the first formula I quoted might come from?

The live cube TPs reported in XG are 'fully live' numbers, i.e. they assume a 100% recube efficiency.

If you make this assumption, then this formula is accurate.

In reality, cube efficiency will vary significantly based on the factors you listed like volatility and gammon threats.

Unless you have an auto-reship situation, looking at a position and attempting to estimate an actual cube efficiency, and calculating an 'actual' take point is a difficult task.

Hope that helped.

btw, one of the several features that GNU has, but XG lacks, is the ability to compile/view stats after a rollout.

Using GNU is the only way I'm aware to put a hard number of cube efficiency. If there are other methods, I'm sure someone will let me know.

I might misunderstand you, but AFAIK GnuBG uses 70% cube lifeness and I assume that XG does something similar. cube lifeness of 100% means that you are you're either too good or not good enough and only by accident on the doubling point.

The Paper of Janowski is to my knowledge the only theretical base for that calculations (I'm glad if someone has additional information) and a good read if you're interested in

Hi bgblitz- Thx for your reply.

My understanding, which may be inaccurate, is that when you rollout a checker play w GNUbg or XG, the evaluations displayed are 'pseudocubeless' results with cube efficiency preset at 70%, just as you have stated.

100% cube efficiency occurs when your opponent has an optional take/pass (the cash point), and is why OP's original formula works.

Also, when you rollout a cube decision in GNUbg, after the rollout is complete, a box appears labelled 'View Statistics'. If you click here, it displays cubeful info on where/what happened to the doubling cube in each game of the rollout. It is from these numbers that a cube efficiency is calculated-- listed below the tables. (% eff = # of DTs/(# of DTs + # of DPs)