nTCEC Season 1 - Stage 2b - starts March 11th at 14:00 UTC+1

By Thomas Müller Date 2013-03-11 05:38

Hi Martin,

you can post www-links better within [url]...[/url] you can use the buttons in the top

Then the links are directly useable

I'm happy you are back with your TCEC!

best wishes
thomas

By Martin Thoresen Date 2013-03-11 09:07

Hi Thomas!

Yes, I know, it's just that I forget that I have to do it manually in this forum.

Also, I've tried to register here but without success, so that I can edit posts and don't have to wait until a mod approves it. I never get an email sent for verification...

Best,
Martin

By Thomas Müller Date 2013-03-11 09:49

ok i know....sorry!
we have trouble with the register here.
I make a register for you and send you the login data per mail.
Which mail can i use?

best
thomas

By Martin Thoresen Date 2013-03-11 10:33 Edited 2013-03-11 11:18

Hi Thomas,

Thanks, that would be great. You can use webmaster at tcec-chess . net

Best,
Martin

By Thomas Müller Date 2013-03-11 11:18

done

By Martin Thoresen Date 2013-03-11 11:21

Thanks Thomas, works great.

By Martin Thoresen Date 2013-03-11 12:26

Now live!

http://www.tcec-chess.net/live.php

By Claude Gable Date 2013-03-11 21:16

Wie viele Kn/s hatte Houdini 3 am Anfang der Partie bzw. nach den ersten paar Zügen?

By Martin Thoresen Date 2013-03-13 11:21

Standings after 2 of 14 rounds:

Code:

N Engine          Pts   SB G V K C S Q H S 

1 Gull R375       2.0 1.00           1 1   
2 Vitruvius 1.19  1.5 1.25       1     =   
3 Komodo 4534     1.5 1.25         1 =     
4 Chiron 1.5      1.0 0.00   0           1 
5 Shredder 12     1.0 0.00     0         1 
6 Quazar 0.4      0.5 0.75 0   =           
7 Hannibal 200213 0.5 0.75 0 =             
8 Scorpio 2.75    0.0 0.00       0 0

By Martin Thoresen Date 2013-03-16 02:44

Standings after 4 of 14 rounds:

Code:

N Engine          Pts   SB K G V S Q C S H 

1 Komodo 4534     3.0 5.75   =   1 =     1 
2 Gull R375       3.0 4.50 =       1   = 1 
3 Vitruvius 1.19  2.5 4.00       = = 1   = 
4 Shredder 12     2.5 3.75 0   =     1 1   
5 Quazar 0.4      1.5 3.25 = 0 =       =   
6 Chiron 1.5      1.5 1.50     0 0     1 = 
7 Scorpio 2.75    1.0 2.25   =   0 = 0     
8 Hannibal 200213 1.0 2.00 0 0 =     =

Finished games:
http://www.tcec-chess.net/stage_2b.php

By Martin Thoresen Date 2013-03-17 23:52

Standings after 6 of 14 rounds:

Code:

N Engine          Pts    SB K V G Q S H C S 

1 Komodo 4534     4.5 12.00   = = = 1 1   1 
2 Vitruvius 1.19  4.0 12.50 =   1 = = = 1   
3 Gull R375       4.0 10.25 = 0   1   1 1 = 
4 Quazar 0.4      3.0  8.75 = = 0   1   = = 
5 Shredder 12     3.0  6.25 0 =   0   = 1 1 
6 Hannibal 200213 2.5  5.50 0 = 0   =   = 1 
7 Chiron 1.5      2.0  3.75   0 0 = 0 =   1 
8 Scorpio 2.75    1.0  3.50 0   = = 0 0 0

Finished games:
http://www.tcec-chess.net/stage_2b.php

By Claude Gable Date 2013-03-18 07:10

Etwas enttäuschend war die Partie Vitruvius gegen Komodo, wo Vitruvius den Gewinn nicht fand und Ka2 spielte mit Remis.

Sogar einige Spieler im Chat sahen den Gewinnzug.

By Michael Scheidl Date 2013-03-18 07:28

Die Singlecore-Engine Quazar 0.4 ist bis jetzt die größte Überraschung dieser TCEC-Saison.

By Martin Thoresen Date 2013-03-20 22:07

Standings after 9 of 14 rounds:

Code:

N Engine          Pts    SB Ko Vi Qu Gu Ch Sh Ha Sc 

1 Komodo 4534     6.5 26.00    =  == =  =  11 1  1  
2 Vitruvius 1.19  6.5 25.25 =     =  1  1= =  =1 1  
3 Quazar 0.4      5.5 24.50 == =     01 =  1  1  =  
4 Gull R375       4.5 19.00 =  0  10    1  =  10 =  
5 Chiron 1.5      4.0 14.00 =  0= =  0     0  =  11 
6 Shredder 12     4.0 13.50 00 =  0  =  1     =  1= 
7 Hannibal 200213 3.5 13.25 0  =0 0  01 =  =     1  
8 Scorpio 2.75    1.5  7.00 0  0  =  =  00 0= 0

Finished games:
http://www.tcec-chess.net/stage_2b.php

By Claude Gable Date 2013-03-21 07:16

Wiso ist Vitruvius zweiter, obwohl er stärkere Gegner besiegt hatte?

By Martin Thoresen Date 2013-03-22 00:46

Claude,

http://en.wikipedia.org/wiki/Neustadtl_score

By Martin Thoresen Date 2013-03-22 22:04

Standings after 11 of 14 rounds:

Code:

N Engine          Pts    SB Vi Ko Qu Gu Ch Ha Sh Sc 

1 Vitruvius 1.19  8.5 42.25    =  =1 1  1= =1 =1 1  
2 Komodo 4534     7.5 36.25 =     == =1 =  10 11 1  
3 Quazar 0.4      6.5 31.25 =0 ==    01 =  1  1  =1 
4 Gull R375       5.5 25.00 0  =0 10    1  10 =  =1 
5 Chiron 1.5      5.5 23.25 0= =  =  0     == 01 11 
6 Hannibal 200213 5.0 26.25 =0 01 0  01 ==    =  1  
7 Shredder 12     4.0 17.25 =0 00 0  =  10 =     1= 
8 Scorpio 2.75    1.5  8.00 0  0  =0 =0 00 0  0=

Finished games:
http://www.tcec-chess.net/stage_2b.php

By Martin Thoresen Date 2013-03-25 23:09

Final standings after 14 rounds:

Code:

N Engine           Pts    SB Vi Ko Ch Qu Ha Gu Sh Sc 

1 Vitruvius 1.19  10.0 64.25    == 1= =1 =1 1= =1 1= 
2 Komodo 4534      9.5 57.75 ==    == == 10 =1 11 11 
3 Chiron 1.5       8.0 49.25 0= ==    =1 == 01 01 11 
4 Quazar 0.4       7.5 46.75 =0 == =0    1= 01 1= =1 
5 Hannibal 200213  6.5 42.00 =0 01 == 0=    01 == 1= 
6 Gull R375        6.5 41.00 0= =0 10 10 10    == =1 
7 Shredder 12      5.5 33.50 =0 00 10 0= == ==    1= 
8 Scorpio 2.75     2.5 18.00 0= 00 00 =0 0= =0 0=

Vitruvius, Komodo, Chiron and Quazar moves on to Stage 3 which will start in a few days, with the new 16-core Xeon machine.

By Michael Scheidl Date 2013-03-26 05:56

Zitat:

with the new 16-core Xeon machine.

Das wird der Überhammer. Die (dzt. noch) Singlecore-Engines Komodo, sowie relativ überraschend Quazar 0.4, werden einen schweren Stand haben.

Die anderen vier sind Houdini, Stockfish, Rybka und Hiarcs. Da sehe ich hauptsächlich ein Duell um Rang 4 zwischen Vitruvius, Hiarcs und möglicherweise Chiron. Vitruvius hat in Saison 1 bisher keine einzige Partie Partie verloren und insgesamt 15,5/21 geholt.

By Thomas Müller Date 2013-03-26 10:03

wird mit P-off oder P-on gespielt?

By Michael Scheidl Date 2013-03-26 13:44 Edited 2013-03-26 13:46

So wie es aussieht, (leider) mit P-off...

Es gibt hierzu eine Umfrage für die nächste Saison im TCEC-Forum, wo die Befürworter von P-on bis jetzt in der Minderheit bleiben.

By Martin Thoresen Date 2013-03-26 13:45

The poll is from Season 2 and beyond. No changes will be made to the current Season with regards to ponder or time control.

By Thomas Müller Date 2013-03-26 13:51

hmm...16 core bringt aber meiner meinung nach nicht viel bei den meisten engines.
2x8 p-on wäre da sicher besser.
Aber gut....

By Martin Thoresen Date 2013-03-26 17:07

You are wrong, most engines that can use 8 cores can also use 16 cores.

There are a few exceptions, but not many.

By Michael Scheidl Date 2013-03-26 19:37

He means, most engines will not gain much from 16 compared to 8 cores.

By Martin Thoresen Date 2013-03-27 00:16

Well, how does he know that? I've seen no tests around that confirms either way.
Claims like that should be followed up with data.

By Ingo Bauer Date 2013-03-27 05:20

How can anyone really deliver data? No one owning a 16 core machine is willing to run some serious tests as they are so excited about high node counts and don't wont to spoil there investment ...

But it is well known that chesss is scaling not too well and with every doubeling of the cores it is less efficiant (See Amdals and Gustafsons law here: http://en.wikipedia.org/wiki/Parallel_computing). Of course some are doing a little better and some a litte worse, but in general the gain from 8 to 16 is minimal. Every programmer would prefer a double speed single then a half speed dual. If you extrapolate that to a 16 core system not much is left of gaining strength there. Of course your nodes go up all the time but the time to depth is not doing this nessesarily.

This is a good start point for some research on the matter: http://chessprogramming.wikispaces.com/Parallel+Search

In general there is no disscusion about the fact but sometimes it is confused with cluster architecture. Of course the laws are valid there to, but the use of 64 or more cores in a cluster is a little bit more effective with different parallel implementations. (Can be found in the programming wiki). In short: 2x8 cores individualy is more effective than 1x16 cores - and of course most effective would be 16x1 core but no one wants to see that

Btw: I miss the guest function in your chat window. I am not willing to use Google, Twitter, Facebook or to sign in to a complete unknown social thing to make a chat. I really would appreciate if you can bring that back.

Bye
Ingo

By Michael Huber Date 2013-03-27 06:56

What could such a serious test look like?
If it is not too time consuming I might have an opportunity to do such a test.
I could imagine to play two equal tournaments exept the no of threads used by one particular engine (eg. 8 versus 16) and compare the results.
Any suggestions ?

Best regards, Michael

By Ingo Bauer Date 2013-03-27 14:24

[quote="Michael Huber"]
What could such a serious test look like?
If it is not too time consuming I might have an opportunity to do such a test.
I could imagine to play two equal tournaments exept the no of threads used by one particular engine (eg. 8 versus 16) and compare the results.
Any suggestions ?

Best regards, Michael
[/quote]

Hehe, "time" might be the biggest problem!

But the above law are actually already proven so there is no need to do that again. If you run a test and you get a bigger elo increase from 2 to 4 then from 1 to 2 cores your setup is simply wrong or yout do not have enough games.

Actually you have to make a big ranking list with many engines on one core first. Then you choose one engine (presumably one in the mid field) and run EXACTLY (including openings, opponents and time control) the same engine on 2 cores into that field of engines again. Then you repeat with 4, 8 and 15 (16) cores. The more games the better.
With very short time controls MP Engines might show a smaller increase a bit earlier with a longer time control it might be later.

Send the computer to me - I will run 1 engines with 2850 games for 1, 2, 4, 8 and 16 cores into the IPON if you pay the power bill

Another possibility is to use just two engines. Eg. Houdini on 1 core, and and Gull II on 1,2,4,8 and 16.YOu need a lot of oppening position as you realy want to repeat exactly the same with more cores. I have a 5000 position set (send me a PM if you want to have that). With changing colors you get 10000 games. NOw you choose a time control and start.
Assuming a 3 + 2 and a 10 min per game the first round with 1 core vs. 16 (Ponder OFF) is already finished after 69 days. 1 vs 8 is done in 35, 1 vs 4 in 17 days, 1 vs 2 in 9 1 vs 1 in a little less than 5 days. Overall the test would last ~135 days

Of course you could choose less games but then the result is less "secure"

Good luck
Ingo

PS: Wir können das auch gerne auf Deutsch machen.

By Michael Huber Date 2013-03-28 10:09

Hi Ingo,
what do you think about the following test setting.

Time: 30"+1"
1500 games against 10 opponents
75 starting positions (Silver suite standard+gambit)

Two similar tournaments.
1.) All engines 8 Cores
2.) All engines 8 cores but one engine (mid class e.g. HIARCS) with 16 cores (24 would also be possible)

Comparison of the results.

Michael

By Martin Thoresen Date 2013-03-27 11:04

Ingo,

The ponder hit rate for Stage 1 was 50%. See this post: http://www.tcec-chess.net/viewtopic.php?f=13&t=9&start=20#p80
So if the ponder hit rate indeed is about 50% with this long time control, it means that about half of the cpu time spent during pondering is wasted if run with ponder on.

I am not against ponder on, the point right now is that the ponder rules are already set for the remainder of Season 1: ponder off, and will not be changed. That is why the poll is there, about Season 2 and the future.

I can tell you one thing though: 16 x 1 cores is not going to happen in nTCEC.

The guest function is enabled again... for now.

By Ingo Bauer Date 2013-03-27 14:03

[quote="Martin Thoresen"]
Ingo,

The ponder hit rate for Stage 1 was 50%. See this post: http://www.tcec-chess.net/viewtopic.php?f=13&t=9&start=20#p80
So if the ponder hit rate indeed is about 50% with this long time control, it means that about half of the cpu time spent during pondering is wasted if run with ponder on.

I am not against ponder on, the point right now is that the ponder rules are already set for the remainder of Season 1: ponder off, and will not be changed. That is why the poll is there, about Season 2 and the future.

I can tell you one thing though: 16 x 1 cores is not going to happen in nTCEC.

The guest function is enabled again... for now.

[/quote]

Hi Martin

I have not mentioned Ponder ON with a word, thats a different story and your business ... I am refering to the fact that from 8 to 16 cores the win in elo is marginal regardless of Ponder condition ...
You can read again the original post, your answer and question for data and I provided some evidence - thats all.

Good that the 'guest' is working again!

Bye
INgo

By Martin Thoresen Date 2013-03-27 14:34

Ingo,

Yes I am well aware that MP gives "lesser and lesser value" the more cores you have.
But the point is that I don't see any point in running with 8 cores if ponder is off, when there are 16 cores available. That's all. Hence my reference to ponder when talking about 8 cores.

I am not interested in running several events at the same time (for example 2 events with 8 cores each, ponder off), that would be too much work for me.

Best,
Martin

By Ingo Bauer Date 2013-03-27 15:03 Edited 2013-03-27 15:11

[quote="Martin Thoresen"]
Ingo,

Yes I am well aware that MP gives "lesser and lesser value" the more cores you have.
But the point is that I don't see any point in running with 8 cores if ponder is off, when there are 16 cores available. That's all. Hence my reference to ponder when talking about 8 cores.

I am not interested in running several events at the same time (for example 2 events with 8 cores each, ponder off), that would be too much work for me.

Best,
Martin
[/quote]

Martin, if you know that you get "lesser and lesser value" why did you ask?



>He means, most engines will not gain much from 16 compared to 8 cores.
>Mit freundlichen Grüßen, M.S. --> P.B.

>Von Martin Thoresen (profil) Datum 2013-03-27 00:16
>Well, how does he know that? I've seen no tests around that confirms either way.
>Claims like that should be followed up with data.

I don't get your pointing at me, I just showed some IT technologie background aka "data"...!?

And btw, I fully agree. If you have 16 cores and you play ponder off use them! I dont have a problem with it, I ONLY answered your question ... (which seems to get dangerous nowadays)!

BYe
Ingo

By Martin Thoresen Date 2013-03-27 16:12

Yes, that was my point too, the 16 cores are there, I will use them.

And while I know about MP scalability, I guess the point is our definition of "most engines will not gain much".
What is much? 10 ELO? 100 ELO?

Regarding the data (proof):
If someone would play 30.000 games with the same engine (for example H3) at fast TC (not super fast), for example 4+2, with 8 cores vs 16 cores, then we have some data.

My guess is that no one has done this before, therefore it is just wrong to state that "most engines will not gain much".

Are you following my logic?

Best,
Martin

By Ingo Bauer Date 2013-03-27 16:20

[quote="Martin Thoresen"]
Yes, that was my point too, the 16 cores are there, I will use them.

By Robert Houdart Date 2013-03-27 14:36

[quote="Ingo Bauer"]
I am refering to the fact that from 8 to 16 cores the win in elo is marginal regardless of Ponder condition ...[/quote]
Ingo, do you have any real data to validate this claim? (not just a link to a general Wiki article about Parallel Computing).

And if you truly believe this, why did you play the Houdini 3 - Rybka Cluster match with Houdini 3 on Clemens' 16-core computer, and not on a 6-core computer - as I suggested before the match?

Robert

By Ingo Bauer Date 2013-03-27 14:45 Edited 2013-03-27 14:50

[quote="Robert Houdart"]
[quote="Ingo Bauer"]
I am refering to the fact that from 8 to 16 cores the win in elo is marginal regardless of Ponder condition ...[/quote]
Ingo, do you have any real data to validate this claim? (not just a link to a general Wiki article about Parallel Computing).

And if you truly believe this, why did you play the Houdini 3 - Rybka Cluster match with Houdini 3 on Clemens' 16-core computer, and not on a 6-core computer - as I suggested before the match?

Robert
[/quote]

Yes, I really believe (actually I don't think that this is something to "believe"!) that and I do not have data - how can I as I dont have a 16 core comp.

I played with a 16 core as that was the best available I cóuld get and I hoped that from 8 to 16 there might be still a 20 ELo increase - but mabe the 8 core would be as good, who knows? ANd I hoped that Houdini, as you develop on a 4 CPU system (right?) might be a better enigne utilizing the 16 cores ... The problem is if that Elo increase is worth the money?

But becaue you are asking, I have a question back: Do you believe that from 8 to 16 cores Houdini gains as much Elo as from 1 to 2 core?

Bye
Ingo

By Ingo Bauer Date 2013-03-27 15:08

PS: AND I asked for a donation of credits for a match vs the cluster with the best available computer! If I would have played with a 6 core machine many of the donors would be disapointed!

By Robert Houdart Date 2013-03-27 15:12 Edited 2013-03-27 15:16

[quote="Ingo Bauer"]
But becaue you are asking, I have a question back: Do you believe that from 8 to 16 cores Houdini gains as much Elo as from 1 to 2 core?[/quote]
The gain from 8 to 16 cores will be smaller than from 1 to 2 cores, just like the Elo gain by doubling the TC from 10 to 20 minutes will be smaller than by doubling from 1 to 2 minutes.
But it's a big step from understanding that the gains are inevitably decreasing (an objective truth), to saying that the gain from 8 to 16 cores is marginal or negligible.

As far as I can see, playing all the engines with 16 cores will provide a new landmark in computer chess tournaments, and will generate some extra excitement for the spectators.
So why not give some enthusiasm to Martin instead of producing negative vibes?

Robert

By Ingo Bauer Date 2013-03-27 15:19 Edited 2013-03-27 15:21

[quote="Robert Houdart"]
[quote="Ingo Bauer"]
But becaue you are asking, I have a question back: Do you believe that from 8 to 16 cores Houdini gains as much Elo as from 1 to 2 core?[/quote]
The gain from 8 to 16 cores will be smaller than from 1 to 2 cores, just like the Elo gain by doubling the TC from 10 to 20 minutes will be smaller than by doubling from 1 to 2 minutes.
But it's a big step from understanding that the gains are inevitably decreasing (an objective truth), to saying that the gain from 8 to 16 cores is marginal or negligible.

As far as I can see, playing all the engines with 16 cores will provide a new landmark in computer chess tournaments, and will generate some extra excitement for the spectators.
So why not give some enthusiasm to Martin instead of producing negative vibes?

Robert
[/quote]

Negative vibes? It seems I missed some conversations somewhere else! He asked for data, I gave some ...

Actually I am puzzled about the negative vibes toward me because I answered a question! Martin and you basicaly agree with what I have written, nonetheless you (both) accusing me of something? I am really stunned a bit here!

Ingo

By Martin Thoresen Date 2013-03-27 16:07

I am definitely not accusing you of anything Ingo.

Thanks Robert, for clearing up a few things.

By Ralf Mueller Date 2013-03-26 18:52

I'm a bit disappointed concerning the use of the 16-core machine. So several engines from beginning on have no chance for advancing to the next round.
You're changing the conditions while the tournament is running. That's unfair for the 1-core engines, which have advanced to stage 3 and unfair for very good scaling engines, which are eliminated, but would stay in the tournament, if you would play with 16 cores from the beginning on.

Of course you can't play all games with one core for equal chances, because 4 cores are now the standard. But this argument doesn't make any sense for the use of 16 cores, when they're several engines, which can't use them.

I hope you see my point.

Best regards!

By Martin Thoresen Date 2013-03-27 00:14

I wouldn't do this if the programmers did not approve of it.

http://www.tcec-chess.net/viewtopic.php?f=2&t=85
.

By Ralf Mueller Date 2013-03-27 06:45

I know, but this doesn't change my arguments: it is unfair, not only for the 1-core engines and it gives a distortet result without any practical meaning.

By Michael Scheidl Date 2013-03-27 10:10

TCEC ist kein Ranglistenbewerb, das war ja nie das Ziel. Es ist ein "Showcase" mit Beispielen bestmöglichen Computerschachs. Die relative geringe Partienzahl enthebt TCEC der Verantwortung jeglicher statistischer Relevanz. Beispielsweise ist ja Critter bereits ausgeschieden - und Quazar immer noch drin! So ist das halt in einer echten Praxis, im Gegensatz zu der sterilen Laborsituation bei Ranglistentests.

By Ralf Mueller Date 2013-03-27 10:29

Das ist mir auch klar und das schätze ich an dem Wettbewerb, dass eben nicht nur Zahlen das einzige Resultat des Turnieres sind.

Du setzt jedoch in deinem Beitrag Fairness mit statistischer Relevanz gleich. Jedes Turnier, auch jedes "Showcase"-Turnier sollte unter fairen, gleich bleibenden Bedingungen abgehalten werden.
Ich habe sicher nichts gegen 16 Kerne, aber dann bitte gleich zu Beginn und nicht während des Turniers die Wettbewerbsbedingungen ändern.

Beispiel: Ein Spaßveranstaltung mit Carlsen, Kramnik, Anand, Topalov - Blind-Blitz. Während des Turnieres entscheidet der Turnierleiter, dass die zwei besten im Finale nun lange Partien mit Ansicht des Brettes spielen, damit das Niveau steigt. Was hat das Turnier dann noch für einen Wert?

Ich frage euch, ob ihr wirklich denkt, als Zuschauer steigt das Vergnügen von 4 auf 16 Kernen in dem Maße, dass es diese Unfairness rechtfertigt? Auch Spaßturniere sollten gewissen Richtlinien folgen, zumal dieses Turnier nicht ganz Spaß ist, da es von vielen Zuschauern als inoffizielle Computer-WM angesehen wird.

By Michael Scheidl Date 2013-03-27 11:50

Zitat:

(...) von vielen Zuschauern als inoffizielle Computer-WM angesehen wird.

Eine große Anerkennung, aber seitens TCEC wurde niemals so etwas behauptet. - Ich sehe es als eine hochklassige Exhibition welche in Stufen abläuft, sodaß ich kein Problem darin sehe wenn zwischen (nicht während!) Stufen sich die Bedingungen ändern.

Es gibt auch dem Prinzip "Augenmaß" den Vorzug: Es war bekannt daß ein 16 Core-Xeon einsatzbereit vorhanden ist. Warum soll man Season 1 justament monatelang auf einer nachrangigen Hardware dahinschleppen? Das ist doch nicht attraktiv. Es ist auch nicht unfair, denn innerhalb der jeweiligen Stufen waren die Bedingungen für alle gleich. Es gibt kein Grundgesetz daß alle Stufen mit derselben Hardware absolviert werden müssen.

Insofern bin ich persönlich hochzufrieden mit TCEC; die Meinungen werden geteilt sein...

By Ralf Mueller Date 2013-03-27 12:13

Da man sich für die nächsten Stufen qualifizieren muss, widerspricht es der Turnierlogik, wenn man sich unter anderen Bedingungen qualifiziert, wie dann das Finale gespielt wird. Wie schon in meinem Beispiel aufgezeigt, wäre eine Qualifikation für das WM-Finale im klassischen Schach über ein Blind-Blitz-Turnier einfach nicht angemessen. Natürlich ist es alles ein großes Turnier und nicht mehrere kleine Turniere, die in Stufen abgehalten werden. Insofern werden doch während des Turnieres die Regeln geändert.

Was heißt denn "monatelang dahinschleppen"? Die Partien dauern durch eine schnellere Hardware doch nicht kürzer bzw. schneller, was deine Formulierung asoziiert. Sicherlich spielen 16 Kerne stärker als 4 Kerne, keine Frage - mein Punkt ging in die Richtung, ob es für die Fans außer der größeren Zahl irgendeinen nennenswerten Unterschied macht. Darauf hast du nicht geantwortet.

Es gewinnt nun nicht mehr die Engine, die die beste Suche und Bewertung zeigt, sondern die Engine, die am besten skaliert. Das senkt für mich leider den Anspruch des Turnieres, den es für mich seit Beginn hatte.

Auch ich bin ein großer Fan der TCEC und möchte die Idee und Umsetzung ausdrücklich loben. Nichtsdestotrotz bin ich mit dieser Entscheidung unzufrieden.