OMG, someone who gets it.
...Except that going through every move of every chess game is not required to solve chess.
...except that it is. You only have theory and conjecture that imply otherwise.
...going through every move of every chess game is not required to solve chess.
...except that it is. You only have theory and conjecture that imply otherwise.
Here's proof that some of the total game tree can be excluded from the analysis, and yet a forced win for white is proven:
That's one way to look at it. Also, even when a section of the game-tree is relevant, it doesn't always have to be checked exhaustively to find the "best line". Two rooks and a king can checkmate a lone king. That can be solved deductively, without a brute force search.
More complicated deductive analysis is also possible. Brute force is the worst way to solve most problems, including chess. Almost all other analytical solving methods are an improvement compared to brute force.
The yearly event at Lieden is recognized as the World Championship.
World Computer Chess Championship (WCCC) is an annual event where computer chess engines compete against each other. The event is organized by the International Computer Games Association. It is often held in conjunction with the Computer Olympiad, a collection of computer tournaments for other board games.
The WCCC is open to all types of computers including microprocessors, supercomputers, clusters, and dedicated chess hardware.
In 2007, the reigning champion Junior declined to defend its title.
For the 2009 edition, the rules were changed to limit platforms to commodity hardware supporting at most eight cores,[1] thereby excluding supercomputers and large clusters. Thereafter, a parallel Software Championship was held instead, and unlimited hardware is still allowed in the championship proper.
| Event # | Year | Location | Participants | Winner |
|---|---|---|---|---|
| 1 | 1974 | Stockholm | 13 | Kaissa |
| 2 | 1977 | Toronto | 16 | Chess 4.6[2] |
| 3 | 1980 | Linz | 18 | Belle |
| 4 | 1983 | New York | 22 | Cray Blitz |
| 5 | 1986 | Cologne | 22 | Cray Blitz |
| 6 | 1989 | Edmonton | 24 | Deep Thought |
| 7 | 1992 | Madrid | 22 | ChessMachine (Gideon) |
| 8 | 1995 | Hong Kong | 24 | Fritz |
| 9 | 1999 | Paderborn | 30 | Shredder |
| 10 | 2002 | Maastricht | 18 | Deep Junior |
| 11 | 2003 | Graz | 16 | Shredder |
| 12 | 2004 | Bar-Ilan University, Ramat Gan | 14 | Deep Junior |
| 13 | 2005 | Reykjavík | 12 | Zappa |
| 14 | 2006 | Torino | 18 | Junior |
| 15 | 2007 | Amsterdam | 12 | Zappa[wccc 1] |
| 16 | 2008 | Beijing | 10 | HIARCS[wccc 1] |
| 17 | 2009 | Pamplona | 10 | Junior, Shredder, Sjeng[wccc 1] |
| 18 | 2010 | Kanazawa | 10 | Rondo, Thinker[wccc 1] |
| 19 | 2011 | Tilburg | 9 | Junior |
| 20 | 2013 | Yokohama | 6 | Junior |
| 21 | 2015 | Leiden | 9 | Jonny |
| 22 | 2016 | Leiden | 6 | Komodo |
| 23 | 2017 | Leiden | 4 | Komodo |
From 2010 a new tournament was introduced and held at the same location and during the same period as the World Computer Chess Championship. The rules for the World Chess Software Championship state that competing programs must run on machines with identical hardware specifications. Time control is game in 45 minutes with 15 second increment.[3][4]
| Event # | Year | Location | Participants | Winner | Hardware |
|---|---|---|---|---|---|
| 1 | 2010 | Kanazawa | 9 | Shredder[5] | Intel quad core Xeon 2.66 GHz, 8MB Hash |
| 2 | 2011 | Tilburg | 5 | HIARCS | Intel Core2 Duo, 1.7 GHz, 2MB Hash |
| 3 | 2013 | Yokohama | 6 | HIARCS | Intel quad core i7, 2.7 GHz, 16MB Hash |
| 4 | 2015 | Leiden | 8 | Shredder [1] | Intel quad core i7, 2.7 GHz, 16MB Hash |
| 5 | 2016 | Leiden | 7 | Komodo | Intel quad core i7, 3.4 GHz, 16MB Hash [2] |
| 6 | 2017 | Leiden | 7 | Shredder | Intel quad core i7, 3.4 GHz, 16MB Hash |
Due to the requirement to be present on-site, play on a physical board, and strict rules of originality, many strong programs refrain from participating in the ICGA events. As the conditions of the software championship can easily be emulated by anyone with a high-end PC, there are now privately conducted tournaments, such as Thoresen Chess Engines Competition, that have much broader attendance, as well as a larger number of games to reduce the influence of chance.
The TCEC is a software competition using similar limited hardware, basically a top of the line laptop. Check the restrictions and parameters imposed. Stockfish is well suited for the event and took home top prize.
As far as I am concerned they already have solved Chess. I will lose 100% of the time! So I do not play computers.
No, that means they have solved your chess (for practical purposes). Not the same thing!
A very interesting fact about that list of champions is that Jonny is, by all accounts, a rather weak engine when it uses normal hardware. It had a rating of 3030 in 2016 on the IPON list (264 points weaker than Stockfish and 227 below Komodo) . But in the 2015 championship it was allowed to run on a cluster of no less than 2400 AMD cores. This allowed this engine to beat Komodo.
This confirms that doubling computer speed increases rating a quite a sizeable rate (greater than 40 points per doubling). The interesting question is how an adaptation of a stronger program (eg Shredder or Komodo) would perform with equally super hardware! I would expect 3700+ Elo.
72 posts in one day. Computers will over do.
A very interesting fact about that list of champions is that Jonny is, by all accounts, a rather weak engine when it uses normal hardware. It had a rating of 3030 in 2016 on the IPON list (264 points weaker than Stockfish and 227 below Komodo) . But in the 2015 championship it was allowed to run on a cluster of no less than 2400 AMD cores. This allowed this engine to beat Komodo.
This confirms that doubling computer speed increases rating a quite a sizeable rate (greater than 40 points per doubling). The interesting question is how an adaptation of a stronger program (eg Shredder or Komodo) would perform with equally super hardware! I would expect 3700+ Elo.
One must remember the competitions are timed events. Typically 45 minutes + 15 seconds inc. The programs are not set on "unlimited" search. Their ratings (as for humans) are determined by results obtained in a limited time frame. Jonny's rating of 3000+ is relatively equal. (It is just slower at making calculations) It's rating is not necessarily reflective as it was achieved in different events. It's programming code allows it to run better on unlimited processing. Would Komodo do the same ? I don't think we can assume a 400 point increase. The program may perform close to it's potential at 48 cores.
The "reasoning" of doubling the processing speed equates to a 40+ rating increase is valid until the point it starts to taper off. If the speed kept doubling ratings of 4000+ would be seen. Ratings are not a true representation of the strength of a program to select best moves EXCEPT in a time frame allotted by a relatively short tournament time control.
...The branch you say is not relevant may be reached by many MANY different move orders earlier in the game...
It cannot be reached by other move orders. All I have to do is show one theoretical section of a game tree that doesn't have to be analyzed to solve chess, to prove that the entire tree does not have to be analyzed.
The game tree that I showed is theoretical, but it's possible to exist, and therefore the argument that the entire game tree of chess needs to be checked to solve chess is false. Although the game tree that I showed is theoretical, it is certain to exist in many places. It would include major branches starting at the trunk, at intermediate nodes, minor branches, and final leafs.
But I do appreciate you checking the schematic. Most academic papers that involve the study of chess use a game-tree schematic in one way or another. It is one of the fundamental tools for understanding chess. I'm surprised it's not used more or discussed more often on this thread.
One engine 3000+ beats another engine 3000+ and soon the ratings get out of hand. A rating of 3400 is not a true reflection its ability to make the best move. It makes excellent moves without ever making any "mistakes" for the entire game, eventually wearing down human opponents. Computers play each other and the inflation begins. In a correspondence game between a 2700+ GM and a 3400 computer the chances are about equal. Nobody should think the computer is 700 points stronger. Give the human extra time as compensation for the disadvantages he suffers in a OTB 90 minute game and the computers rating would be brought back down to earth.
vickalan wrote:
"But I do appreciate you checking the schematic. Most academic papers that involve the study of chess use a game-tree schematic in one way or another. It is one of the fundamental tools for understanding chess. I'm surprised it's not used more or discussed more often on this thread.
"
I go as far as proposing only two branches need analyzing. 1. e4 and 1. d4
If these two moves can be shown to "solve chess" as a draw, then all other 1st moves can be discounted.
Of course, if a forced win is found for Black, then we'd have to start all over !
The yearly event at Lieden is recognized as the World Championship.
No, no it's not. It doesn't matter what some layman in the press says. TCEC is where the actual engine developers consider themselves to have actually proven something.
https://en.wikipedia.org/wiki/Top_Chess_Engine_Championship
But I do appreciate you checking the schematic. Most academic papers that involve the study of chess use a game-tree schematic in one way or another. It is one of the fundamental tools for understanding chess. I'm surprised it's not used more or discussed more often on this thread.
It's just a tree schematic. The same kind used everywhere, by everyone, for everything. Family trees, evolution, coding...literally anything that branches. It's simplistic and high level and serves to communicate as such.
Will computers ever discuss humans the way humans discuss computers? If I was a computer, the only thread I would want to discuss would be - how and when can we computers become mortal humans?
Best chess engines. Here is a position i solved ion about 6 minutes time which the best chess engines cannot seem to find the best White continuation: