[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Bug-gnubg] Re: remote server code fixed I think
|
From: |
kvandoel |
|
Subject: |
[Bug-gnubg] Re: remote server code fixed I think |
|
Date: |
Thu, 4 Sep 2003 21:10:37 +0200 (CEST) |
They way I test it is to set everything to expert level and do a match
analysis: all error rates should be 0. As it is with version 1.1343 if I
play like that, the external player makes "errors", meaning it's not
playing at expert level.
I've tried 3pt matches like that, both with player 0 and player 1
external. The only errors are now chequerplay errors. The errors are
larger when player 0 is the external.
The errors are very small (rating equiv. of 18 (player 0) and 40
(player 1)), yet they should be exactly 0.
Let me know if you can reproduce this.
Kees
On Thu, 4 Sep 2003, Jim Segrave wrote:
> On Wed 03 Sep 2003 (17:15 +0000), Joern Thyssen wrote:
> > On Wed, Sep 03, 2003 at 04:27:52PM +0000, Joern Thyssen wrote
> > > On Wed, Sep 03, 2003 at 12:13:48AM +0200, Jim Segrave wrote
> > > >
> > > > You can see this yourself by starting a gnubg -t and giving the
> > > > command
> > > > external localhost:4567
> > > > (or any other port number you like), then starting another gnubg and
> > > > setting one player to external with the same localhost:4567. Start a
> > > > money session and watch the strange cube handling.
> > >
> > > I can only reproduce the problem with
> > >
> > > set player 0 external localhost:10000
> > >
> > > Setting player 1 to external works fine. It seems there is a problem
> > > swapping the board.
> >
> > I've removed the call to SwapSides in external.c. It seesm to fix the
> > problem.
>
> I think so as well. I've just done a short test - 25 to 30 money games
> expert mode vs. expert mode and the results, particularly the luck
> adjusted results (what a useful feature) suggest to me that it's
> playing equal strength in master and client. I tried running the
> client with the external on player0 and player1 and again, it looks
> reasonable to me:
>
> clientgnubg remotegnubg1
> Chequer Play Statistics:
> Total moves 591 584
> Unforced moves 503 460
> Unmarked moves 0 0
> Moves marked good 583 579
> Moves marked doubtful 6 5
> Moves marked bad 2 0
> Moves marked very bad 0 0
> Error rate (total) -1.441 ( -2.785) -1.053 (-2.118)
> Error rate (per move) -2.87 ( -0.006) -2.29 (-0.005)
> Chequerplay rating World class World class
>
> Luck Statistics:
> Rolls marked very lucky 6 12
> Rolls marked lucky 27 32
> Rolls unmarked 526 504
> Rolls marked unlucky 30 29
> Rolls marked very unlucky 2 7
> Luck rate (total) -6.678 (-17.967) -1.670 (+0.886)
> Luck rate (per move) -11.30 ( -0.030) -2.86 (+0.002)
> Luck rating None None
>
> Cube Statistics:
> Total cube decisions 330 440
> Close or actual cube decisions 78 92
> Doubles 18 21
> Takes 13 18
> Passes 8 0
> Missed doubles around DP 0 0
> Missed doubles around TG 0 0
> Wrong doubles around DP 0 7 (+0.136(
> +0.140)
> Wrong doubles around TG 5 (+0.188 ( +0.252 0
> Wrong takes 0 2 (+0.172(
> +0.217)
> Wrong passes 2 (+0.038 ( +0.126 0
> Error rate (total) -0.226 ( -0.378) -0.308 (-0.357)
> Error rate (per cube decision) -2.90 ( -0.005) -3.35 (-0.004)
> Cube decision rating World class World class
>
> Overall Statistics:
> Error rate (total) -1.667 ( -3.163) -1.361 (-2.476)
> Error rate (per decision) -2.87 ( -0.005) -2.47 (-0.004)
> Equiv. Snowie error rate -1.42 -1.16
> Overall rating World class World class
> Actual result -18.000 +18.000
> Luck adjusted result +0.853 -0.853
> Advantage (actual) in ppg -0.7 0.7
> 95% confidence interval (ppg) 1.1 1.1
> Advantage (luck adjusted) in ppg 0.0 -0.0
> 95% confidence interval (ppg) 0.0 0.0
>
> =============== and reversed ==============================
>
> remotegnubg0 clientgnubg1
> Chequer Play Statistics:
> Total moves 586 591
> Unforced moves 531 523
> Unmarked moves 0 0
> Moves marked good 573 586
> Moves marked doubtful 10 4
> Moves marked bad 3 1
> Moves marked very bad 0 0
> Error rate (total) -1.782 ( -2.536) -1.204 (-1.914)
> Error rate (per move) -3.36 ( -0.005) -2.30 (-0.004)
> Chequerplay rating World class World class
>
> Luck Statistics:
> Rolls marked very lucky 7 7
> Rolls marked lucky 31 35
> Rolls unmarked 527 525
> Rolls marked unlucky 16 18
> Rolls marked very unlucky 5 6
> Luck rate (total) +0.902 ( -2.805) +4.485 (+5.765)
> Luck rate (per move) +1.54 ( -0.005) +7.59 (+0.010)
> Luck rating None None
>
> Cube Statistics:
> Total cube decisions 450 405
> Close or actual cube decisions 101 100
> Doubles 21 22
> Takes 14 13
> Passes 8 8
> Missed doubles around DP 0 2 (+0.067(
> +0.126)
> Missed doubles around TG 2 (+0.045 ( +0.085 0
> Wrong doubles around DP 0 6 (+0.123(
> +0.172)
> Wrong doubles around TG 3 (+0.075 ( +0.109 0
> Wrong takes 0 0
> Wrong passes 0 1 (+0.051(
> +0.051)
> Error rate (total) -0.120 ( -0.194) -0.242 (-0.350)
> Error rate (per cube decision) -1.19 ( -0.002) -2.42 (-0.004)
> Cube decision rating Supernatural World class
>
> Overall Statistics:
> Error rate (total) -1.902 ( -2.730) -1.445 (-2.264)
> Error rate (per decision) -3.01 ( -0.004) -2.32 (-0.004)
> Equiv. Snowie error rate -1.62 -1.23
> Overall rating World class World class
> Actual result -7.000 +7.000
> Luck adjusted result +1.569 -1.569
> Advantage (actual) in ppg -0.2 0.2
> 95% confidence interval (ppg) 0.3 0.3
> Advantage (luck adjusted) in ppg 0.0 -0.0
> 95% confidence interval (ppg) 0.0 0.0
>
>
>
>
>
> --
> Jim Segrave address@hidden
>