Why is the elo on chess.com so bad?

Sort:
Ca_boom

Its honestly a really bad system because it is encouraging people to spam new accounts. When you first make an account and win a random game you get +200 elo so you jump to 1400 with one game or if you lose one game you jump down to 1000/1050. This is DEFINITELY intentional by the site so that people make new accounts constantly and thus the site seem to have a lot of traffic because after 10-20 games you only get 6-7 elo PER GAME. That's incredibly low because lets say i have an old account at chess.com where my elo is 600 (when in reality my real elo is 1300 now). In order to get to my REAL elo i have to win more than 100++ games.... and imagine if those games are 30 mints RAPID. This is RIDICULOUSLY slow. Problem number 2 is that with so many new accounts on chess.com you cannot possibly know how strong your opponent is at the range of 1100 to 1400 elo or the range were the new accounts are being put in general. You can face a PRO who just started or made a new account, a cheater who made a fresh account to cheat or a complete beginner. It makes the elo 1100-1400 and any other were new accounts are put an illusion. Why cant we just get more points when we are on a win streak? Why do we get 200+ points when we have a fresh account but when we dont we get 6-7? 

 

This elo system is super unbalanced and needs fixing. Please just improve it because unless you are not a pro and you play against pros at high elo (were you basically play with the same people all the time because they are very few up there) this number means nothing. 

And even if you are not convinced about my argument i have one more. I asked 5 friends with a real elo at chess who also play chess.com. 2 of them had way higher elo at chess.com than their real elo and 3 of them had lower with the second person having way lower (250+ elo difference) . This also proves that chess.com elo is unbalanced.

KetoOn1963

I dont know since chess.cm doesnt use ELO.

KetoOn1963

Maybe they should use REO Speedwagon.

Ca_boom
KetoOn1963 wrote:

I dont know since chess.cm doesnt use ELO.

You get my point. Whatever this number that represents your skill is called

KetoOn1963
Ca_boom wrote:
KetoOn1963 wrote:

I dont know since chess.cm doesnt use ELO.

You get my point. Whatever this number that represents your skill is called

That number does not represent your skill level.  A rating is an estimate of your most recent performance.

KetoOn1963

Say you have an OTB rating of 1800.  You play in an OTB tournament, and finish with a rating of 1784.  You did not suddenly forget 16 rating points of chess knowledge.  That 1784 represents your performance in that tournament.

About 600 years ago i played in Las Vegas at the National Open.  I was a 1600 player at the time.  I played up in the Under 2000 section.  I finished with a rating of 1794.  I did not suddenly develop 100+ rating points of chess knowledge.  It is an estimate of how i did in that tournament.

GambitShift

Yea, I don't understand how I lose to people lower rated and then get paired with people higher rated. Then all of a sudden I beat someone who is my highest opponent for the day. It doesn't make sense.

 

Even when you play the bots, they change their level. A few games will go down with blunders, then all of a sudden no blunders. It's like they are programmed to play the 3rd best move in one game and then the 1st for a while in the next.

blueemu

https://en.wikipedia.org/wiki/Glicko_rating_system

The big rating swings when you first join are intentional, yes. It is to enable your rating to "home in" on its proper value as quickly as possible.

Consider this case:

A new player has no idea what his proper rating should be. All he knows is that he regularly beats the other kids at his school. When he joins, he picks a strong initial rating (1800).

In fact, he is only 1000 strength. He doesn't know this, because he has never faced real competition before.

If he was using a simple rating system (say, plus 8 points for a win, minus 8 points for a loss) it would take 100 lost games before his rating sank from 1800 to its natural value of 1000. At that point, he would be matched against other 1000-strength players, and would win about as many games as he lost.

But with the Glicko-2 rating system, he will drop to 1000 within six or eight games, and from then on will face competition of his own skill level.

- "lets say i have an old account at chess.com where my elo is 600 (when in reality my real elo is 1300 now). In order to get to my REAL elo i have to win more than 100++ games...."

Not true. Your RD (which determines how many points you gain with a win or lose with a loss) goes down with more games played, but it goes up as time passes without playing. So if you have "an old account" where your rating was much lower, then your RD will be high (because you must not have been playing recently, in order for a 700-point skill gap to open up) and high RD means that you will gain LOTS of points with each win.

I suggest that you learn how the rating system works... and THEN criticize it.

Ca_boom
blueemu wrote:

https://en.wikipedia.org/wiki/Glicko_rating_system

The big rating swings when you first join are intentional, yes. It is to enable your rating to "home in" on its proper value as quickly as possible.

Consider this case:

A new player has no idea what his proper rating should be. All he knows is that he regularly beats the other kids at his school. When he joins, he picks a strong initial rating (1800).

In fact, he is only 1000 strength. He doesn't know this, because he has never faced real competition before.

If he was using a simple rating system (say, plus 8 points for a win, minus 8 points for a loss) it would take 100 lost games before his rating sank from 1800 to its natural value of 1000. At that point, he would be matched against other 1000-strength players, and would win about as many games as he lost.

But with the Glicko-2 rating system, he will drop to 1000 within six or eight games, and from then on will face competition of his own skill level.

Yes, but thats only if the only other option is for him to lose 6-7 elo per game and not more. What if everyone starts at a base elo lets say 1200 or 1000 or even BETTER you start with no elo and you get your placement elo after 10 games that you will play depending on how much you win or lost and by who and then you get or lose 25+- points every game instead of 6-7. When you are having a winning streak you will get more points and when you are in a losing streak you will get less and less until you start winning again. That way your elo will have the ability to accelerate with your skill lvl. I feel this would be a much more balanced environment. 

blueemu

Did you actually read the link I game you?

The Glicko-2 rating system has had a lot of thought put into it... probably far more thought than your suggestion, which would make established ratings too volatile.

It is also a sad fact that people who have GAINED 250 points in the past month (instead of LOSING 250 points) seldom call for a revision to the rating system.

Ca_boom
blueemu wrote:

Did you actually read the link I game you?

The Glicko-2 rating system has had a lot of thought put into it... probably far more thought than your suggestion, which would make established ratings too volatile.

It is also a sad fact that people who have GAINED 250 points in the past month (instead of LOSING 250 points) seldom call for a revision to the rating system.

Its not a competition of who put more thought into it. It is simple logic and reasoning. Having no elo and getting a placement after you play 10 games is way more balanced than picking an elo and then telling all this weirdness happen with you getting less and less points the more you play until you barely get anything anymore.

Ca_boom
blueemu wrote:

Did you actually read the link I game you?

The Glicko-2 rating system has had a lot of thought put into it... probably far more thought than your suggestion, which would make established ratings too volatile.

It is also a sad fact that people who have GAINED 250 points in the past month (instead of LOSING 250 points) seldom call for a revision to the rating system.

By the way your wiki link is inaccurate. Counter strike doesnt use the Glicko rating anymore and i think it never really did. It only used a version like Glicko. neither does Guild wars 2. I play both games.

blueemu
Ca_boom wrote:
blueemu wrote:

Did you actually read the link I game you?

The Glicko-2 rating system has had a lot of thought put into it... probably far more thought than your suggestion, which would make established ratings too volatile.

It is also a sad fact that people who have GAINED 250 points in the past month (instead of LOSING 250 points) seldom call for a revision to the rating system.

Its not a competition of who put more thought into it. It is simple logic and reasoning. Having no elo and getting a placement after you play 10 games is way more balanced than picking an elo and then telling all this weirdness happen with you getting less and less points the more you play until you barely get anything anymore.

And what happens in the case you gave earlier: you have an old account that is 600 rating, but your real strength now is 1300. Under the system you just proposed, it will take dozens of games to catch up.

Meanwhile, for a player with an established rating that is a good match for his playing strength, one bad patch (say, a bad day at work that makes it hard to concentrate, or construction work near his home is making his internet connection unstable) will cost him over 100 rating points.

Your "rating system" is poorly though out. I'll stick with Glicko.

Ca_boom
blueemu wrote:
Ca_boom wrote:
blueemu wrote:

Did you actually read the link I game you?

The Glicko-2 rating system has had a lot of thought put into it... probably far more thought than your suggestion, which would make established ratings too volatile.

It is also a sad fact that people who have GAINED 250 points in the past month (instead of LOSING 250 points) seldom call for a revision to the rating system.

Its not a competition of who put more thought into it. It is simple logic and reasoning. Having no elo and getting a placement after you play 10 games is way more balanced than picking an elo and then telling all this weirdness happen with you getting less and less points the more you play until you barely get anything anymore.

And what happens in the case you gave earlier: you have an old account that is 600 rating, but your real strength now is 1300. Under the system you just proposed, it will take dozens of games to catch up.

Meanwhile, for a player with an established rating that is a good match for his playing strength, one bad patch (say, a bad day at work that makes it hard to concentrate, or construction work near his home is making his internet connection unstable) will cost him over 100 rating points.

Your "rating system" is poorly though out. I'll stick with Glicko.

I don't understand where you got those numbers. With my system a player with an old account will be able to catch up to his real rank much quicker than with Glicko because of the acceleration of his win streak that he will have, while someone who had a bad day will never lose 100 points. He will have to lose more than 10 games in a row to get close to that.  A lose by having a bad day will cost him nothing more than the basic 20/25 elo he will get back easily the next day.

I think you are just too quick to dismiss my idea because i am a nobody in your eyes.

BouseSause

The system makes sense but you are looking to use it wrong, I'd be curious as to why you want to use the elo system like this as in my mind there is no point in creating an 10 accounts so you can have 1 game and a 1800+ elo when in reality you are 1300. Elo systems aren't designed to evaluate your skill level on the result of 1 game.

It's worth noting that the more games you play the more accurately an elo system will find an accurate skill level for you so the idea behind having huge swings of points is to allow players to get to their skill bracket the fastest, But give it a couple of 100 games on any account and you will end up with roughly the same Elo rating, regardless of if your first few games 

ThroughTheStorm

Unfortunately, that is how the Elo system works. If you truly believe you are a 1300 rated player. You should go to your 600 account and queue against 1000+ rated players. (You can choose the specific rating of players you want to be matched against)

By winning games against 1000+ rated players you will gain Elo very quickly since you are technically 600. 

The Elo system is exactly the same as the one used in FIDE. the limitation here is that unlike in FIDE were they play Over-the-board and there is no room for cheaters, chess.com is online and sad humans are bound to abuse the software and cheat.

Ca_boom
BouseSause wrote:

The system makes sense but you are looking to use it wrong, I'd be curious as to why you want to use the elo system like this as in my mind there is no point in creating an 10 accounts so you can have 1 game and a 1800+ elo when in reality you are 1300. Elo systems aren't designed to evaluate your skill level on the result of 1 game.

It's worth noting that the more games you play the more accurately an elo system will find an accurate skill level for you so the idea behind having huge swings of points is to allow players to get to their skill bracket the fastest, But give it a couple of 100 games on any account and you will end up with roughly the same Elo rating, regardless of if your first few games 

elo systems are also not designed to push people into making new accounts because they are not build for a free to play website but for other places like irl chess. I stated all the problems i have with the current elo system and what it is causing to the site. You didnt address any of them, instead you claimed that the system makes sense without disproving my points and also mentioned how the current elo system works.

lcravethatmineral
Ca_boom wrote:
blueemu wrote:

https://en.wikipedia.org/wiki/Glicko_rating_system

The big rating swings when you first join are intentional, yes. It is to enable your rating to "home in" on its proper value as quickly as possible.

Consider this case:

A new player has no idea what his proper rating should be. All he knows is that he regularly beats the other kids at his school. When he joins, he picks a strong initial rating (1800).

In fact, he is only 1000 strength. He doesn't know this, because he has never faced real competition before.

If he was using a simple rating system (say, plus 8 points for a win, minus 8 points for a loss) it would take 100 lost games before his rating sank from 1800 to its natural value of 1000. At that point, he would be matched against other 1000-strength players, and would win about as many games as he lost.

But with the Glicko-2 rating system, he will drop to 1000 within six or eight games, and from then on will face competition of his own skill level.

Yes, but thats only if the only other option is for him to lose 6-7 elo per game and not more. What if everyone starts at a base elo lets say 1200 or 1000 or even BETTER you start with no elo and you get your placement elo after 10 games that you will play depending on how much you win or lost and by who and then you get or lose 25+- points every game instead of 6-7. When you are having a winning streak you will get more points and when you are in a losing streak you will get less and less until you start winning again. That way your elo will have the ability to accelerate with your skill lvl. I feel this would be a much more balanced environment. 

You're a genius! 

Ca_boom
ThroughTheStorm wrote:

Unfortunately, that is how the Elo system works. If you truly believe you are a 1300 rated player. You should go to your 600 account and queue against 1000+ rated players. (You can choose the specific rating of players you want to be matched against)

By winning games against 1000+ rated players you will gain Elo very quickly since you are technically 600. 

The Elo system is exactly the same as the one used in FIDE. the limitation here is that unlike in FIDE were they play Over-the-board and there is no room for cheaters, chess.com is online and sad humans are bound to abuse the software and cheat.

This is my point exactly. That the elo system needs improving because it was not build specifically for online chess on a website. The people who own this site need to make adjustments to it to make it more balanced.

lcravethatmineral
Ca_boom wrote:
BouseSause wrote:

The system makes sense but you are looking to use it wrong, I'd be curious as to why you want to use the elo system like this as in my mind there is no point in creating an 10 accounts so you can have 1 game and a 1800+ elo when in reality you are 1300. Elo systems aren't designed to evaluate your skill level on the result of 1 game.

It's worth noting that the more games you play the more accurately an elo system will find an accurate skill level for you so the idea behind having huge swings of points is to allow players to get to their skill bracket the fastest, But give it a couple of 100 games on any account and you will end up with roughly the same Elo rating, regardless of if your first few games 

elo systems are also not designed to push people into making new accounts because they are not build for a free to play website but for other places like irl chess. I stated all the problems i have with the current elo system and what it is causing to the site. You didnt address any of them, instead you claimed that the system makes without disproving my points and also mentioned how the current elo system works.

We are lucky to have geniuses like you in this world. Suprisingly, people like you are very common.