Trinket V1.0.0 Up For Testing

[DkeRee]

I'm very happy that my account is finally activated so that I can submit a request for my engine, Trinket, to be tested!

My engine is open source on github: https://github.com/DkeRee/Trinket

A release is also published on the side for easy testing without the hassle to compile

It runs on UCI protocol and is written in Rust.

-DkeRee

[Yinuo Huang]

Hi,

(Just FYI I’m not a CCRL operator, just another humble engine author.)

I don’t see your engine supporting “stop” or “winc” “binc” “movestogo” options for the “go” command, as they may be crucial in testing (for example, the CCRL 40/15 tc requires movestogo).
You might want to consider completing the UCI interface. Just wanted to point it out there, hope it helps

Good luck to your engine being tested!

Thanks.

P.S. you might also want to add a README including description, compilation instructions, features, estimated elo, etc. so non-rust devs know how to compile & run it.

[DkeRee]

Hi,

(Just FYI I’m not a CCRL operator, just another humble engine author.)

I don’t see your engine supporting “stop” or “winc” “binc” “movestogo” options for the “go” command, as they may be crucial in testing (for example, the CCRL 40/15 tc requires movestogo).
You might want to consider completing the UCI interface. Just wanted to point it out there, hope it helps

Good luck to your engine being tested!

Thanks.

P.S. you might also want to add a README including description, compilation instructions, features, estimated elo, etc. so non-rust devs know how to compile & run it.

Hello! Thanks for the heads up. I will do just that right now.

[DkeRee]

Hi,

(Just FYI I’m not a CCRL operator, just another humble engine author.)

I don’t see your engine supporting “stop” or “winc” “binc” “movestogo” options for the “go” command, as they may be crucial in testing (for example, the CCRL 40/15 tc requires movestogo).
You might want to consider completing the UCI interface. Just wanted to point it out there, hope it helps

Good luck to your engine being tested!

Thanks.

P.S. you might also want to add a README including description, compilation instructions, features, estimated elo, etc. so non-rust devs know how to compile & run it.

Hello again,

I implemented the stop command in my engine (updated github src and release too), and after further research I have found that movestogo, winc, and binc just help fine tune the search time for a move before aborting. Wtime and Btime should do this just fine for my engine at the current moment (which my engine does have). Feel free to correct me if I'm wrong.

And while I'm at it:
I estimate my engine to be between 1900-2000 CCRL elo. I will add a readme in the near future.

Thanks, DkeRee

[bastiball]

I'll put this up on queue.

[Gabor Szots]

Two remarks:

1. I recommend putting the version number in the file name and engine id as well.
2. Master and 1.0.0 release both have id 1.0.0 although they are surely different. That may cause confusion.

Right

[Gabor Szots]

I may have missed it but I don't find your real name and nationality anywhere. Do you prefer to keep them to yourself?

[DkeRee]

I'll put this up on queue.

Thanks! I assume that when it's ready to be tested in the next wave you'll send over a tlcv link, correct?

-Dke

[DkeRee]

I may have missed it but I don't find your real name and nationality anywhere. Do you prefer to keep them to yourself?

I just realized that it was mismatched. I simply added a few more bug fixes since I first published my release, and so I simply updated my release, not knowing that the ID would be different. Will do!

Also, my real name is Daniel Ke and I'm from the US. Should I put this up on Github and phpBB? I will do that right now.

Thanks for checking out my engine!

-dke

[bastiball]

I'll put this up on queue.

Thanks! I assume that when it's ready to be tested in the next wave you'll send over a tlcv link, correct?

-Dke

Hello, I think Mr. Graham is the one that uses tlcv.

[DkeRee]

I'll put this up on queue.

Thanks! I assume that when it's ready to be tested in the next wave you'll send over a tlcv link, correct?

-Dke

Hello, I think Mr. Graham is the one that uses tlcv.

Oh, I see. Well I'll sit tight and wait for my results then.

[Gabor Szots]

I have started a tournament and I can see that almost all Trinket games are lost owing to illegal moves. This is what I found in the Arena debug:

2022-04-28 10:49:33,116<--1:info depth 7 nodes 5945633 pv e8h8 a1d1 d7c5 g2g1 f6e4 d4d5 e4c3 score cp 15 nps 1780061
2022-04-28 10:49:33,136<--1:bestmove e8h8
2022-04-28 10:49:33,136*moveerror* Rstart piece>0 and target piece>0, can't capture own piece !...!
2022-04-28 10:49:33,136*1*---------> Arena:Illegal move!: "bestmove e8h8" ()
2022-04-28 10:49:33,138****Tour**Game end: 1-0 {1-0 Arena Adjudication. Illegal move!}

I think Arena expects e8g8 or 0-0 for castling.

[DkeRee]

I have started a tournament and I can see that almost all Trinket games are lost owing to illegal moves. This is what I found in the Arena debug:

2022-04-28 10:49:33,116<--1:info depth 7 nodes 5945633 pv e8h8 a1d1 d7c5 g2g1 f6e4 d4d5 e4c3 score cp 15 nps 1780061
2022-04-28 10:49:33,136<--1:bestmove e8h8
2022-04-28 10:49:33,136*moveerror* Rstart piece>0 and target piece>0, can't capture own piece !...!
2022-04-28 10:49:33,136*1*---------> Arena:Illegal move!: "bestmove e8h8" ()
2022-04-28 10:49:33,138****Tour**Game end: 1-0 {1-0 Arena Adjudication. Illegal move!}

I think Arena expects e8g8 or 0-0 for castling.

Thank you for the notice. I've been using the python program lichess-bot on lichess API and cutechess for my testing or automated plays, and it worked perfectly. It appears to be that lichess-bot and cutechess automatically converts chess960 styled castling into the standard O-O or O-O-O. The reason for this is because I use cozy-chess API for my board represesntation, which uses chess960 styled castling. I will immediately fix this when I have time tommorow. Will this affect anything?

-dke

[Gabor Szots]

Thank you for the notice. I've been using the python program lichess-bot on lichess API and cutechess for my testing or automated plays, and it worked perfectly. It appears to be that lichess-bot and cutechess automatically converts chess960 styled castling into the standard O-O or O-O-O. The reason for this is because I use cozy-chess API for my board represesntation, which uses chess960 styled castling. I will immediately fix this when I have time tommorow. Will this affect anything?

I don't think so. By the time your fix is ready I'm going to have a free core for restarting the tournament. The already played games won't count.

BTW, could you make a minor change so that after the uci command the displayed id includes the verson number as well? In fact the exe in the assets might also include the version number in its name, to remain on the safe side. Of course only if this won't be the last released version, which I hope.

[DkeRee]

Thank you for the notice. I've been using the python program lichess-bot on lichess API and cutechess for my testing or automated plays, and it worked perfectly. It appears to be that lichess-bot and cutechess automatically converts chess960 styled castling into the standard O-O or O-O-O. The reason for this is because I use cozy-chess API for my board represesntation, which uses chess960 styled castling. I will immediately fix this when I have time tommorow. Will this affect anything?

I don't think so. By the time your fix is ready I'm going to have a free core for restarting the tournament. The already played games won't count.

BTW, could you make a minor change so that after the uci command the displayed id includes the verson number as well? In fact the exe in the assets might also include the version number in its name, to remain on the safe side. Of course only if this won't be the last released version, which I hope.

Alright! Glad to hear that. I just finished fixing the previous issue (e1a1 -> e1c1) and implemented your suggestions. I added a version identification in the executable file name and uci command (so that it would say "id name Trinket 1.0.0" after the "uci" command is given). I published it onto my github repository and replaced my V1.0.0 release with my updated executable. It is still version 1.0.0 since I feel that the changes aren't grand enough to qualify for a new release. Of course, in the future I will not be changing my releases like so anymore. I simply wanted a solid initial release before I got it all going

Thanks for testing out my engine and feel free to message me if you encounter any further issues! (hope not)

-dke

[Gabor Szots]

I just finished fixing the previous issue (e1a1 -> e1c1) and implemented your suggestions. I added a version identification in the executable file name and uci command (so that it would say "id name Trinket 1.0.0" after the "uci" command is given). I published it onto my github repository and replaced my V1.0.0 release with my updated executable. It is still version 1.0.0 since I feel that the changes aren't grand enough to qualify for a new release. Of course, in the future I will not be changing my releases like so anymore. I simply wanted a solid initial release before I got it all going

Excellent! Tournament already running.

BTW, I tried the 'go depth n' command. When n was 10 the reached depth was only 7 then it displayed bestmove. When n was 9 the reached depth was 8. I don't think that is correct behaviour, what is your opinion?

Gabor

[DkeRee]

I just finished fixing the previous issue (e1a1 -> e1c1) and implemented your suggestions. I added a version identification in the executable file name and uci command (so that it would say "id name Trinket 1.0.0" after the "uci" command is given). I published it onto my github repository and replaced my V1.0.0 release with my updated executable. It is still version 1.0.0 since I feel that the changes aren't grand enough to qualify for a new release. Of course, in the future I will not be changing my releases like so anymore. I simply wanted a solid initial release before I got it all going

Excellent! Tournament already running.

BTW, I tried the 'go depth n' command. When n was 10 the reached depth was only 7 then it displayed bestmove. When n was 9 the reached depth was 8. I don't think that is correct behaviour, what is your opinion?

Gabor

Thanks again! Glad to hear that it is running smoothly. As for the go depth n command, I suppose I misread the protocol then. I just made it so that when no wtime or btime is specified it will default it to 5 minutes in milliseconds, and so that restriction would be used in the go depth n command. I made it so that go depth n would simply set a limit to how far it can search the depth to. Thus, the search would be aborted when it either hits the depth limit or time limit (whichever comes first, depth restriction will be ignored if not specified). I could fix it in the near future.

-dke

[Gabor Szots]

Seems an engine playing in a very enterprising style. However, there are too many time forfeits, so far 7 out of 41 games.

BTW, does it gives the score from white's point of view?

[Gabor Szots]

Having played some games I had to discover that the engine is much weaker than the announced 1900 (I don't remember where I got that number from).
That is not a problem in itself but I have selected wrong opponents already twice which was a waste of time.

Also, the frequent time forfeits do pose a problem for the tester. I have to edit the whole resulting PGN in order to prevent the processing scripts to remove games which Trinket would have lost anyway. That is a tedious work for which I don't have the desire.
I am surprised that with an incremental time control that can happen at all. For this reason I have decided to stop my tests until the issue is solved.

[DkeRee]

Having played some games I had to discover that the engine is much weaker than the announced 1900 (I don't remember where I got that number from).
That is not a problem in itself but I have selected wrong opponents already twice which was a waste of time.

Also, the frequent time forfeits do pose a problem for the tester. I have to edit the whole resulting PGN in order to prevent the processing scripts to remove games which Trinket would have lost anyway. That is a tedious work for which I don't have the desire.
I am surprised that with an incremental time control that can happen at all. For this reason I have decided to stop my tests until the issue is solved.

I estimated that it was about 1900 elo, and I was terribly wrong. I will see what I can do about the time management system and I will get back to you in the near future (I didn't think that it would be THIS bad). However, I didn't experience too many issues with my time management on my end when I was testing stuff out. Still, thanks for everything!

P.S. It gives the score based off of the side it's playing.

P.S.S. I will still keep the version as 1.0.0 until I get a solid release that can participate in CCRL testing.

-dke

[DkeRee]

Having played some games I had to discover that the engine is much weaker than the announced 1900 (I don't remember where I got that number from).
That is not a problem in itself but I have selected wrong opponents already twice which was a waste of time.

Also, the frequent time forfeits do pose a problem for the tester. I have to edit the whole resulting PGN in order to prevent the processing scripts to remove games which Trinket would have lost anyway. That is a tedious work for which I don't have the desire.
I am surprised that with an incremental time control that can happen at all. For this reason I have decided to stop my tests until the issue is solved.

Now that I think about it, could it be a huge bug gone unnoticed? If you can, can you check if my engine timed out because it genuinely had no time left (due to bad time management) or did it freeze with a few minutes left on the clock? If it's because of the former I believe that I can fix it by adding winc, binc, and movestogo commands. Previously, I assumed that these simply fine tuned the time management a bit more. Thus, I didn't decide to add them quite yet. I ask this since I haven't had any troubles with time management beforehand.

-dke

[Gabor Szots]

Now that I think about it, could it be a huge bug gone unnoticed? If you can, can you check if my engine timed out because it genuinely had no time left (due to bad time management) or did it freeze with a few minutes left on the clock? If it's because of the former I believe that I can fix it by adding winc, binc, and movestogo commands.

I'm afraid it is the latter. Look at these two games:

[Event "Trinket 1.0.0 64-bit - Apr 30"]
[Site "Szentendre"]
[Date "2022.04.30"]
[Round "1"]
[White "Trinket 1.0.0 64-bit"]
[Black "BibiChess 0.5"]
[Result "0-1"]
[ECO "C06"]
[Opening "French"]
[Time "07:53:40"]
[Variation "Tarrasch, Closed, 8...f6 9.exf6, 11...O-O 12.Bf4"]
[TimeControl "120+1"]
[PlyCount "70"]

1. e4 e6 2. d4 d5 3. Nd2 Nf6 4. e5 Nfd7 5. c3 c5 6. Bd3 Nc6 7. Ne2 cxd4 8.
cxd4 f6 9. exf6 Nxf6 10. O-O {-0.06/6 4} Bd6 {+0.45/7 4} 11. Nf3 {+0.04/7
3} O-O {+0.37/6 4} 12. Bf4 {-0.05/6 4} Bxf4 {+0.36/7 4} 13. Nxf4 {-0.28/7
2} Ne4 {+0.36/6 3} 14. g3 {-0.11/6 4} g5 {+0.44/7 3} 15. Nh5 {-0.30/7 3}
Bd7 {+0.50/6 3} 16. h4 {-0.03/6 4} Be8 {+1.34/7 3} 17. Ne5 {-0.27/7 2} Nxd4
{+1.20/7 3} 18. hxg5 {-0.87/6 3} Qxg5 {+2.40/7 3} 19. f4 {-2.14/6 4} Qxh5
{+2.40/7 3} 20. Bxe4 {-3.66/7 4} Ne2+ {+2.43/7 3} 21. Kg2 {-3.66/8 2} dxe4
{+2.43/6 3} 22. Rh1 {-3.21/7 2} Nxf4+ {+2.54/8 3} 23. gxf4 {-3.02/8 3} Qxd1
{+2.54/8 3} 24. Rhxd1 {-3.06/7 3} Rxf4 {+2.83/8 3} 25. Rd4 {-3.41/7 3} Bh5
{+2.83/6 3} 26. Kg3 {-4.83/7 3} Raf8 {+2.74/7 3} 27. Re1 {-4.79/7 3} Bf3
{+2.74/7 3} 28. Rd6 {-4.79/7 4} R4f5 {+3.31/8 3} 29. Rxe6 {-4.32/8 2} Rg5+
{+3.10/8 2} 30. Kh3 {-4.25/7 2} Rh5+ {+2.96/7 2} 31. Kg3 {-3.57/8 3} Rg5+
{+2.95/8 2} 32. Kh3 {0.00/9 3} Bg2+ {+2.95/7 2} 33. Kh4 {-3.98/8 3} Rg7
{+2.88/8 2} 34. Rc1 {-3.78/7 3} Rf4+ {+3.01/7 2} 35. Kh5 {-3.89/8 2} Bf3+
{+2.95/7 2} 36. ... {White forfeits on time} 0-1

[Event "Trinket 1.0.0 64-bit - Apr 30"]
[Site "Szentendre"]
[Date "2022.04.30"]
[Round "2"]
[White "IACH 0.56z"]
[Black "Trinket 1.0.0 64-bit"]
[Result "1-0"]
[ECO "A87"]
[Opening "Dutch"]
[Time "09:38:13"]
[Variation "Leningrad, Main Line, 7.Nc3 Qe8"]
[TimeControl "120+1"]
[PlyCount "91"]

1. d4 f5 2. Nf3 Nf6 3. g3 g6 4. Bg2 Bg7 5. c4 O-O 6. O-O d6 7. Nc3 Qe8 8.
Bf4 {+0.28/11 2} Ne4 {-0.71/6 5} 9. Nxe4 {+0.44/11 2} fxe4 {-0.61/7 5} 10.
Ng5 {+0.42/11 2} e5 {-0.88/6 4} 11. dxe5 {+1.12/12 2} dxe5 {-1.67/7 3} 12.
Be3 {+1.14/11 2} Qd8 {-1.69/6 4} 13. Qb3 {+1.32/11 2} c6 {-1.93/6 4} 14.
Nxe4 {+1.40/11 2} Qc7 {-2.05/6 4} 15. c5+ {+1.44/10 2} Qf7 {-1.95/7 4} 16.
Qxf7+ {+1.44/11 2} Rxf7 {-2.03/7 3} 17. Nd6 {+1.34/12 2} Rc7 {-2.11/7 4}
18. Nxc8 {+1.36/11 2} Rxc8 {-1.90/8 2} 19. Bh3 {+1.32/11 2} Re8 {-1.70/8 4}
20. Rad1 {+1.28/10 2} e4 {-1.73/8 2} 21. b4 {+1.14/11 1} a5 {-1.23/7 3} 22.
bxa5 {+1.16/12 1} Rxa5 {-1.08/7 3} 23. Rd2 {+1.12/11 1} Na6 {-0.52/7 4} 24.
Rc1 {+1.02/11 1} Nb4 {-0.46/6 2} 25. Rc4 {+1.08/11 1} Nd5 {+0.20/7 3} 26.
Bg2 {-0.02/11 1} Nxe3 {+0.30/6 2} 27. fxe3 {+0.02/15 1} Bh6 {+0.89/7 3} 28.
Rxe4 {0.00/15 1} Bxe3+ {+1.24/8 2} 29. Kh1 {0.00/16 1} Rxe4 {+0.83/8 2} 30.
Rd8+ {0.00/15 1} Kf7 {+0.82/9 1} 31. Bxe4 {0.00/15 1} Ra4 {+0.57/8 1} 32.
Bd3 {+1.00/13 1} Bxc5 {-0.34/7 2} 33. Rd7+ {+1.04/13 1} Kf6 {-0.86/8 2} 34.
Rxb7 {+1.10/13 1} Rxa2 {-1.05/8 1} 35. Rxh7 {+1.10/12 1} Ra7 {-1.37/7 1}
36. Rh8 {+1.10/13 1} Bd4 {-1.32/6 2} 37. Rc8 {+2.04/14 1} Ra1+ {-2.26/9 2}
38. Kg2 {+2.06/15 1} Rc1 {-2.56/8 2} 39. Rg8 {+2.06/14 1} Rg1+ {-2.83/8 2}
40. Kf3 {+2.12/13 1} g5 {-3.08/7 2} 41. e3 {+2.18/13 1} Bb6 {-3.59/8 2} 42.
Rg6+ {+2.20/13 1} Ke7 {-4.15/9 1} 43. Rxc6 {+2.26/13 1} Bd8 {-4.30/8 2} 44.
e4 {+2.28/13 1} Rd1 {-4.29/8 1} 45. Bc4 {+2.28/13 1} Rc1 {-4.34/8 1} 46. e5
{+2.36/14 1} ... {Black forfeits on time} 1-0

If you add the times spent for each move, you get 87 for the first game, 98 for the second one. That means Trinket had plenty of time left on its clock when it exceeded time.

[DkeRee]

Now that I think about it, could it be a huge bug gone unnoticed? If you can, can you check if my engine timed out because it genuinely had no time left (due to bad time management) or did it freeze with a few minutes left on the clock? If it's because of the former I believe that I can fix it by adding winc, binc, and movestogo commands.

I'm afraid it is the latter. Look at these two games:

[Event "Trinket 1.0.0 64-bit - Apr 30"]
[Site "Szentendre"]
[Date "2022.04.30"]
[Round "1"]
[White "Trinket 1.0.0 64-bit"]
[Black "BibiChess 0.5"]
[Result "0-1"]
[ECO "C06"]
[Opening "French"]
[Time "07:53:40"]
[Variation "Tarrasch, Closed, 8...f6 9.exf6, 11...O-O 12.Bf4"]
[TimeControl "120+1"]
[PlyCount "70"]

1. e4 e6 2. d4 d5 3. Nd2 Nf6 4. e5 Nfd7 5. c3 c5 6. Bd3 Nc6 7. Ne2 cxd4 8.
cxd4 f6 9. exf6 Nxf6 10. O-O {-0.06/6 4} Bd6 {+0.45/7 4} 11. Nf3 {+0.04/7
3} O-O {+0.37/6 4} 12. Bf4 {-0.05/6 4} Bxf4 {+0.36/7 4} 13. Nxf4 {-0.28/7
2} Ne4 {+0.36/6 3} 14. g3 {-0.11/6 4} g5 {+0.44/7 3} 15. Nh5 {-0.30/7 3}
Bd7 {+0.50/6 3} 16. h4 {-0.03/6 4} Be8 {+1.34/7 3} 17. Ne5 {-0.27/7 2} Nxd4
{+1.20/7 3} 18. hxg5 {-0.87/6 3} Qxg5 {+2.40/7 3} 19. f4 {-2.14/6 4} Qxh5
{+2.40/7 3} 20. Bxe4 {-3.66/7 4} Ne2+ {+2.43/7 3} 21. Kg2 {-3.66/8 2} dxe4
{+2.43/6 3} 22. Rh1 {-3.21/7 2} Nxf4+ {+2.54/8 3} 23. gxf4 {-3.02/8 3} Qxd1
{+2.54/8 3} 24. Rhxd1 {-3.06/7 3} Rxf4 {+2.83/8 3} 25. Rd4 {-3.41/7 3} Bh5
{+2.83/6 3} 26. Kg3 {-4.83/7 3} Raf8 {+2.74/7 3} 27. Re1 {-4.79/7 3} Bf3
{+2.74/7 3} 28. Rd6 {-4.79/7 4} R4f5 {+3.31/8 3} 29. Rxe6 {-4.32/8 2} Rg5+
{+3.10/8 2} 30. Kh3 {-4.25/7 2} Rh5+ {+2.96/7 2} 31. Kg3 {-3.57/8 3} Rg5+
{+2.95/8 2} 32. Kh3 {0.00/9 3} Bg2+ {+2.95/7 2} 33. Kh4 {-3.98/8 3} Rg7
{+2.88/8 2} 34. Rc1 {-3.78/7 3} Rf4+ {+3.01/7 2} 35. Kh5 {-3.89/8 2} Bf3+
{+2.95/7 2} 36. ... {White forfeits on time} 0-1

[Event "Trinket 1.0.0 64-bit - Apr 30"]
[Site "Szentendre"]
[Date "2022.04.30"]
[Round "2"]
[White "IACH 0.56z"]
[Black "Trinket 1.0.0 64-bit"]
[Result "1-0"]
[ECO "A87"]
[Opening "Dutch"]
[Time "09:38:13"]
[Variation "Leningrad, Main Line, 7.Nc3 Qe8"]
[TimeControl "120+1"]
[PlyCount "91"]

1. d4 f5 2. Nf3 Nf6 3. g3 g6 4. Bg2 Bg7 5. c4 O-O 6. O-O d6 7. Nc3 Qe8 8.
Bf4 {+0.28/11 2} Ne4 {-0.71/6 5} 9. Nxe4 {+0.44/11 2} fxe4 {-0.61/7 5} 10.
Ng5 {+0.42/11 2} e5 {-0.88/6 4} 11. dxe5 {+1.12/12 2} dxe5 {-1.67/7 3} 12.
Be3 {+1.14/11 2} Qd8 {-1.69/6 4} 13. Qb3 {+1.32/11 2} c6 {-1.93/6 4} 14.
Nxe4 {+1.40/11 2} Qc7 {-2.05/6 4} 15. c5+ {+1.44/10 2} Qf7 {-1.95/7 4} 16.
Qxf7+ {+1.44/11 2} Rxf7 {-2.03/7 3} 17. Nd6 {+1.34/12 2} Rc7 {-2.11/7 4}
18. Nxc8 {+1.36/11 2} Rxc8 {-1.90/8 2} 19. Bh3 {+1.32/11 2} Re8 {-1.70/8 4}
20. Rad1 {+1.28/10 2} e4 {-1.73/8 2} 21. b4 {+1.14/11 1} a5 {-1.23/7 3} 22.
bxa5 {+1.16/12 1} Rxa5 {-1.08/7 3} 23. Rd2 {+1.12/11 1} Na6 {-0.52/7 4} 24.
Rc1 {+1.02/11 1} Nb4 {-0.46/6 2} 25. Rc4 {+1.08/11 1} Nd5 {+0.20/7 3} 26.
Bg2 {-0.02/11 1} Nxe3 {+0.30/6 2} 27. fxe3 {+0.02/15 1} Bh6 {+0.89/7 3} 28.
Rxe4 {0.00/15 1} Bxe3+ {+1.24/8 2} 29. Kh1 {0.00/16 1} Rxe4 {+0.83/8 2} 30.
Rd8+ {0.00/15 1} Kf7 {+0.82/9 1} 31. Bxe4 {0.00/15 1} Ra4 {+0.57/8 1} 32.
Bd3 {+1.00/13 1} Bxc5 {-0.34/7 2} 33. Rd7+ {+1.04/13 1} Kf6 {-0.86/8 2} 34.
Rxb7 {+1.10/13 1} Rxa2 {-1.05/8 1} 35. Rxh7 {+1.10/12 1} Ra7 {-1.37/7 1}
36. Rh8 {+1.10/13 1} Bd4 {-1.32/6 2} 37. Rc8 {+2.04/14 1} Ra1+ {-2.26/9 2}
38. Kg2 {+2.06/15 1} Rc1 {-2.56/8 2} 39. Rg8 {+2.06/14 1} Rg1+ {-2.83/8 2}
40. Kf3 {+2.12/13 1} g5 {-3.08/7 2} 41. e3 {+2.18/13 1} Bb6 {-3.59/8 2} 42.
Rg6+ {+2.20/13 1} Ke7 {-4.15/9 1} 43. Rxc6 {+2.26/13 1} Bd8 {-4.30/8 2} 44.
e4 {+2.28/13 1} Rd1 {-4.29/8 1} 45. Bc4 {+2.28/13 1} Rc1 {-4.34/8 1} 46. e5
{+2.36/14 1} ... {Black forfeits on time} 1-0

If you add the times spent for each move, you get 87 for the first game, 98 for the second one. That means Trinket had plenty of time left on its clock when it exceeded time.

Ah, I love coding! Fixing one bug creates another! Thank you for the confirmation, for that confirms my suspicions. I do know for sure that my program didn't crash before since I ran tests in bulk without much issue. Thus, I must've changed something recently that broke some part of my program. When I fix this I will get back to you. Thanks!

(At least I know that my time management is fine)

-dke

[Gabor Szots]

Just occurred to me that it can be related to my playing a game of logic while the tournament is running. That game has a clock running while I play. Maybe it interferes with the chess clock.
Just a wild guess because I do it often and other engines are not sensitive to this. But some are.

EDIT: I have played about 30 more games without playing myself simultaneously. 1 time forfeit occurred, in lost position. Unfortunately the small sample size does not confirm anything but I can't do more.

[DkeRee]

Just occurred to me that it can be related to my playing a game of logic while the tournament is running. That game has a clock running while I play. Maybe it interferes with the chess clock.
Just a wild guess because I do it often and other engines are not sensitive to this. But some are.

EDIT: I have played about 30 more games without playing myself simultaneously. 1 time forfeit occurred, in lost position. Unfortunately the small sample size does not confirm anything but I can't do more.

I'm pretty sure it's an issue on my end. Sorry for putting you through so much trouble ‍

I will notify you when I fix it!

-dke