|Subject:||[Bug-gnubg] Problems with Gnubg 04-Mar-2009 release|
|Date:||Fri, 13 Mar 2009 11:04:12 +0000 (GMT)|
Dear Christian, Øystein and all other Gnu developers,
First of all, I would like to congratulate all of you for developing such a great product! I am a backgammon aficionado from Istanbul, Turkey. I have been using your software only since September, 2008 and am greatly impressed with its capabilities so far (especially the Tutor and Hint modes which don't exist in Snowie, but are so important for learning-by-doing).
I never downloaded any upgrades since the 05-Sep-08 release, as I was already very happy with the performance of the code, barring some minor glitches. Couple of days ago, I upgraded my code to the 04-Mar-09 release. However, I noticed some problems in the newer version. I wanted to share these with you as a feedback for positive contribution by subscribing to this forum (hoping this is the right place).
In order to systematically determine the differences (being an engineer myself :), I opened 2 instances of gnubg side-by-side (05-11-08 release vs. 04-03-09 release) in Windows XP SP3. My settings for both versions are for gnugb to be "supremo" checker play and "world class" cube (analysis and evaluation settings are just the same, too). I also set the dices to the same seed in order to play the exact same matches in both versions with the exact same moves of mine.
Here is a list of differences / problems that I noticed so far (some of them very minor - maybe even intentional?) I hope I am hereby not repeating any previous posts. The bullets are not listed in any particular order of importance:
1) The best move in the "Analysis panel" and "Hint window" is not highlighted with red any more. Being accustomed to see most of my moves in the first line in red, this constantly makes me think that I made a mistake and my move is listed way far down in the list.
2) In addition, the line spacings in the "Analysis panel" and "Hint window" are now much narrower which leads to a more cramped / crowded list of moves.
3) The "Analysis panel" is not showing under rolled dice the "delta" equity of a particular move made with respect to the average roll equity [Previously, it was saying above the toggle menu for luck for instance: Rolled 25 (-0.076) - this value in parenthesis is not there anymore]..
4) When you perform any rollout for any move(s) in the Hint or Analysis window and the rollout finishes, you cannot go back to the main program by clicking OK. The program just freezes and crashes.
5) For unknown reasons, there were couple of instances when the newer version arbitrarily crashed during games/matches. The only common characteristic that I noticed was that they were all happening when I was clicking on the dice to make the next move (either for gnu or for myself).
6) The size of the new Temperature Map window and the fonts are bigger now which is obviously a huge plus! However, the best moves inside the squares are listed upto 4 lines now (intentional?). For example, a double six played 13/7(4) is not represented 13/7(4) anymore, but on four separate lines as 13/7. Also, when you resize the temperature map bigger, the fonts of best moves and equity over-adjust such that the characters overlap.
7) When an offered cube is taken by the other player, the previous version played a sound file ("take.wav"). This file is not played in the newer version. After one accepts the cube, you immediately hear the sound of next dice rolling. Very minor issue but just wanted to point it out.
8) Finally, when I analyze the same match that I played in the 2 instances of gnubg, the older version takes 5 seconds, whereas the newer version does it in 45 seconds! (even starting the analysis first with the older version!) This applies to any previous saved games of mine, too. The newer version with exactly the same settings takes much more time to analyze the games/matches? I have no idea why.
Last but not least, there are 4 general problems in both versions that one would hope to be fixed in the future versions:
A) The tutor does NOT warn you (however big of a blunder your cube decision is) when you are "Too Good To Double" and gnubg penalizes you by taking your wrongly offered cube and ending the game, whereas you should have continued to play for gammon. Everytime this happens I think that my cube was correct, just to be disappointed later when I run an analysis of the match to see that it was either doubtful, or even bad.
B) When your actual move is marked doubtful (maybe even bad) in the Hint or Analysis window and you perform a rollout just to be sure, sometimes you find out that it is indeed the best move according to the rollout results. But gnubg does not somehow "save" or "feed" this information into its game evaluation. Then, when you go ahead and play the confirmed best move in the game, it is still marked as doubtful, bad etc. which affects your final statistics :( This was not the case when I was trial testing Snowie 4 where seemingly incorrect moves were accepted later after a proper roll-out.
C) In the bear-in or bear-off phase, Gnubg does some strange checker plays when it is either definitely winning OR definitely losing the game. For example, with two checkers left on 1 and 2 points and having a last winning roll of a 6-1, it moves the checker on the 2 point to the 1 point and bears off the same checker, leaving one checker behind for an additional unnecessary roll? I sometimes wonder whether it is psychologically messing and teasing me by prolonging the pain :)) Likewise, when I am certain to win a gammon, it does not continue the fight by trying to bear-in as much checkers as possible to save the gammon and sometimes makes irrelevant moves in its own homeboard?
D) The statistics numbers in the "Player Records" do not match those of "Match Statistics". One can verify this by erasing all player records, playing a fresh, brand new match, running a match analysis and adding the results to the "Player Records". The numbers in the two windows are completely different (unless "Player Records" calculates something else?)
I hope that this list was not too long for you to read and that it would serve beneficial to further improve an already fantastic program! I would appreciate it if you would comment on the issues mentioned above and look forward to your replies (please not too technical - I have no programming knowledge). If you need any further concrete evidence (screenshots, saved games/matches etc.), I would be happy to send you these separately, too. And thanks again to everyone in the team for such a superior work!
|[Prev in Thread]||Current Thread||[Next in Thread]|