BGonline.org Forums

GNU Testing

Posted By: Stick
Date: Friday, 12 December 2008, at 8:20 a.m.

There are a few things I've been wondering about GNU of late and while my tests will be far from exhaustive, I'll post the odd bits of here & there and if anyone is interested (or has already done them before) they can chime in.

The rollouts below are the same position, the same number of trials, the same seed number, but the first position was performed using my personal modified world class settings whereas the second position used the default supremo settings.

The second rollout took a full 36 minutes longer to finish than the first one. I had switched over this week to the supremo settings and noticed they seemed more sluggish than I'd have expected. This short test confirms it, that's a lot of wasted trials. (computer is a quad core 2.7 GHZ btw)

The purpose of this wasn't to test the strength of the settings but it is nice to see they have roughly the same answer. To come to any sort of conclusion I'd have to run a lot of these tests (which I won't do) but I can't imagine supremo adding enough playing strength to make up for all the trials that it would waste.

gnubg	153
user	128
Position ID: jOfCgAxj5oYDQA Match ID: QQkVAAAAAAAA

#		Ply	Move	Equity
•	1	R	bar/18	+0.2054
			0.6343 0.1462 0.0056 - 0.3657 0.0875 0.0020 +0.3308 +0.2054 0.0031 0.0038 0.0008 - 0.0031 0.0027 0.0005 0.0075 0.0110
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 811798147 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	2	R	bar/20 8/6	+0.1310 ( -0.0744)
			0.6038 0.1759 0.0029 - 0.3962 0.1011 0.0035 +0.2817 +0.1310 0.0032 0.0043 0.0013 - 0.0032 0.0021 0.0006 0.0081 0.0119
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 811798147 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 12 more moves within equity 0.16
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

---

gnubg	153
user	128
Position ID: jOfCgAxj5oYDQA Match ID: QQkVAAAAAAAA

#		Ply	Move	Equity
•	1	R	bar/18	+0.2043
			0.6339 0.1458 0.0056 - 0.3661 0.0878 0.0020 +0.3296 +0.2043 0.0031 0.0038 0.0008 - 0.0031 0.0027 0.0005 0.0075 0.0110
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 811798147 and quasi-random dice
			Play: supremo 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 16 more moves within equity 0.32
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	2	R	bar/20 8/6	+0.1298 ( -0.0745)
			0.6039 0.1761 0.0034 - 0.3961 0.1017 0.0035 +0.2821 +0.1298 0.0033 0.0043 0.0014 - 0.0033 0.0022 0.0006 0.0082 0.0130
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 811798147 and quasi-random dice
			Play: supremo 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 16 more moves within equity 0.32
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Messages In This Thread

• GNU Testing
  Stick -- Friday, 12 December 2008, at 8:20 a.m.
  • GNU Testing
    David Rockwell -- Friday, 12 December 2008, at 3:32 p.m.
    • GNU Testing
      Stick -- Friday, 12 December 2008, at 4:07 p.m.