Not logged inCSS-Forum
Forum CSS-Online Help Search Login
CSS-Shop Impressum Datenschutz
Up Topic Hauptforen / CSS-Forum / TCEC 14 Superfinal Mystery?
- - By Emil Vlasák Date 2019-02-25 09:00
I would like to open a conspiracy theory about TCEC 14 Superfinal.  It seems me, there were not two opponents here, but three.

The normal Stockfish played games say 1-22 and 66-100. In the first period he overcame his opponent and led by three points 6:3. And in the third period, LC0 was unable to get a full point at all.
But in the middle of the Superfinal I felt like that LC0 played-out positionally  Stockfish in almost every game. I was not the only chessplayer, who had written-off Stockfish at this time.

Such things happen in human‘s  matches, but not in computer‘s. Computers do not know fatigue and nervousness.

My suspicion falls on the Stockfish contempt factor.  The TCEC rules say:  This setting is not changed from the default unless it is a request from the programmer.  Should we understand that the programmer can ask for a change at any time?  If you click the king figurine icon beside the engine name in the TCEC screen, you get details about engine settings. But for some reason the contempt factor is not displayed here. 

So my question for TCEC:  was the contempt factor or any other parameters changed during the Superfinal?
Parent - By Benno Hartwig Date 2019-02-25 09:53 Edited 2019-02-25 10:00
"Häufchenbildung" ist normal und zu erwarten in solchen stochastischen Abfolgen.
Eine besonders gleichmäßige Verteilung von Siegen und Niederlagen hingegen hätte Verdacht erregen können. Weil eben das nicht zu erwarten wäre.

Wenn du zufällig 10 Pfennige auf 100 Plätze verteilst, dann wirst du hinterher Bereiche finden, wo die Pfennige irgendwie als "zu häufig"  wirken.

Mir gehet das übrigens regelmäßig so, wenn ich bei so einem Onlinediest zufällig einen Vorschlag für die Lottozahlen generieren lasse.
"Das ist doch nicht zufällig, hier in der Ecke sind doch viel zu viele!" ist eigentlich jedes Mal mein Eindruck.
Oder auch, dass sich da für mich dann Muster andeuten.
Zufall ist einfach so!
Das Umgehen von Mustern oder Häufungen ist ein Eingriff in die Zufälligkeit.
Parent - By Achim Müller Date 2019-02-25 11:06 Edited 2019-02-25 11:09
No conspiration!

I didn't follow all games during the super final but I guess I got a good grip on the whole match. First of all it's all about the opening choices. The match had games where both bots had good chances to get the iniative or even better, chances to win a game. Very often these were games including early pawn sacrifices or unbalanced pawn structures. And there were games where you could see already in the opening and due to very »dull« pawn structures that these games will pretty sure end in a draw.

The statistical deviation (winning and draw series) is most likely just based on these openings, nothing else. Furthermore there were at least three games, where lc0 had a winning position but failed to convert. So the match could have easily finished with a +2 for lc0 also. You can call it »unlucky«, but winning a winning position also belongs to the skills of chess. I can't remember a single game where Stockfish failed to convert an advantage into a win (but I might have missed such a game).

Ciao

Achim
Up Topic Hauptforen / CSS-Forum / TCEC 14 Superfinal Mystery?

Powered by mwForum 2.29.3 © 1999-2014 Markus Wichitill