[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Thu, 3 Apr 2003 20:43:25 +0000
gnubg should now be able to play hypergammon.
The CVS repository now includes a new program called makehyper that can
generate the necessary databases for hypergammon with 1, 2 or 3
The generation of the databases are, unlike the usual bearoff databases,
not exact, but is rather an iterative process.
Database Size Total Time Time/iteration
hyper1.bd 18,968 <1s < 1s
hyper2.bd 3,449,628 ~20 min 39s
hyper3.bd 300,500,968 unknown 6400s
The reported times are for a 1GHz PIII processor. Unfortunately I didn't
record the time for the hyper3.bd database, but it used some 40-50
iterations of 6400s each, so that's 4+ days of CPU time :-)
Generation of hyper3.bd requires approximately 400MB of memory.
The databases includes gwc + gammon percentages (for cubeless non-Jacoby
money game), and cubeful money equities (centered cube, centered cube
with Jacoby rule, owned cube, and unavailable cube).
For match play gnubg will use gwc + gammon percentages and feed those
into the usual formulae we use to calculate cubeful equities. This part
is not tested thouroughly yet.
- [Bug-gnubg] hypergammon,
Joern Thyssen <=