TalkChess.com

Guenther wrote: ↑Thu Dec 19, 2019 2:22 pm A suggestion for a future improvement:

Logarithmic scaling would make the graph look better, if one does late adjudication.
Otherwise most part of the graph will look flat, even when the scores were already quite high.

phhnguyen wrote: ↑Fri Jan 31, 2020 10:18 am Version 0.15 released

* Fixed bugs: almost all known bugs so far.  The most important ones are (hopefully) the root bugs of crashing/exiting problems of BSG. Look like BSG now can run tournaments much more smoothly. Other bugs: couldn’t load some online PGN files, show some wrong stats for tournaments
* Some small improvements/adjustments. Graphical lagging issues have been improved
* New features:
- Pickup openings in sequence (for tournaments when using PGN or EPD books)
- Draw adjudication by engines scores

Notes:
- Lichess bots function has been temporally turned off due to it has some unsolve problems dealing with missing/wrong received data
- BSG has been passed all my tests for durable/crashing with some sets of engines and computers. However, it is clearly too small and not diversity enough. Please help me to test more on your hardware and your sets of engines
- A neutral flag “Earth” has been added and now flags can be shown for both engines and human

Ferdy wrote: ↑Sat Feb 08, 2020 8:57 am

phhnguyen wrote: ↑Fri Jan 31, 2020 10:18 am Version 0.15 released

* Fixed bugs: almost all known bugs so far.  The most important ones are (hopefully) the root bugs of crashing/exiting problems of BSG. Look like BSG now can run tournaments much more smoothly. Other bugs: couldn’t load some online PGN files, show some wrong stats for tournaments
* Some small improvements/adjustments. Graphical lagging issues have been improved
* New features:
- Pickup openings in sequence (for tournaments when using PGN or EPD books)
- Draw adjudication by engines scores

Notes:
- Lichess bots function has been temporally turned off due to it has some unsolve problems dealing with missing/wrong received data
- BSG has been passed all my tests for durable/crashing with some sets of engines and computers. However, it is clearly too small and not diversity enough. Please help me to test more on your hardware and your sets of engines
- A neutral flag “Earth” has been added and now flags can be shown for both engines and human

While the tournament is in progress and you re-arrange some windows, BSG will crash.
Test conditions:
* System is i7 -2600K 3.4 Ghz, 4-cores, 8-threads, 12GB RAM, Windows 10
* Tournament concurrency 4
* Each engine is using 1 thread and 128MB Hash

phhnguyen wrote: ↑Sat Feb 08, 2020 6:53 am

Guenther wrote: ↑Thu Dec 19, 2019 2:22 pm A suggestion for a future improvement:

Logarithmic scaling would make the graph look better, if one does late adjudication.
Otherwise most part of the graph will look flat, even when the scores were already quite high.

I started working on this suggestion. Just have seen some issues:

- Logarithmic scaling is good for exponential values. However, chess scores seem not to be exponential. Sometimes scores may be very high but most of the time scores are small values. Logarithmic scaling may make the graph looks much flatter
- score unit may affect seriously the shape of the graph since logarithms of values below 1 (if using pawn unit) are negative when the above values are positive. Centipawn gives always positive. Which one is better?
- I don't know how to deal with scores of 0 and negative (they can't be calculated logarithms)

Any ideas/experiences? Thanks

phhnguyen wrote: ↑Fri Jan 31, 2020 10:18 am Version 0.15 released

* Fixed bugs: almost all known bugs so far.  The most important ones are (hopefully) the root bugs of crashing/exiting problems of BSG. Look like BSG now can run tournaments much more smoothly. Other bugs: couldn’t load some online PGN files, show some wrong stats for tournaments
* Some small improvements/adjustments. Graphical lagging issues have been improved
* New features:
- Pickup openings in sequence (for tournaments when using PGN or EPD books)
- Draw adjudication by engines scores

Notes:
- Lichess bots function has been temporally turned off due to it has some unsolve problems dealing with missing/wrong received data
- BSG has been passed all my tests for durable/crashing with some sets of engines and computers. However, it is clearly too small and not diversity enough. Please help me to test more on your hardware and your sets of engines
- A neutral flag “Earth” has been added and now flags can be shown for both engines and human

BrendanJNorman wrote: ↑Sat Feb 08, 2020 10:32 am

Ferdy wrote: ↑Sat Feb 08, 2020 8:57 am
While the tournament is in progress and you re-arrange some windows, BSG will crash.
Test conditions:
* System is i7 -2600K 3.4 Ghz, 4-cores, 8-threads, 12GB RAM, Windows 10
* Tournament concurrency 4
* Each engine is using 1 thread and 128MB Hash

Yeah, the crashing is really killing it for me. Moved back to Arena/Fritz GUI while bugs are fixed.

I'm a patient guy and Nguyen is working like a machine, so Banksia will be running again on my machine soon enough.

Guenther wrote: ↑Sat Feb 08, 2020 10:33 am You could look into the WB/XB source it does some clever scalings for the graph (with user settings available, eg. evalZoom and evalThreshold).

Edit: I guess a lot of programs also use a ceiling value, which treats all values above it as equal.

phhnguyen wrote: ↑Sat Feb 08, 2020 1:37 pm To BrendanJNorman: I guess on your computer BSG crash not only by re-arranging windows but for other situations (thus you could not use it), is it correct? If yes, please tell me more details. Thanks again.

phhnguyen wrote: ↑Sat Feb 08, 2020 1:37 pm Clearly BSG of you both got crashes. However, I can't reproduce even I have tried on a few computers. On my very weak laptop, it becomes so laggy when I re-arrange windows (showing multi-playing games) but did not crash.
- Did BSG crash when it is showing tournament's playing games (select Open & view) or not?

- Did BSG crash randomly or after doing a certain action?