12892 Players currently online!
Man vs. Machine - good luck!
Turn-based games at any time!
Vote for the best move to win!
Do you have what it takes?
Sharpen your tactical vision!
Get advice and game insights!
Learn from top players & pros!
View millions of master games!
Your virtual chess coach!
Perfect your opening moves!
Test your skills vs. computer!
Find the right private coach!
Can you solve it each day?
Bring it all together!
Beginners, start here!
Make friends & play team games!
News from the world of chess!
Search all Chess.com members!
Find local clubs & events!
Who's the best of your friends?
Read what members are saying!
Also, there is a significant flaw with the Elo system in that it estimates all percentages by rating differential.
What I mean is, a 2700 is expected to score 64% against someone 100 points below him, and 76% against someone 200 points below him. But a problem in this prediction starts if he for example scores 80% against someone 200 points below him and only 60% against someone 100 points below him, these percentages would both be wrong and he would still come out at about the same rating.
For this reason I believe the margin of error for a 2700 vs a 1300 could very well be great enough that this mathematical estimate of 3/10000 could easily become 0.
"But a problem in this prediction starts if he for example scores 80% against someone 200 points below him and only 60% against someone 100 points below him, these percentages would both be wrong and he would still come out at about the same rating."
It's not really a problem. If a person can get those kinds of results that seems to be a perfectly good 2700 to me, if a more streaky one. I mean sure, not all 2700s or 1300s are equal in that there are styles they can have that can be good against some of equal rating and not others of equal rating but a different style. This doesn't change the overall reliability of ratings.
And there is no reason to me why simply the big rating difference between the 2700 and 1300 means a greater chance or magnitude of anomaly -- I see no need to make that logical jump. If the 2700 has a certain kind of style against that 1300 he will perform as he would against the average 1400 instead for that particular player, maybe -- not such a huge deal.
I mean, this problem of style and others will come up in any rating comparison -- it may be that player A, 1400, and player B, 1500, have perfectly legit ratings, yet the rating system is totally wrong when player A beats the 1500 9 out of 10 in a 10 game match, perhaps because player A's style makes player B uneasy. This is of course an extremely unlikely scenario, but my point is the 2700 vs. 1300 is not inherently subject to problems any more than any rating comparison will be.
General chess knowledge is the main variable, and style is another variable in chess performance depending on how two styles clash -- this fact need not multiply when the rating difference increases. Remember the 1300 and 2700 got their ratings established over their long run performance -- they didn't just play some random guy who had an easy to beat style and sap points from him.
One more post before I head out this evening. Quantum mechanics give me a very low chance of walking through the door. It's not going to happen even though my chances are non-zero.
My point is that whatever equation is used is not completely accurate in predicting extreme cases. They are very accurate within a neighborhood. The only thing I am objecting to is the exact decimal answer that was posted previously.
I do think that over long term, in experiment, a 2700 player would get the same result vs. a 1300 and a 1400. He would win them all.
Likewise, if I play a true 800 or 900 player (not an under rated kid), I would expect to win every single game. 100,000/100,000. I'm sure this would be true. See you guys in a few days!
How much margin of error do you think there is? Are you talking 3 in 11000 or more like 3 in 40000 or something even more drastically different from 3 in 10000? Personally I don't think that, proportionally, the margin of error with the 2700 vs 1300 match-up would be so different from a match-up like 2000 vs 1900.
Guys dont get crazy! The question is just: Can a 1300 rated Player beat a 2700 Gm and the simple answer is .... NO! Magnus Carlsen played once a game against a 2000 rated Player without a Queen and he was nearly winning this. its just hopeless for the 1300.
I searched chess tempo's database, and out of 94 games only 4 were lost with 2700+ players vs 2000 and below players.
Two of these games were simuls, and two of them were wins by David Christopherson over Peter and Preben Nielson at the same tournament. While chess tempo lists Preben as having a 2700 rating, chessgames.com states that his highest rating in their database was in the 1900s and fide lists him as being a 1793 with no titles. Perhaps there was some confusion in chesstempo's database.
There were also 8 draws, of which only 2 were not simuls, one of which included Peter Nielson at the same tournament. Perhaps there was some confusion, because I just can't see him having that bad of tournament.
There were also only 28 losses out of 540 for players ranging in the low 2000s against players under 1500 in the database.
So now, we can multiply these two to get the odds of a 2700 losing to a 1500, right? No. Multiplying incorrect information is senseless.
It's interesting that if you find a win by someone rated under 2000 against a 2700, chances are better that the information is incorrect than the result actually happened!
Peter Heine Nielsen was exactly 2700 for one rating period, May of 2010. There has never been anyone named "Nielson" who was +2700. Clearly, whoever entered the data changed 17xx to 27xx.
Also, we aren't counting simuls. We aren't counting games almost played against players rated 2600 (someone mentioned nearly getting to play in the same tournament as Jon Ludwig Hammer, a great player whose highest rating is 2647).
How many serious tournament games (not classes) have been played between someone rated 13xx and a player rated 2700 FIDE or better at the the time of the game in the entire database?
Maybe the 1300 can win whith the rule we used at the last party i was: Each move you have to drink a short-drink.
Better dont do this vs a Russian GM though :)
Shepi and SmyslovFan:
Chessbase Megadatabase 2012 makes the same mistake, they confuse:
Peter Nielsen: A 1749 player
Peter Heine Nielsen: A GM with a 2700 peak rating
Ten thousand is a larger number than people think. If you do ten thousand of something condensed into a relatively short period of time (e.g., a month) you'll feel it like it's become a part of you.
Assuming you are awake 16 hours a day, and the month is 31 days long that means doing that thing every three minutes for every waking moment of the month in question.
I think a more reasonable question is what kind of advantage would a 1300 player have to be given to be able to beat a 2700? for example, having the 2700 player play without a queen or rook? without a queen and both rooks?
When I was 14 I was playing solastic chess. I had a rating of perhaps 1200 at the time, perhaps even 1000. At the club we would play unrated games swiss style and after beating two of the other solastics, I played the club top player, who was 2075-2100. I had lost to him most likely 100 times, I played a 4-knights that stayed closed, and he got in time trouble. He hung a knight when he didn't see that a pawn recapture opened up my bishop, and then lost 2 more pawns that the knight was defending. And in the endgame he still almost beat me down a knight and two pawns. But I was able to beat him. And In Corr. play, I hung a queen out to a player who was rated 800 points below me. I hope he enjoyed those rating points. It can happen, but it is rare.
Perhaps to refine this question further: how much would he need to win consistently against the 2700? I think a 1300 has decent chances to win a queen up for example (depending on how much the position is simplified too), but the 2700 coming back is also a definite possibility.
Well, I think generally being up both a queen and a rook should lead to a win for the 1300 a very large amount of the time :)
yes, there is a chance
I did point out that there was probably some error due to some suspicious data, but didn't have any evidence so I didnt throw out the games.
This is the best answer so far, chess is so complex anything can happen, sometimes when I play against lo rated players the style of play can be so chaotic that one can miss something or get caught out by ones own pieces. I remember being 12 pieces up and nearly losing having trapped myself in.
Making the Chess.com Forums Better
by pelly13 a few minutes ago
The candidates tournament is here!
by KnobSocket a few minutes ago
3/10/2014 - A Quick End
by pinkpawnther a few minutes ago
by chessmicky 3 minutes ago
Daily Puzzle 3/10/14-Helpmate - #8/10
by GSHAPIROY 6 minutes ago
abuse of the holiday button
by TheGrobe 6 minutes ago
What is your favorite chess quote?
by JariIkonen 8 minutes ago
Where are the Ameriacn made chess sets?
by chessmkr1 9 minutes ago
Live Chess on Android Tablet
by dylandog7 14 minutes ago
Chess.com handicap for only one player
by Hinanawi-Tenshi 15 minutes ago
Why Join | Chess Topics |
Help & Support |
© 2014 Chess.com
• Chess - English
We are working hard to make Chess.com available in over 70 languages. Check back over the year as we develop the technology to add more, and we will try our best to notify you when your language is ready for translating!