Wow, beeindruckend wie der TCAII gegen London gewinnt. Danke fürs teilen!

Der London Phoenix steht ja mit 2460ELO in der Wikiliste. Aber Du hast ja den neuen Phoenix? Hast Du eine Schätzung, welche ELO Zahl der mit London erreicht?

Viele Grüße
Jürgen

PS: Bei meinem Proelum war auch dieser Aufkleber drauf, bei dem die Farbe nicht gut hielt. Habe ich auch abgemacht...

Hallo Jürgen,
Dann gibt es ja jetzt schon 2 Chess Academy plus

Ich denke der London spielt ca 50 Elo stärker

Viele Grüße
Markus

Hello all,

I have breaking news... aggressivity definitely fits well the Proelium (aka Talking Chess Academy Plus and -maybe?- featuring the same chess engine as the Talking Chess Academy II).

To assess the most efficient style out of five available ones, I leveraged the "Spacious Mind reloaded" test. The one based on five classic games from the 18th century, updated using Stockfish 16 and Komodo Dragon 3.3 (details here).

Well, here are the resulting table (style number is between brackets, level was set to 15 secs/move), and graph:


I also checked the results filtered for White moves only on one hand, and also only black moves on the other hand; they both display the same result: the more aggressive the style, the better the result.

MfG,
Tibono

Miniaturansicht angehängter Grafiken

   

Hi Eric,

Thanks for the Test. My 10 game normal style should be ending today. So I might try Style aggressive next and see if there is an improvement in a 10 game match against the robot.

Regards
Nick

Interesting stuff. Do you have an idea of how closely those numbers may track real OTB play? For example, the TCA II (same as the Tabutronic Proelium), based on the 40 games I've played with it at 30 secs/move, using the "Normal" style, has scored around 2725 Elo. At longer time controls it's likely even stronger. Also, do you think, since the Aggresive setting of 5 scored almost 130 points more than the Normal setting, a similar improvement in OTB performance will occur?

I do trust Style 5 setting will perform percertibly stronger than normal/3 setting. Up to what an amount, I can't really say. The Elo values returned by the test should be correct as an order of magnitude; yet not that accurate. What is clear though is the trend: the higher the style setting, the stronger the play. Pretty sure.

Das der Proelium Schachcomputer bei dem Preis ein richtiges Sensorbrett hat, finde ich sehr gut. Aber ansonsten, was hat so ein Gerät eigentlich für ein Schachprogramm installiert? Deutsch kann es laut Webseite leider nicht.

Hello dear all,

as planned, I ran the Khmelnitsky test. The computed score is 2233, that's in between the Millennium CGP first version (2144) and the Millennium The King Performance (2339).

Here is the usual graph:



I point out the unusual strength regarding standard endgame positions. There is knowledge inside!

I also drew the % achievement graph to compare to average skills of a human player achieving the same Khmelnitsky test performance, here you are:



The typical human player is better at strategy, endgame and sacrifice - that's rather common a pattern - but he is as well a better attacker. The Proelium leverages recognizing threats, strong defense and counterattack (these are usual strengths of chess programs).

Conditions were: Level TOU3 eventually halted once the maximum granted thinking time is reached (I use as a standard for this test 3 minutes to compute a position, if required 1 minute to compute a suggested move). I used the most aggressive setting (Style 5) as it performed best with the Spacious Mind reloaded test; I consider this choice somehow confirmed by the slightly weak score achieved for attack, despite the use of style 5. Pushing the aggressivity bar looks required. Of course, actual games are needed to make sure this choice pays back, maybe depending on the opponent's level (as suggested by Nick).

MfG,
Tibono

Miniaturansicht angehängter Grafiken

   

Hi Eric

That is almost a perfect circle. It is a shame though that it did not score higher.

Regards
Nick

ps. I just looked at the Robot website and it says this:

Self-Set Chess Game

You can set up any position you want and choose from different AI levels to play against. It’s a great way to analyze your own games, test strategies, or explore specific situations.

So maybe I need to reinstall the App and do some tests as well sometime.

Regards
Nick

habt ihr den Proelium hier bestellt? https://tabutronic.com/product/proelium/
oder auf Amazon?

Rudi