TCEC results after all programs have played 6 rounds.

[beram]

It's far from over, but here are the standings after 6 rounds.

Please note that some programs have played more challenging games than others, so this table only tells a partial story. In particular Komodo has not played black yet against Stockfish, although it did manage to draw with black against Houdini and win with white against Stockfish. Komodo also took a loss against Critter. Komodo has played 4 blacks and 2 whites. Stockfish has also played 4 blacks and 2 whites so you can expect it's slow start to improve, I expect it to finish this stage in the top 3 easily as it's still very early.

Houdini has played 5 whites and 1 black but shares 2nd place so it's having a bad tournament in my opinion as I would guess that with 5 whites and 1 black it would normally be on top. But anything can happen and 6 games doesn't tell much.

I believe Komodo, Stockfish and Houdini will finish in the top 3 here and it's pretty much impossible to predict the order, by I will take a crack at it and say the most likely finish order is Houdini, Komodo, Stockfish. This tournament and other results puts Houdini's performance at very long time controls in question so I am least clear on Houdini, but I still give it a slight nod for first.

Code: Select all

Nr Engine           Pts    SB K N B R H C H G S J 

 1 Komodo 1121.05   4.0  9.50   =     = 0   1 1 1 
 2 Naum 4.6         3.5 11.00 =   = =     1 = =   
 3 Bouquet 1.8b     3.5 10.25   =     = = = = 1   
 4 Rybka 4.1        3.5  9.75   =     = 1 =   = = 
 5 Houdini 9601     3.5  9.25 =   = =   =   =   1 
 6 Critter 1.6a     3.0 10.25 1   = 0 =   = =     
 7 Hiarcs 14        3.0  7.25   0 = =   =     = 1 
 8 Gull 2.3         2.5  7.25 0 = =   = =       = 
 9 Stockfish 151013 2.5  6.00 0 = 0 =     =     1 
10 Junior 13.3      1.0  3.00 0     = 0   0 = 0   

It is far to early to make conclusions
Just one round later after another white(!) loss for Stockfish against Gull, one can also say that Stockfish perhaps will not qualify
And after a beautiful positional black win for Houdini you now must say "the new Houdini 4 dev plays a hell of nice chess"

[Don]

It's far from over, but here are the standings after 6 rounds.

Please note that some programs have played more challenging games than others, so this table only tells a partial story. In particular Komodo has not played black yet against Stockfish, although it did manage to draw with black against Houdini and win with white against Stockfish. Komodo also took a loss against Critter. Komodo has played 4 blacks and 2 whites. Stockfish has also played 4 blacks and 2 whites so you can expect it's slow start to improve, I expect it to finish this stage in the top 3 easily as it's still very early.

Houdini has played 5 whites and 1 black but shares 2nd place so it's having a bad tournament in my opinion as I would guess that with 5 whites and 1 black it would normally be on top. But anything can happen and 6 games doesn't tell much.

I believe Komodo, Stockfish and Houdini will finish in the top 3 here and it's pretty much impossible to predict the order, by I will take a crack at it and say the most likely finish order is Houdini, Komodo, Stockfish. This tournament and other results puts Houdini's performance at very long time controls in question so I am least clear on Houdini, but I still give it a slight nod for first.

Code: Select all

Nr Engine           Pts    SB K N B R H C H G S J 

 1 Komodo 1121.05   4.0  9.50   =     = 0   1 1 1 
 2 Naum 4.6         3.5 11.00 =   = =     1 = =   
 3 Bouquet 1.8b     3.5 10.25   =     = = = = 1   
 4 Rybka 4.1        3.5  9.75   =     = 1 =   = = 
 5 Houdini 9601     3.5  9.25 =   = =   =   =   1 
 6 Critter 1.6a     3.0 10.25 1   = 0 =   = =     
 7 Hiarcs 14        3.0  7.25   0 = =   =     = 1 
 8 Gull 2.3         2.5  7.25 0 = =   = =       = 
 9 Stockfish 151013 2.5  6.00 0 = 0 =     =     1 
10 Junior 13.3      1.0  3.00 0     = 0   0 = 0   

It is far to early to make conclusions

That's probably why I started with the statement, "It's far from over ..."

Just one round later after another white(!) loss for Stockfish against Gull, one can also say that Stockfish perhaps will not qualify
And after a beautiful positional black win for Houdini you now must say "the new Houdini 4 dev plays a hell of nice chess"

[lucasart]

Trying to draw conclusions from TCEC is always pointless. People should look at rating lists for that.

For example:
* in last TCEC, Critter got kicked out in Stage 2, yet it is certainly better than many of the other engines that passed that Stage 2.
* in the current TCEC, Stockfish has high chances of not making it through Stage 3, yet (if that happens) many of the qualified will be weaker than SF.

Stage 3 is only 18 games. No sensible conclusion can be drawn from such a ridiculously small sample. It's amazing how people extrapolate theories about scaling properties of engines based on TCEC results.

It's a knock-out format, and as such it is always prone to such effects. For example, in the last Winbledon Tennis championship, Raphael Nadal lost in the first round. Yet now he's #1 at the ATP ranking (comparable to rating lists, if you like). If you lost a single match you're out, and they don't let you play hundreds of other ones to do an average and build robust rating statistics from that. It's the name of the game (KO tourney).

[Don]

Trying to draw conclusions from TCEC is always pointless. People should look at rating lists for that.

So far nobody has tried to draw any conclusions, but everyone is quick to say that you should not draw conclusions. What's with that?

For example:
* in last TCEC, Critter got kicked out in Stage 2, yet it is certainly better than many of the other engines that passed that Stage 2.
* in the current TCEC, Stockfish has high chances of not making it through Stage 3, yet (if that happens) many of the qualified will be weaker than SF.

Stage 3 is only 18 games. No sensible conclusion can be drawn from such a ridiculously small sample. It's amazing how people extrapolate theories about scaling properties of engines based on TCEC results.

It's a knock-out format, and as such it is always prone to such effects. For example, in the last Winbledon Tennis championship, Raphael Nadal lost in the first round. Yet now he's #1 at the ATP ranking (comparable to rating lists, if you like). If you lost a single match you're out, and they don't let you play hundreds of other ones to do an average and build robust rating statistics from that. It's the name of the game (KO tourney).