Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / When will computers 2851 + 800 (3651) Elo?
- - By BankShots (***) Date 2011-06-13 14:37 Edited 2011-06-13 16:20
I just thought this would be an interesting question--as of course +800 elo corresponds to/= going 100% in head-to-head competition and that 2851 elo = highest recorded human elo.
:eek::evil::grin:
Parent - By BankShots (***) Date 2011-06-13 19:11
So ie. when computers would be considered to be going 100% on highest recorded human(-level) in head-to-head competition*. :twisted::lol::roll:

*  Also, I guess from CCRL having had a rating adjustment [from subtracting 100-200 elo], this would come 100-200 elo earlier than reported on CCRL--at approximately 3551-3451 ELO.
Parent - - By BankShots (***) Date 2011-06-16 20:38

>*  Also, I guess from CCRL having had a rating adjustment [from subtracting 100-200 elo], this would come 100-200 elo earlier than reported on CCRL--at approximately 3551-3451 ELO.


Has the Rybka Cluster achieved this?
Parent - - By NATIONAL12 (Gold) Date 2011-06-16 22:39 Edited 2011-06-16 22:42
none of us know strength of cluster as it will not play corr games. i dont blame Lukas or Vas for this.saying that a friend of mine Werewolf hired cluster and gave some estimates of strength against an overclocked 12 core(R4 at the time not R4.1),doubt it would hold up on deep AN or corr chess.however this is my OPO.

btw,despite what most think if you can get to similar depths to Houdini in corr games,Rybka is stronger.:smile:
Parent - - By BankShots (***) Date 2011-06-17 00:23

> none of us know strength of cluster


Well......I guess it's only _History_ that we are missing. :fat:

> btw,despite what most think if you can get to similar depths to Houdini in corr games,Rybka is stronger.


Yes, thank-you. :smile:  Houdini.....what a shark! :lol:
Parent - - By Geomusic (*****) Date 2011-06-25 07:36 Edited 2011-06-25 07:54
As a function of the ceilings going up due to non-recentering in the rating lists. The top rated human will probably be rated 3050 in 50 years. Just look at how many 2700+ there are today vs 30 years ago on the FIDE lists.
The assumption being +50 elo gain on the top end every decade. Computers have gone through something similar with its rating lists and yes eventually they too will be rated 3800+ and even 4000+ if nobody recenters them. You can see this to an extreme level on certain chess servers as an example whose rating lists go into the high 3600+s for both (C) and humans alike. A joke for sure, putting it mildly. It's hard to know what anyone is truly rated anymore. If things keep going as they are in 100 years online average will be like 2400+ and a GM will be rated 5000+ :) LOL
Parent - - By Banned for Life (Gold) Date 2011-06-25 12:13
Uly will still be rated 1400 though... :smile:
Parent - By Geomusic (*****) Date 2011-06-25 19:34
"Experts agree" LOL
Parent - - By Labyrinth (*****) Date 2011-06-25 14:13

>The assumption being +50 elo gain on the top end every decade.


Karpov was at 2700 in 1980.

Kasparov crossed 2800 in 1990.

Kasparov was ~2850 in 2000.

There was no climb from 2850 to 2900 in 2000 to 2010. In fact most stayed at or under 2800.
Parent - - By Geomusic (*****) Date 2011-06-25 18:54 Edited 2011-06-25 20:09
"1970 FIDE rating list. Top 7 players
  
    1 Fischer Robert James..............   USA  2720
  2-3 Spassky, Boris....................   URS  2670
      Kortchnoi, Viktor.................   URS  2670
  4-5 Geller, Efim......................   URS  2660
      Keres, Paul.......................   URS  2660
  6-7 Larsen, Bent......................   DEN  2650
      Petrosian, Tigran.................   URS  2650

1980

1 Karpov, Anatoly...................   URS  2725
    2 Tal, Mikhail......................   URS  2705
    3 Kortchnoi, Viktor.................   SWZ  2695
    4 Portisch, Lajos...................   HUN  2655
    5 Polugaevsky, Lev..................   URS  2635
  6-8 Spassky, Boris....................   URS  2615
      Petrosian, Tigran.................   URS  2615
      Mecking, Enrique..................   BRA  2615

1990

January 1990 FIDE rating list. Top 11 players

    1 Kasparov, Gary....................   URS  2800 11
    2 Karpov, Anatoly...................   URS  2730 30
    3 Timman, Jan.......................   NLD  2680 23
    4 Ivanchuk, Vassily.................   URS  2665 21
  5-6 Gurevich, Mikhail.................   URS  2645 38
      Salov, Valery.....................   URS  2645 17
    7 Beliavsky, Alexander..............   URS  2640 32
    8 Short, Nigel......................   ENG  2635 36
    9 Andersson, Ulf....................   SVE  2630 39
10-11 Kortchnoi, Viktor.................   SWZ  2625 37
      Ljubojevic, Ljubomir..............   JUG  2625 11

January 2000 - FIDE Rating List
1 . Kasparov, Garry : 2851
2 . Anand, Viswanathan : 2769
3 . Kramnik, Vladimir : 2758
4 . Shirov, Alexei : 2751
5 . Morozevich, Alexander : 2748
6 . Leko, Peter : 2725
7 . Kamsky, Gata : 2717
8 . Adams, Michael : 2715
9 . Bareev, Evgeny : 2709
10 . Ivanchuk, Vassily : 2709
11 . Topalov, Veselin : 2702"

In January of 2010
there were 36, 2700+ players and 3 2800+ players
--------------------------------------------------
Something odd happened between 2000-2010 There was a 2700+ explosion.
Parent - - By Leto (***) Date 2011-07-10 14:17
The important thing is only 3 players are 2800+ at the moment.  So I don't see where you are getting evidence that the top player will be 3050 within 50 years.
Parent - - By NATIONAL12 (Gold) Date 2011-07-11 23:35
only by inflated ELO's,which has been happening for some years now.
Parent - By turbojuice1122 (Gold) Date 2011-07-12 19:23
I have long believed that the "inflation" is due to various changes in society.  For example, in the early 1990's, you're dealing with players who started young in the Fischer era.  The brightest people out there were significantly more likely to become interested in chess because of these events (though average people, not necessarily so, since chess ideas might not interest them much in the first place, thus arguing against those who note a somewhat higher percentage of players rated over such and such in the mid-1980's).  In the late 1990's, you're having the fruition of the information age, where young players have lots and lots of games at their fingertips due to the Internet, giving far more study opportunities.  Similarly, during this time, chess programs were becoming fairly strong, and so 6-year-old Junior doesn't constantly have to wait for when his dad feels like playing a chess game--he always has a strong opponent against whom to practice.  This was also the time when the Internet Chess Club had been around for awhile.

However, I don't really notice major events occurring now that would cause a sudden increase in the ability of players to improve or an impetus that would encourage really smart kids out there to take up chess as their main hobby.  Thus, I really don't anticipate anyone getting much over 2900 during my lifetime.  Breaking 2900 wouldn't surprise me, but breaking 2950 would.
Parent - - By Geomusic (*****) Date 2011-06-25 18:59
You can't refute an argument on 1 case Lab.
Parent - By yanquis1972 (****) Date 2011-07-12 19:08
especially when that 'case' is arguably the greatest chess player of all time.
Parent - - By Vempele (Silver) Date 2011-06-25 15:05

> If things keep going as they are in 100 years online average will be like 2400+ and a GM will be rated 5000+ :) LOL


ICC's average is actually still ~1600, and the rating system used by FICS actually makes the average rating go down.

Whoa, look at the Chess960 list! Only two IMs have ever been over 2200, the other 15 are GMs! http://staff.cs.utu.fi/~juhkivij/chess/icc_best_en.shtml
Parent - By Geomusic (*****) Date 2011-06-25 19:29 Edited 2011-06-25 19:32
Yes Vemp, I don't have a problem with the FICS rating system. And the average was centered basically on the top 50 chessplayers not 50,000 or however many FIDE members there are in the world. I was being silly when I said 2400+ to the avg player. It effects those at the edges of the distribution not the middle.
Parent - By Geomusic (*****) Date 2011-06-27 21:11
Give it 5-10 years the top on the 960 list will be 2600+
Parent - - By Gr00vy (**) Date 2011-06-25 16:35
There may be an argument in here somewhere, but there are actual numbers and claims being made that aren't even internally supported.  Get those right first, and then we can talk about whether the ratings should be "recentered ", whether or not there are simply more elite players with more elite knowledge, whether or not outliers do extend further out on the rating chain, etc.

But lets start with "The assumption being of a +50 elo gain on the top end every decade). we are at 2800 and change, and then "If things keep going the way they are..." in 100 years a GM will be rated 5000+. 

Now most chess players are good at math.  And we are going to reject nearly your entire argument, because your math is not even close.
Parent - - By Geomusic (*****) Date 2011-06-25 18:58 Edited 2011-06-25 19:26
Of course there is an argument in there the facts are in front of you. 1 or 2, 2700 players were in the FIDE list from 1970-1990 suddenly 7 appear in 2000 and 30+ appear in the 2010 list. I always check my facts, read that Chessbase article at the bottom. My 50 elo per decade isn't so far off. And my quote on 5000+ rated GMs was a joke though partially serious in that online ratings have been going stratospheric in the past 10-15 years. From 2700+ to 3700+ It doesn't take a stretch of the imagination to see how GMs could be rated 5000+ online in 100 years hell it might even be sooner. So I would appreciate it if you guys check your facts first before you disqualify mine off the cuff.

http://www.chessbase.com/newsdetail.asp?newsid=5608
Parent - By dcorbit (***) Date 2011-06-27 18:41
When we are talking about extrapolations into the future, I think it is not possible to know for sure in either direction.
Things that seemed ridiculous 25 years ago do not seem so ridiculous now.
There is something about extrapolation in that as soon as you go beyond the bounds of the experimental data, the error bars go exponential.

So, a prediction about a long time from now really has not a lot of meaning other than it is interesting to talk about.

Ray Kurzweil has an interesting page on exponential trends:
http://www.kurzweilai.net/the-law-of-accelerating-returns
Parent - - By Gr00vy (**) Date 2011-07-10 13:52
what is 10 x 50?
Parent - - By BankShots (***) Date 2011-07-10 14:34
Noone seemed to really get this post! [:mad:]

The whole point is that the Rating Lists are off by 100 or 200 ELO points (my source could not remember which).  I was told they are UNDERstating the strengths of their engines by this amount from old standardized FIDE lists.  AND if the ClusterRybka is +200 ELO points on the Rating List stated amounts, then we are talking about ClusterRybka being more like an ACTUAL +300 or +400 ELO on the FIDE (Human!) lists.  Whereas normal DeepRybka is listed as 3262 ELO (CCRL @ time of post), this places their FIDE equivalent rating at 3562 or 3662 ELO.  So it comes down to how far off the rating lists are from their FIDE equivalent.  If they are only 100 points off, then we are not quite there yet.  Soooon right.  However, if the rating lists are the 200 ELO points understated, THEN we have already! reached some History in Chess!!!!!  There is a theoretical number which has then been exceeded--800 ELO + Human World Record ELO which I think is arguably MORE IMPORTANT than that >>a computer beat a human world champion in a match!!<<.  This then is the theoretical point where computers are TRULY DOMINATING!! the Human World Champion at the game of Chess at their highest level over the course of Many games.  Where the Human World Champion might be only winning a match point < 1% of the time!

It should be on the news as BIG as the LOSS to DeepBlue.  It is what everyone meant, and yet WE DON'T EVEN KNOW!!!!???? :cry:

Too trapped in our caves watching dumb shadows! :red:
Parent - - By chesstango (**) Date 2011-07-10 15:19
I guess that when anybody finds a chess game  algorythm or something close to it, first: personally for me it would be sad, because chess would probably dissapear..... second, I am insisting since 15 years ago, that ELo strength will rise only by software work, more than improving processors (hard).....
Thx in advance.
Parent - - By TheHug (Bronze) Date 2011-07-10 15:25
Well you have some good points in your statement and I do agree for the most part. But computer clusters do have a big effect, the horizon effect they have is very impressive. And software and hardware can go hand in hand as a programmer can make changes to the software to have better search algorithms.
Parent - - By chesstango (**) Date 2011-07-10 15:32
Dear friend: despite i am not a programmer nor   a systems specialist, i insist that if an algorythm is find out, then u could apply it even in an old DOS...... sucessfully!!!!
Maybe i am wrong; i have a hunch: its a matter of algorythm.....
Parent - - By TheHug (Bronze) Date 2011-07-10 15:36
I see your point, the one thing I may point out is that you would like to see the moves in a more timely fashion ply wise. But you made a good point.
Parent - By chesstango (**) Date 2011-07-10 15:38
thx a lot!!!
Parent - - By chessmaster15 (**) Date 2011-07-10 16:10
i agree that rating in computer chess are way off . one thing i find funny is an old chess computer was rated say 2018 when it came out and years later its rated at 1500 or 1600 . im sure the computer plays just as good now as it played back then hehehehehehehe so to me thats very very funny .
Parent - By Leto (***) Date 2011-07-11 21:58
Probably a flawed testing method led to the claimed 2018 rating.  Someone probably complained and the company retested it and lowered the rating to 1500-1600.
Parent - - By chessmaster15 (**) Date 2011-07-10 16:06
i believe 1 day chess will be solved completely . i dont know what elo that translates into . but in 10 years time chess will be solved e2-e4 mate in 230 . then the human plays c7-c5 then the computer plays g1-f3 mate in 172 indicating a mistake was made and now instead of a loss in 229 more moves you will lose in 172 . thats where it heading . even though chess is a deep game there are only so many ways to play that dont lead to a decisive loss vs correct play . since there is already egtb thats taken care of already . so only thing left is the opening and part of the middle game .
Parent - - By Moz (****) Date 2011-07-11 22:18

> in 10 years time chess will be solved


Dream on. Chess will not be solved for many, many decades. The numbers are too big.
Parent - By NATIONAL12 (Gold) Date 2011-07-11 23:37
i agree,i will be long dead and buried before chess is solved.
Parent - - By chessmaster15 (**) Date 2011-07-11 23:52
think of it this way ,,, lets say you want to go to the store thats 2 blocks away . ok how many diff ways are there to go to that store (alot) but how many make sense (only a few) . i mean you could walk backwards all the way to the store true enough. you can walk around the world and you will run into it . but there are alot less ways to get there that make sense . chess is exactly like that . you can start the game with h2-h4 follow by a2-a4 but vs correct play you will get stomped . so chess does not have that large a number of correct moves . now if your playing a 6 years old kid who just learned the game im sure the number of ways to play seem infinite . but if you set down and play rybka you will find out rather quickly that alot of those moves dont work . so if you follow my line of thinking it wont be that long because most those moves are rediculas .
Parent - - By Moz (****) Date 2011-07-12 04:22 Edited 2011-07-12 04:25

> so if you follow my line of thinking it wont be that long because most those moves are rediculas .


It doesn't work that way. Until chess is solved, we'll never know for sure which moves are ridiculous. After several hundred years of human praxis we can surmise that 1.h4 isn't a winning move but until chess is solved we can't be certain. Until you work out the end result for every conceivable move you'll never know one way or the other and the numbers are simply too big to solve any time in the near future.

Engines will never solve chess because pruning will never yield a solution.
Parent - By chessmaster15 (**) Date 2011-07-12 13:09
it does work that way , it really does . if you want to do an experiment then try this . create an opening book have 2 engines very close in elo 1 or 2 points from each other . give 1 engine a opening book with h2-h4 follow by a2-a4 and see what happens . on a more practical chess like in real games . here is the way it works for me if i get a advantage out of the opening(i considder the opening most important) ill win that game even if it is rybka im playing . but if im playing slightly lesser player and i get a really bad opening ill work very very hard just to save the game (draw ?). so with that in mind chess itself dictates that you dont waste time in the openings .i guess understanding  CHESS not just scientific numbers like there are  1,000,000,000,000,000,000,000,000,000,000,000,000 ways to do something is why i stated chess will be solved . true the numbers of ways you could play chess are huge , but the number of ways you can safely play vs a 3200 elo player are not huge narrow i would say . because any move that waste time will be eliminated very fast . if you make a move in chess you better have a good reason . not just g1-f3 f3-g1 g1-f3 f3g1 . while your doing that a good strong player is developeing his pieces or positioning them to destroy you . endings are solved already (egtb)true or false ? also i get up 1 knight vs anyone in the world computer or otherwise im win that game . so that narrows chess down even more because it means if your opening causes you to lose up to a knights worth of material without considerable compensation in position you lose .that eliminates lots and lots of variations over 15 moves (opening) .i guess to make my little store example more accurate i should say there is only a few ways you can get to the store without getting mugged . lots of ways to try to get there but only a few will give you a chance to actually get there . as you notice when you play engines at a fast rate like bullet 1 min game you get lots wins loses but if you slow the game down to 1 hour per move you get alot more draws and here is why . the longer the program thinks the better move it will play tactical mistakes are cut down alot . those wastefull moves alot get eliminated . without mistakes i predict the end result of chess = a draw . its like a gigantic game of tic tac toe . the place where chess can be solved is the opening . the opening is the soul of chess , thats where to look first . because if you figure out how to gain a advantage in the opening the game is solved .
Up Topic Rybka Support & Discussion / Rybka Discussion / When will computers 2851 + 800 (3651) Elo?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill