On-line engine blitz tourney May

Discussion of chess software programming and technical issues.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Joost Buijs
Posts: 846
Joined: Thu Jul 16, 2009 8:47 am
Location: Almere, The Netherlands

Re: On-line engine blitz tourney May

Post by Joost Buijs » Sun May 26, 2019 5:33 am

MikeB wrote:
Sun May 26, 2019 1:31 am
flok wrote:
Sat May 25, 2019 9:29 pm
hgm wrote:
Sat May 25, 2019 9:27 pm

Code: Select all

:Tourney Players: Round 9 of 9 
:     Name              Rating Score Perfrm Upset  Results 
:     ----------------- ------ ----- ------ ------ ------- 
: 28 +Spartacus         [1470]  6.5  [1389] [  68] +26w +22b -21w -25b -23w +29b +32w -31w -18b 
: 31 +EmblaTwo          [1163]  6.0  [1363] [ 448] +33w +32b -26w +35b -29b +34w =30w +28b -19w 
Nemeton was suffering from a bad connection, which caused it to forfeit all initial games on time. It was forfeited from the tourney after it permamently disconnected.
Something's wrong with this system of play: emblatwo won of spartacus, yet, it has a lower rating and score.
You to have to look the ratings of the opponents, the score is just one factor - it's all automated and tried and true. Also the tournament score shown here is modified. High rated engines automatically get more points (fictitious) to their score to improve pairings. The real final results , would not include those fictitious points. Not sure if HGM plans to publish that or not - but nobody really cares - it's just for fun. MCBrain scored 8.5 out of 9, but is shown with 13.5 point. For pairing purposes, MCBRain had 5 points added at the start of tourney to ensure it would not get paired with an engine that is rated over 1000 points below it.
Maybe HGM can add a column with true scores? I assume that should not be to difficult.

This tourney is getting harder and harder each time, I still play with my old engine from 2013, maybe I should should ditch it and start working on the new one that is nearly finished, but you know how it goes. Even now that I'm retired from my job I still have very little time to work on the engine.

User avatar
Posts: 23230
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller

Re: On-line engine blitz tourney May

Post by hgm » Sun May 26, 2019 8:09 pm

The score without the starting points is even more meaningless as with those. Because it does not account for the fact that the players starting with very few points only have played very weak opponents, while those starting with many points had to face tough opposition. One could see the McMahon points as a (most of the time self-fulfilling) guess for the difference in opponent quality. Spartacus might have scored 0.5 actual points less than EmblaTwo, but since it scored those against much stronger opponents it still had the better performance. The performance ratings that are calculated, which should be the most accurate measure (although I have no idea how mamer exactly calculates those). This is indeed higher for Spartacus than for EmblaTwo. All in all there is a pretty good correlation between these performance ratings and the total points including the bonus points. There would hardly be any correlation at all between these ratings and the actual points.

E.g. micro-Max would be ranked above Rookie, based on true points (5 vs 4).

Post Reply