So much for Elo ratings

Sort:
Avatar of Vease

I wonder if Carlsen fails to win the Candidates Tournament if FIDE will finally look at overhauling the rating system? If a man with the highest rating in history, 60 points clear of his nearest rival cannot win a qualifying tourney for a World Championship match then surely they have to admit that the numbers have gone crazy.

The best recent example of how using numbers to predict chess results is flawed  is actually only a few days old on the Chessbase website. Jeff Sonas (Chessmetrics ratings inventor) 'proved' mathematically that Carlsen had an 87 percent chance to win the Candidates after round 11. After round 12, Kramnik now has a 65 percent chance of winning...ONE GAME LATER!

Chess is not a science, people play well or badly for reasons that have nothing to do with their innate ability so assigning 'hard' numbers to their results is pointless.

Avatar of Scottrf

Why should the #1 player be guaranteed to challenge?

Do the #1 and #2 national teams always play in the world cup final? Do the two best club sides always play in the FA cup final?

Avatar of Fevly_P

he deserves for his rating , and his competitors deserve for their own win too, if carlsen failed , it just nothing..

Avatar of Vease

Exactly,but in Chess the ratings are used to determine invitations to the most elite tournaments ('We need a 2740 player to get a 21 category - lets invite Wang Hao') and who is worthy to challenge for the world title. With something as 'scientific' as chess how can the number one rated player NOT be the world champion?

The ratings are flawed anyway because one man's 2700 has been achieved by playing a whole bunch of different players to another 2700 player. The only way the ratings could be accurate is if every rated player had exactly the same number of games against every other rated player.

Avatar of Scottrf
Vease wrote:

With something as 'scientific' as chess how can the number one rated player NOT be the world champion?

Seriously?

Avatar of -waller-
Vease wrote:

The best recent example of how using numbers to predict chess results is flawed  is actually only a few days old on the Chessbase website. Jeff Sonas (Chessmetrics ratings inventor) 'proved' mathematically that Carlsen had an 87 percent chance to win the Candidates after round 11. After round 12, Kramnik now has a 65 percent chance of winning...ONE GAME LATER!.

Why does this mean its flawed? It said Carlsen had an 87% chance to win, not a 100% chance to win. That means he's expected not to win 1 time in 8, which isn't a negligible proportion of times!

Avatar of Ubik42
Vease wrote:

I wonder if Carlsen fails to win the Candidates Tournament if FIDE will finally look at overhauling the rating system? If a man with the highest rating in history, 60 points clear of his nearest rival cannot win a qualifying tourney for a World Championship match then surely they have to admit that the numbers have gone crazy.

The best recent example of how using numbers to predict chess results is flawed  is actually only a few days old on the Chessbase website. Jeff Sonas (Chessmetrics ratings inventor) 'proved' mathematically that Carlsen had an 87 percent chance to win the Candidates after round 11. After round 12, Kramnik now has a 65 percent chance of winning...ONE GAME LATER!

Chess is not a science, people play well or badly for reasons that have nothing to do with their innate ability so assigning 'hard' numbers to their results is pointless.

The chess rating system is about as perfect as you are going to get, its not flawed. 

If the weatherman tells you there is a 80% chance of rain today, and it doesn't rain, was the weatherman proven wrong? No. That simply implied there was a 20% chance it wouldn't rain.

You want to see if he is right, you have to look at multiple days with predictions. The last 100 times he made the 80% call, it should have rained in about 80 of those days. This will give you a clearer picture.

There is nothing at all problematic about it being Carlsen 87% one game, then a game later it is now Kramnik 65%. 

If you doubt the ratings are the best way top predict, go enter a tournament with these guys. How many games you think you will win? Why?

Avatar of Sunofthemorninglight
Vease wrote:

 With something as 'scientific' as chess how can the number one rated player NOT be the world champion?

The ratings are flawed anyway because one man's 2700 has been achieved by playing a whole bunch of different players to another 2700 player. The only way the ratings could be accurate is if every rated player had exactly the same number of games against every other rated player.

statistically speaking, the more games that are played, the more clearly we can see who is the best tournament player over a year.

to see who is the best match player, a lot of games in matches can determine that.

for small numbers of games, final standings are subject to more fluctuation and inconsistent outcomes more common.

it's just how the stats work.

Avatar of VLaurenT

The ratings are not flawed : they describe statistically what can happen. The abrupt change from one round to another comes from the fact that both Kramnik winning and Carlsen losing was a highly unprobable event.

Avatar of Vease
hicetnunc wrote:

The ratings are not flawed : they describe statistically what can happen. The abrupt change from one round to another comes from the fact that both Kramnik winning and Carlsen losing was a highly unprobable event.

Only 'improbable' if you take the ratings at face value, Ivanchuk is one of the greatest players in history, it was not 'improbable' that he could defeat Carlsen.

Avatar of Xilmi

You also are aware that the ratings are not set in stone and if Carlsen loses more than his rating would imply, it infact is corrected downwards to reflect that.

Avatar of CapAnson

The numbers haven't gone crazy.. people are just giving too much weight to them.  Every leading grandmaster today would say Carlsen is by far the strongest player active today.. now whether he's the strrongest by 6 million  points or 40 or  just 1 point is just a feature of whatever rating system. But one bad game doesn't invalidade his rating.

Avatar of VLaurenT
Vease wrote:
hicetnunc wrote:

The ratings are not flawed : they describe statistically what can happen. The abrupt change from one round to another comes from the fact that both Kramnik winning and Carlsen losing was a highly unprobable event.

Only 'improbable' if you take the ratings at face value, Ivanchuk is one of the greatest players in history, it was not 'improbable' that he could defeat Carlsen.

improbable, yes, and not highly I agree Smile Thanks for the correction.

Avatar of TetsuoShima
Scottrf wrote:
Vease wrote:

With something as 'scientific' as chess how can the number one rated player NOT be the world champion?

Seriously?

lol

Avatar of TetsuoShima
Ubik42 wrote:
Vease wrote:

I wonder if Carlsen fails to win the Candidates Tournament if FIDE will finally look at overhauling the rating system? If a man with the highest rating in history, 60 points clear of his nearest rival cannot win a qualifying tourney for a World Championship match then surely they have to admit that the numbers have gone crazy.

The best recent example of how using numbers to predict chess results is flawed  is actually only a few days old on the Chessbase website. Jeff Sonas (Chessmetrics ratings inventor) 'proved' mathematically that Carlsen had an 87 percent chance to win the Candidates after round 11. After round 12, Kramnik now has a 65 percent chance of winning...ONE GAME LATER!

Chess is not a science, people play well or badly for reasons that have nothing to do with their innate ability so assigning 'hard' numbers to their results is pointless.

The chess rating system is about as perfect as you are going to get, its not flawed. 

If the weatherman tells you there is a 80% chance of rain today, and it doesn't rain, was the weatherman proven wrong? No. That simply implied there was a 20% chance it wouldn't rain.

You want to see if he is right, you have to look at multiple days with predictions. The last 100 times he made the 80% call, it should have rained in about 80 of those days. This will give you a clearer picture.

There is nothing at all problematic about it being Carlsen 87% one game, then a game later it is now Kramnik 65%. 

If you doubt the ratings are the best way top predict, go enter a tournament with these guys. How many games you think you will win? Why?

well actually the wetherman is wrong because there is no real 80 % chance, he is just guessing, at least according to the documentary i watched.

Avatar of Markle

Carlsen is human, he lost a game to a strong player nothing more nothing less it does not mean the rating system is flawed

Avatar of ChrisWainscott

Odds change based on a singular event all the time.

 

If we are playing a hand of hold em poker and I have three eights and you have three kings and there are still two cards to come I have roughly a fourteen percent chance of either getting another eight or a card to match one of my kickers.

 

On the next card an eight comes up.  I now have a 98 percent chance of winning.

 

Is the entire game of poker invalidated because the odds swung so mightily based on a single card?

Avatar of JM3000

Carlsen wons the last edition of Bilbao Masters, Tata Steel chess, Tal memorial and London Classic. When a player is the best he wins more elite championship that their opponents, but he don't win every tournaments.

The Elo is for measure a group of performance, but in one tournament the most important is the best peak in the tournament. The second part of the tournament for Kramnik is fantastic. If Kramnik Finally wins and recovered the world's tittle He obtain a record. He lost her crow in 2007 and any champion has recovered the title after years. 

Avatar of Vease
ChrisWainscott wrote:

Odds change based on a singular event all the time.

 

If we are playing a hand of hold em poker and I have three eights and you have three kings and there are still two cards to come I have roughly a fourteen percent chance of either getting another eight or a card to match one of my kickers.

 

On the next card an eight comes up.  I now have a 98 percent chance of winning.

 

Is the entire game of poker invalidated because the odds swung so mightily based on a single card?

Poker stats are based on pure Mathematics, theres nothing subjective about the numbers. ELO ratings are just an opinion reflected in numbers, they can't possibly predict with 100 percent certainty what will happen in an individual chess game. My argument isn't really with the number per se, just how it is used to support arguments that cannot be resolved, e.g player ability in different generations.

Avatar of ChrisWainscott

Elo ratings are also used to determine mathematical probability.

 

You are correct in that they can't be used to compare players of one generation to another.  i.e. you can't say Carlsen is better than Fishcer since he's higher rated. 

 

You can make something of a comparison by looking at the difference between ratings of top players.  i.e. the gap between the highest and second highest rated player.