The things I used to play around with



Regards

Nick

Miniaturansicht angehängter Grafiken

 

Indeed, a single game analysis can provide some clues about a game-scope performance; player's stats across a (hopefully large) database is required for an Elo strength estimate of a player.

Nevertheless, I double-checked the P400 King 2.55 vs Chess Gate Smyslov game and i can report slightly different outcomes, as CARA is highly flexible and setting-dependant. Using 10secs SF18, 1 core as the analysis engine, I got:
And a clue about Smyslov's low-looking Elo estimate:
As you can read, Smyslov's game as Black shew some inconsistency from the opening phase to the middlegame; and this is a criteria that matters for the Elo estimate.

Anyway, let's wait for several enough games to run player's stats.
MfG,
Eric

Hi Eric

Thanks, just a couple of things with Cara.

Based on the speed of the search, since it is only about a second or less per move, it is going to miss a lot of secondary good moves which will always give the ratings inaccuracies as it can't recover due to lack of time to search the next move if it was unexpected well enough to score it good.

Example.

If you follow the moves with analysis, Stockfish shows +6, computer plays move and it is different than expected.

1 seconds +2
2 seconds +3.5
4 seconds +4.8
5 seconds + 5.7

The above is just an example of what I often seen in games with Stockfish. Based on the speed cara will score the alternative moves as very low (SenseRobot for example).

To offset this it needs time of search.

I would recommend that he adds a secondary analysis which the user can control by either setting the per move search or ply search that the user wants.

Also Cara should consider providing ratings for after opening book as well (game play rating so you can physically see where a difference lies)

Regards

Nick

Hallo Kurt

Danke für deine Frage: Ich setze nochmal alle links hier rein damit man nicht so suchen muss:

Chessgate letzter Update mit Timeout fix und screen popup fix, das war eigentlich eine Spezial Anfertigung für das Spacious Mind Turnier aber es wurde da schon getestet und ist also ok:

https://drive.google.com/drive/folde...usp=drive_link

Pioneer 2 Soviet Chess Computer Windows version :

https://drive.google.com/file/d/1sSi...ew?usp=sharing

ChessGate online play :

https://playchessgate.com/

Pioneer 2 online Play :

https://play.pioneer2chess.com/


Hoffentlich is das jetzt übersichtlicher, wir hatten halt viel Anpassungen mit hilfe des Forums

Beste Grüße Arne

Hallo Arne
Vielen Danke und Grüsse
Kurt

Thanks a lot Arne for this fix (a small pop-up window letting you choose to continue or not the game when flagged). Thanks to which I could start playing with ChessGate (PC-version, without this fix, it was a no-go for me).

I appreciate the nice position commentary that acts like a coaching help, yet I think there is a glitch about reporting backward pawns (each time I saw one reported, it was actually an advanced pawn, not a backward one). Please check! I think I also spotted some questionable comments about "holes in the position" but I didn't pay enough attention to that, I am not so sure I am right. I trust Thorsten could help double-checking about.

Anyway a nice piece of software, congrats!

MfG,
Eric

Hi Eric,

Thank you for your kind words and for the detailed feedback!

Regarding the position commentary — you are absolutely right, and I want to be transparent: this is still a beta feature that is not yet fully developed. There are known issues there that I am aware of and plan to address. The move suggestions are working well, but certain parts of the commentary still need significant improvement, and I will need to write a better prompt structure for that in the next build.

Since the core focus was on getting the play against the AI personalities right, that is where I concentrated my efforts first. The commentary will follow, but I want to do it properly rather than rush it.

The project has also grown quite large now, so even small changes require care to avoid breaking anything. The approach going forward will be incremental fixes — one step at a time.

Thanks again for testing it and for the honest feedback. That is exactly what helps!

Best regards,
Arne