Re: [Bug-gnubg] Use of Two-Sided DB in matches - I agree, this looks suspicious.

[Michael Petch]

You know something somewhere along the line I got the scores wrong. . Now
that you point it out, its an automatic recube for match for the person that
is trailing (I thought it was to 7pts). The issue for Gnubg is that at 0 ply
its a cube/drop. That is fascinating in itself because it differs from XBG
by a wide margin.

I¹ll ask this to the other devs. Using our formula  on 0 ply is there a
reason why it should differ so much different to XBG? I am excluding the
other ply levels because there is a basic fundamental difference somewhere
even with no lookahed using what appears to be equivalent data in the
twosided databases.

Now that I realize what the score is, I have to admit XBG's result looks
right (Sorry Neil for all the trouble so far lol). It simply appears like
Gnubg hasn't accounted for the automatic recube. Gnubg has the same cubeless
equities but computed wildly different cubeful ones from the same data.

To recap XBG says this on 1 ply (Gnubg 0ply)

Analyzed in 1 ply
Player Winning Chances:    73.53% (G:  0.00% B:  0.00%)
Opponent Winning Chances:  26.47% (G:  0.00% B:  0.00%)

Cubeless Equities: No Double=+0.471, Double=+1.549

Cubeful Equities:
      No Double:   +0.550 (-0.450)
      Double/Take: +1.435 (+0.435)
      Double/Drop: +1.000

Best Cube action: Double / Drop

-----------

Gnubg Gives this:

Cube analysis
0-ply cubeless equity  +0.470665 (Money:  +0.47066 +0.470665)
  0.735332 0.000000 0.000000 - 0.264668 0.000000 0.000000
Cubeful equities:
  0-ply cubeful [expert]
1. Double, pass         +1.000000
2. Double, take         +1.460251  ( +0.460251)
3. No double            +0.757803  ( -0.242197)
Proper cube action: Double, pass





On 05/09/09 3:36 AM, "Neil Robins" <address@hidden> wrote:

> Gives directly the probability of winning the match after Double/ Take and
> Automatic Redouble.