the algorithm on this site is awful
Sort:
mwagih93
Jul 29, 2022
0
#1
it's Funny how it's been repetitive for years
you get like 10 or 15 normal opponents,you win at every game
then you get like a week of people who play like Kasparov
cheaters or not i don't know
but the whole algorithm is awful
pretty much like betting games
it's always win streak then like 20 games loss streak
it doesn't make since to be a coincidence everytime
Jul 29, 2022
0
#3
mwagih93 wrote:
it's Funny how it's been repetitive for years you get like 10 or 15 normal opponents,you win at every game then you get like a week of people who play like Kasparov cheaters or not i don't know but the whole algorithm is awful pretty much like betting games it's always win streak then like 20 games loss streak it doesn't make since to be a coincidence everytime
The pairing algorithm isn't doing anything other than looking for a compatible match. There is some logic around trying not to pair players with poor connections to those with good ones, those in the poor sports pool get paired together, but it doesn't look at playing history/streaks and just finds someone with a random seek at the same time control and where both ratings are within the seek range.
It doesn't make sense to add too many criteria to the process, which would just slow down the pairings.
You're just not that good. Go study and quit blaming the algorithm. Something tells me you're account has been put in the poor sports pool
0
#6
This is ridiculous. The chess.com randomly pairs u with random people. There's no devious secret algorithm set in place to make you lose games.
Jul 29, 2022
0
#7
mwagih93 wrote:
So everyone is suddenly a genius when you have a win streak ? Doesn't make any sense
Streaks happen.
I'm very curious to see these underrated Kasparovs you're playing against. Please post the games where it took a genius to defeat you. Links, or it didn't happen.
Martin_Stahl wrote:
mwagih93 wrote:
it's Funny how it's been repetitive for years you get like 10 or 15 normal opponents,you win at every game then you get like a week of people who play like Kasparov cheaters or not i don't know but the whole algorithm is awful pretty much like betting games it's always win streak then like 20 games loss streak it doesn't make since to be a coincidence everytime
The pairing algorithm isn't doing anything other than looking for a compatible match. There is some logic around trying not to pair players with poor connections to those with good ones, those in the poor sports pool get paired together, but it doesn't look at playing history/streaks and just finds someone with a random seek at the same time control and where both ratings are within the seek range.
It doesn't make sense to add too many criteria to the process, which would just slow down the pairings.
Do you have a source for this I had assumed it had a hidden MMR and to my knowledge chess.com has not releveled how it pairs players. Can I get a link? I know you can choose the elo range but a spamable website not having MMR seems odd then again i know streamers do speed runs but i had assumed server side differences for those accounts for that purpose.
0
#11
cokezerochess22 wrote:
Martin_Stahl wrote:
mwagih93 wrote:
it's Funny how it's been repetitive for years you get like 10 or 15 normal opponents,you win at every game then you get like a week of people who play like Kasparov cheaters or not i don't know but the whole algorithm is awful pretty much like betting games it's always win streak then like 20 games loss streak it doesn't make since to be a coincidence everytime
The pairing algorithm isn't doing anything other than looking for a compatible match. There is some logic around trying not to pair players with poor connections to those with good ones, those in the poor sports pool get paired together, but it doesn't look at playing history/streaks and just finds someone with a random seek at the same time control and where both ratings are within the seek range.
It doesn't make sense to add too many criteria to the process, which would just slow down the pairings.
Do you have a source for this I had assumed it had a hidden MMR and to my knowledge chess.com has not releveled how it pairs players. Can I get a link? I know you can choose the elo range but a spamable website not having MMR seems odd then again i know streamers do speed runs but i had assumed server side differences for those accounts for that purpose.
Your MMR is literally your rating... it's not hidden at all.
Win streak = higher Leo than your playing strength
Higher elo than your playing strength = better opponents = lose more
You are simply just mad because you are bad. That’s OK, there is no need to take it out on the website lol
0
#14
The guy lost 29 out of his last 50 games (roughly half) which is to be expected when playing people of a similar ELO. He's lost all four of his games today and is playing on a tilt. Do some puzzles and take some time out.
nMsALpg wrote:
cokezerochess22 wrote:
Martin_Stahl wrote:
mwagih93 wrote:
it's Funny how it's been repetitive for years you get like 10 or 15 normal opponents,you win at every game then you get like a week of people who play like Kasparov cheaters or not i don't know but the whole algorithm is awful pretty much like betting games it's always win streak then like 20 games loss streak it doesn't make since to be a coincidence everytime
The pairing algorithm isn't doing anything other than looking for a compatible match. There is some logic around trying not to pair players with poor connections to those with good ones, those in the poor sports pool get paired together, but it doesn't look at playing history/streaks and just finds someone with a random seek at the same time control and where both ratings are within the seek range.
It doesn't make sense to add too many criteria to the process, which would just slow down the pairings.
Do you have a source for this I had assumed it had a hidden MMR and to my knowledge chess.com has not releveled how it pairs players. Can I get a link? I know you can choose the elo range but a spamable website not having MMR seems odd then again i know streamers do speed runs but i had assumed server side differences for those accounts for that purpose.
Your MMR is literally your rating... it's not hidden at all.
Yes but The Glicko System stuff is hidden and much like it will adjust your elo differently based on how long since you and your oppoents last game wouldn't you expect it has a similar mechanic of hot streaks to prevent smurfs?
https://www.chess.com/article/view/chess-ratings---how-they-work
https://www.chess.com/blog/kurtgodden/elo-to-glicko-your-rating-explained
https://support.chess.com/article/210-how-do-ratings-work-on-chess-com
After reading these its certainly not clear to me that's not part of the system they are using.
So as its dscribed the system when you have a rating of 1600 could think "hes between 1200-2000" or it could think " he is between 1650 and 1550"
Are you saying that side of things has no effect on pairing and its all handled on the +1 elo side of the equation? That kind of makes sense it just wasn't very clear form the articles read that's how it works.
https://support.chess.com/article/369-how-does-matching-work-in-live-chess
The blurb at the top of this one makes me think that could be the case and is even implied but the fact that they seem to go out of there way to not directly say that makes me skeptical. In that if they think adding such a mechanic was overall good for players and that them knowing about it would make the upset I could totally see them slipping it by. I was looking for a link that says flat out form chess.com " The rating deviation form the glicko system on your account has no Barings on your pairings" . If they are already making tweaks for connection and behavior in the background I'm not going to assume they don't use a mechanic like this without them flat out saying they don't.
Jul 29, 2022
0
#16
@cokezerochess22 I can't day for certain RD isn't part of that pairing algorithm, but I would almost bet it's not 
If it was, it probably would make the most sense to attempt pairing high RD with low RD players to try and converge ratings more quickly, but it also seems like that would slow the pairing system down more.
0
#17
cokezerochess22 wrote:
nMsALpg wrote:
cokezerochess22 wrote:
Martin_Stahl wrote:
mwagih93 wrote:
it's Funny how it's been repetitive for years you get like 10 or 15 normal opponents,you win at every game then you get like a week of people who play like Kasparov cheaters or not i don't know but the whole algorithm is awful pretty much like betting games it's always win streak then like 20 games loss streak it doesn't make since to be a coincidence everytime
The pairing algorithm isn't doing anything other than looking for a compatible match. There is some logic around trying not to pair players with poor connections to those with good ones, those in the poor sports pool get paired together, but it doesn't look at playing history/streaks and just finds someone with a random seek at the same time control and where both ratings are within the seek range.
It doesn't make sense to add too many criteria to the process, which would just slow down the pairings.
Do you have a source for this I had assumed it had a hidden MMR and to my knowledge chess.com has not releveled how it pairs players. Can I get a link? I know you can choose the elo range but a spamable website not having MMR seems odd then again i know streamers do speed runs but i had assumed server side differences for those accounts for that purpose.
Your MMR is literally your rating... it's not hidden at all.
Yes but The Glicko System stuff is hidden and much like it will adjust your elo differently based on how long since you and your oppoents last game wouldn't you expect it has a similar mechanic of hot streaks to prevent smurfs?
https://www.chess.com/article/view/chess-ratings---how-they-work
https://www.chess.com/blog/kurtgodden/elo-to-glicko-your-rating-explained
https://support.chess.com/article/210-how-do-ratings-work-on-chess-com
After reading these its certainly not clear to me that's not part of the system they are using.
So as its dscribed the system when you have a rating of 1600 could think "hes between 1200-2000" or it could think " he is between 1650 and 1550"
Are you saying that side of things has no effect on pairing and its all handled on the +1 elo side of the equation? That kind of makes sense it just wasn't very clear form the articles read that's how it works.
https://support.chess.com/article/369-how-does-matching-work-in-live-chess
The blurb at the top of this one makes me think that could be the case and is even implied but the fact that they seem to go out of there way to not directly say that makes me skeptical. In that if they think adding such a mechanic was overall good for players and that them knowing about it would make the upset I could totally see them slipping it by. I was looking for a link that says flat out form chess.com " The rating deviation form the glicko system on your account has no Barings on your pairings" . If they are already making tweaks for connection and behavior in the background I'm not going to assume they don't use a mechanic like this without them flat out saying they don't.
Yeah, RD is hidden now, and the explanation may be "the math claims confidence on an interval from 1500 to 1600" but this is not a claim about pairings. At the center of that range will be the player's rating... so pairing by people's rating (visible rating) should be fine... and in any case active players have RDs below 100 (some below 20) and these people are frequently paired up or down by 100 points... so I think it's a safe guess that RD is not taken into account.
Hard to say I don't assume it is part of it I just also don't assume its not and if someone did have such a link I would love to see it.
0
#19
Martin_Stahl wrote:
@cokezerochess22 I can't day for certain RD isn't part of that pairing algorithm, but I would almost bet it's not
If it was, it probably would make the most sense to attempt pairing high RD with low RD players to try and converge ratings more quickly, but it also seems like that would slow the pairing system down more.
I assume the nature of active players is that they account for the majority online at any given time, meaning most players online players are low RD, so yeah, would guess it's redundant / waste of time.
0
#20
cokezerochess22 wrote:
Hard to say I don't assume it is part of it I just also don't assume its not and if someone did have such a link I would love to see it.
In any case, these sorts of complaints pop up now and then. They're always ridiculous. "The system if forcing me to lose" blah blah blah. Angry and dumb, not worth paying attention to.
But sure, you can be investigate the pairing logic if you want. Maybe you'll find something interesting, who knows.
Please Sign Up to comment.
If you need help, please contact our Help and Support team.