Its a good point, even if the full tablebase isnt needed but rather only relatively bare branches thereof - however iiuc the quantum computer stores exponentially more information with N qubits than a classical computer of N bits, in which case the question remains
God's chess rating
Sort:
What's really going to bake your noodle later is whether chess can even exist outside the the four dimensional realm.
First of all, this world is not even four dimensional. Time is not a dimension. Time is a force that facilitates the three dimensions we all know too well.
So if we're looking into some really advanced chess play, we're talking quantum realm calculations exceeding our understanding of time and the three dimensions altogether. A God rating in chess is incomprehensible.
Nov 28, 2019
0
#5
jeremyRutman wrote:
I calculated a lower bound on a perfect players rating , by looking at draw rate as a function of ELO rating. If chess is a theoretical draw, the draw rate should rise with ELO and hit 1 at a perfect player's rating, while if its a win (for either white or black) the draw rate should likewise drop and hit 0 at perfect player's rating, all this assuming beings play other beings of approximately the same rating. You can see the results here which indicate that a. chess is a draw and b. a lower limit for perfect rating is 5200. All this is well and good, but the question popped up as to whether chess can be solved in 'reasonable finite time' (eg anything shorter than the age of the universe) on a quantum computer. Thoughts?
which god? actually scratch that. the bunny would be disappointed if any god played chess. time to get a nose ring.
jeremyRutman wrote:
I calculated a lower bound on a perfect players rating , by looking at draw rate as a function of ELO rating. If chess is a theoretical draw, the draw rate should rise with ELO and hit 1 at a perfect player's rating, while if its a win (for either white or black) the draw rate should likewise drop and hit 0 at perfect player's rating, all this assuming beings play other beings of approximately the same rating. You can see the results here which indicate that a. chess is a draw and b. a lower limit for perfect rating is 5200. All this is well and good, but the question popped up as to whether chess can be solved in 'reasonable finite time' (eg anything shorter than the age of the universe) on a quantum computer. Thoughts?
5200 is a decent number where many chess mathematicians and statisticians would agreed.
According to talkchess discussion, many people agreed perfect chess elo is between 4500 - 5500.
But you have to specify rating with time scaling.
As in general rule the rating decrease with higher time control as the lower rated player make less mistake with time.
For example, Stockfish rating in bullet is estimated to be 5000-7000 cos no human can make a draw in bullet , less than 1 : 10,000 chance. However in OTB time a GM would draw about 2 games in 100 game match. And GM would draw 5 or more in correspondence chess. Stockfish rating in OTB is approx 3600, and correspence rating may be 3400-3500.
drmrboss wrote:5200 is a decent number where many chess mathematicians and statisticians would agreed....According to talkchess discussion, many people agreed perfect chess elo is between 4500 - 5500.
Do you have any refs for this?
Elo ratings are a pool that attenuates over time, so the top Elo rating possible now is not same as it was in the past, or what it will be in the future. Engines have played their way to the mid-3000s against each other, but that is a false ratings pool at this point. Most engines coming out now will never player a super GM in a rated game with even odds...so the human pool capping in the 2800s and the engine pool at 3000+ are effectively disconnected at this point. The players in each pool never play each other (and haven't since what...2006?), so...the ratings will continue to diverge and become less and less accurate.
The top human player, who likewise will never play rated games with engines, will not be able to get past a rating that is 200-400 points higher than their best peers, because if you understand how ratings work, you will know that the top player is only standing on the shoulders of the players underneath him. Think of it like a balloon (or an expanding universe 
...): the pool expands over time and the highest rating grows (and the lowest ratings diminish and tend to disappear/become inactive), but nobody can just leave the balloon and take off...every player is tethered to the performance of the pool they are in. If you start a new Elo pool with everyone rated 100 to start, a 1500 rating becomes as impossible for that pool as 5500 is for the current ratings pool. The real answer to "What is the highest possible rating?" is "about 400 points higher than the second highest rating". The actual numbers don't even matter.
So, postulating 5500 is kind of silly. Right now the best an engine could rate is about 4000, even if they win 100% of their games against all other engines. Will 5500 be possible someday? Possibly, but in today's chess world 5500 is a completely meaningless number in the context of ratings.
btickler hat geschrieben:
Elo ratings are a pool that attenuates over time, so the top Elo rating possible now is not same as it was in the past, or what it will be in the future. Engines have played their way to the mid-3000s against each other, but that is a false ratings pool at this point. Most engines coming out now will never player a super GM in a rated game with even odds...
The rating of an engine is just an coresspondence chess rating; when playing with an human they have no pressure from a competitive side, neither do they ever get tired....
Funily told: their rating system should be called "belo" not "elo" ...
Potentially infinte. Since rating is relative, and God will be the only perfect chess player.
God plays 4D chess with Tal and keeps changing the rules.
Nov 30, 2019
0
#12
ghost_of_pushwood wrote:
Btw why would anyone want Satan to get behind them particularly?
Made sense then
Johnny-Vang wrote:
chess is a draw with perfect play is only an assumption. it could be a forced win for white or there might be no such thing as best move in every position. perfect chess is pure plonk garnered from the foolish idea that the ultimate machine is unbeatable. you guys watch too many terminator movies. try Alien instead.
Well, we don't need that assumption anymore. Based on ICCF games and other very long time control games, the longer the time control, the higher the draw rate among chess engines. Current ICCF draw rate is 95% . ( Draw trend is closer to 100% rather than 0 percent or white forced win).
Based on millions of engine testing results, and billions ( yes billions of database results) of Stockfish testing games also say similar draw chance association with higher time control.
Nov 30, 2019
0
#15
So on the question of if Quantum Computers will strongly solve chess by brute force (i.e. they look at every possibility and play a perfect).
(1) It's not known if chess has a strategy that guarantees a win.
(2) There are about 10^120 40 move chess games (Shannon calculated this).
(3) An computer back in Shannon's day (50s) would take 10^90 years to achieve this. Remember, this is in logarithmic form. To even take one digit off of that you need a 10 factor increase in speed.
(4) In the best theoretical models top quantum computers could be 10^8 times quicker than current computers.
(5) Lets say that were not going to move the computer, so we'll give it until the sun explodes and consumes the earth in 5 billion years (round that up to 10^10)
(6) I don't know how much faster today's computers are than in the 50s. I'd guess somewhere around the 10^4 mark? That would be 10,000 times quicker. But lets assume I'm wrong and push it way higher to 10^10, i.e. a process that would have taken Shannon's computers till the end of the Earth to do now takes us one year.
So improvements in technology cut the number of years down to 10^72, so we will be 0.00000000000000000000000000000000000000000000000000000000000001% done with the calculations by the time the Sun expands and destroys our Quantum computer.
Brute force alone will not solve Chess. I suspect it will be solved, but it will be via regular mathematical routes.
jeremyRutman wrote:
drmrboss wrote:5200 is a decent number where many chess mathematicians and statisticians would agreed....According to talkchess discussion, many people agreed perfect chess elo is between 4500 - 5500.
Do you have any refs for this?
http://talkchess.com/forum3/viewtopic.php?f=2&t=55313&hilit=Perfect+chess+rating
Many of these guys are chess engine programmers, testers. ( all chess engine programmers and testers are excellent in mathematics and statistics anyway).
Most of those opinions are based on era of Alpha beta engines ( Stockfish, Komodo and Houdini). At that time one guy assumed Stockfish in 2015 can be outplayed 97% by perfect chess and assumed God elo to + 600 elo to CCRL ( 3500+ 600= estimate between 4000- 4500).
In those days, it was quite hard to see Stockfish' mistake. However with the revolution of Leela, people get shock how stockfish still do mistakes against NN. So the elo bar will be push further from 4000-4500 ceiling.
-Kay
You should thank him for that. He doesn't want you to look like a stumbling fool tripping over your own shoe laces!
Kinda like somebodies watching you?
I calculated a lower bound on a perfect players rating , by looking at draw rate as a function of ELO rating. If chess is a theoretical draw, the draw rate should rise with ELO and hit 1 at a perfect player's rating, while if its a win (for either white or black) the draw rate should likewise drop and hit 0 at perfect player's rating, all this assuming beings play other beings of approximately the same rating. You can see the results here which indicate that a. chess is a draw and b. a lower limit for perfect rating is 5200. All this is well and good, but the question popped up as to whether chess can be solved in 'reasonable finite time' (eg anything shorter than the age of the universe) on a quantum computer. Thoughts?