matching algorithm
Sort:
It seems to me that the opponent matching algorithm draws people into addictive play. 4-5 easy opponents to get your confidence up and then 4-5 difficult opponents with lower ratings. It leads to streaky play and lets you get almost to the top of your rating before plummeting you downward. Has anyone else noticed this? Is there anything you have found that can keep it from happening?
Aug 28, 2018
0
#2
"4 or 5 difficult opponents with lower ratings"
How does chess.com know which players are difficult other than by using their ratings?
This sounds like the other conspiracy theory stuff that pops up now and then.
If you find it addictive, maybe what you actually like is chess. If one day you win 5 and the next you lose 5, then maybe (like all of us) you have good days and bad days.
It's in chess.com's best interest to pair people of similar skill levels. That's what people enjoy and (other than ranking professional players) that's the whole purpose of the rating system to begin with.
It’s so easy to criticize. Don’t get me wrong. I’m not complaining. And I do love chess. I just wondered if anyone else had noticed the same thing. No conspiracy theories. Just wondering how it all works. Thanks for your judgment of me, though. I will try to be a better human being in the future.
Aug 28, 2018
0
#4
People say a lot of crazy stuff, so yeah, I'm judgmental first and ask questions later.
Maybe look at your last 200-500 games to see if there's an odd win/loss pattern. That's something I'd find interesting.
And, as you probably know, even if it's truly a 50/50 chance, you're going to get streaks. Like flip a coin non-stop for just 5 minutes and you'll probably get 5, 6, or 7, in a row of heads or tails.
Preggo_Basashi escreveu:
"4 or 5 difficult opponents with lower ratings"
How does chess.com know which players are difficult other than by using their ratings?
This sounds like the other conspiracy theory stuff that pops up now and then.
If you find it addictive, maybe what you actually like is chess. If one day you win 5 and the next you lose 5, then maybe (like all of us) you have good days and bad days.
It's in chess.com's best interest to pair people of similar skill levels. That's what people enjoy and (other than ranking professional players) that's the whole purpose of the rating system to begin with.
it can measure by acurracy. i agree with op, it really looks like chess.com uses an algorithim to make people feel bad at chess and make them pay for the game. on lichess every match is a random victory or defeat, but here we only got sequences.
:3
johnco316 wrote:
It seems to me that the opponent matching algorithm draws people into addictive play. 4-5 easy opponents to get your confidence up and then 4-5 difficult opponents with lower ratings. It leads to streaky play and lets you get almost to the top of your rating before plummeting you downward. Has anyone else noticed this? Is there anything you have found that can keep it from happening?
Totally. That is in fact how their opponent matching algorithms work!
Jun 12, 2021
0
#8
mr-clumsy wrote:
johnco316 wrote:
It seems to me that the opponent matching algorithm draws people into addictive play. 4-5 easy opponents to get your confidence up and then 4-5 difficult opponents with lower ratings. It leads to streaky play and lets you get almost to the top of your rating before plummeting you downward. Has anyone else noticed this? Is there anything you have found that can keep it from happening?
Totally. That is in fact how their opponent matching algorithms work!
No it's not. There is no way to even do that. The only measure the site has of strength is the rating, and there is no way to programmatically determine what opponent a particular player is going to have problems beating, in order to pair you with that person.
The she simple fact is the system tries to find another player, in your seek range that I has compatible seek settings. There are some other considerations, but have nothing to do with anything that is strength related.
I guess it is possible there may be an attempt to pair people with win/loss streaks as part of the decision, but I've never heard of that being the case.
Jul 14, 2021
0
#9
Martin_Stahl wrote:
mr-clumsy wrote:
johnco316 wrote:
It seems to me that the opponent matching algorithm draws people into addictive play. 4-5 easy opponents to get your confidence up and then 4-5 difficult opponents with lower ratings. It leads to streaky play and lets you get almost to the top of your rating before plummeting you downward. Has anyone else noticed this? Is there anything you have found that can keep it from happening?
Totally. That is in fact how their opponent matching algorithms work!
No it's not. There is no way to even do that. The only measure the site has of strength is the rating, and there is no way to programmatically determine what opponent a particular player is going to have problems beating, in order to pair you with that person.
The she simple fact is the system tries to find another player, in your seek range that I has compatible seek settings. There are some other considerations, but have nothing to do with anything that is strength related.
I guess it is possible there may be an attempt to pair people with win/loss streaks as part of the decision, but I've never heard of that being the case.
Chess.com states~
========================
When you choose to play a rated game with a specific time control (like 5 min or 2|1), we try to find you an opponent who is closest to your current rating, because this makes for games that are rewarding and fun! It's also the most fair for everyone.
We also try to avoid matching certain pairs:
- We avoid matching players who have poor internet connections with those who have solid connections.
- We avoid matching players who frequently just disconnect or quit without resigning (aka "Bad Sportsmanship") with those who finish their games properly.
- We avoid matching players who are rude or abusive with those who are kind.
===============================
I totally agree with the possibility of "Streaks" being included into the decision process. So if you're on a 7 game streak, you could possibly be more likely to match with someone of a same rating, but a higher streak takes priority.
But other than that. Some players you wont see, if you have good vs bad internet connection, or if you or the other displays bad sportsmanship. But that wouldn't effect your game to game player pool, to the point of picking a harder opponent.
If you're on a streak and you keep plummeting after 10 wins or so. Welcome to your chess rating plateau. You have to learn something new to reach the next bracket.
Jul 14, 2021
0
#10
There are many issues with chess.com, but the matching algorithm is not one of them. You are just a bunch of cry babies who can't deal with losing streaks and desperately look for someone/something else to blame for your losses
Jul 14, 2021
0
#11
Chess.com is not transparent in general, not just its matching algorithm. It does not use FOSS software, so it is not transparent by definition.
This, however, does not prove that the matching algorithm is biased in whatever sense you have in mind. To establish the bias, you would need strong statistical evidence based on a large data sample. All I see here is anecdotes
Jul 14, 2021
0
#12
behappyhavejoy wrote:
The matching algorithm is an issue if it is not transparent and unbiased in how it makes matches and if the algorithm has been designed to manipulate user behavior. I too have noticed that after a successful run, I get hit with a non-random string of underrated players to knock me back down. As I stated in my earlier post, if the algorithm were not being manipulated by chess.com in a way that users would claim is unfair, the algorithm would be made public. The fact that they keep it secret, unlike in real-world chess, is proof that it is not fair. Happy to lose and learn in a fair system. Not a cry baby just because I want the rules to be transparent and fair.
https://support.chess.com/article/369-how-does-matching-work-in-live-chess
The site isn't hiding anything and has nothing to gain by trying to make the pairing algorithm more complex than it needs to be. It pairs you with other members that have compatible settings and attempts not to pair someone with a bad connection, with someone with a good one.
In the real world, randomly challenging people isn't really very common, so appealing to that doesn't make sense. Also, random pairings aren't tournaments, so rules governing those aren't pertinent either.
Jul 14, 2021
0
#13
I've seen no evidence, just anecdotes and speculation. Care to show some actual evidence?
Is it possible chess.com match your usual opening to someone who uses an opening that is known to be good vs yours? I could see this as a way for the algorithm to then challenge you. if you have a reportoire of openings and cycle through them, then you would be less predictable
I bet if I started a thread “Chess.com will always match you against higher rated players to keep your rating down”, no one would challenge me on the math or the logic.
.lMore players get assigned to play black than white “ is something I got away with once on another thread. I was proud of that one.
.lMore players get assigned to play black than white “ is something I got away with once on another thread. I was proud of that one.
I tend to agree with OP, chess is more than a raw rating, if you are in a bad mood you lose, no matter the rating. Chess.com can realize if you are in a bad move: checks if you have a bad win rate in your current run, then you are in a bad mood and it’s not hard to match you with a similar rating who just started to play.
This is just one speculation, another is that, when you just want to start to play, you are fresh and play better than when you played 10 games already, wo at start you get more points. There could be other reason for usual easy starts.
This is just one speculation, another is that, when you just want to start to play, you are fresh and play better than when you played 10 games already, wo at start you get more points. There could be other reason for usual easy starts.
Oct 15, 2021
0
#17
frankquietly wrote:
Is it possible chess.com match your usual opening to someone who uses an opening that is known to be good vs yours? I could see this as a way for the algorithm to then challenge you. if you have a reportoire of openings and cycle through them, then you would be less predictable
Is it technically possible? Yeah, if they wanted to drastically decrease the speed of pairings. It doesn't make any sense to manipulate pairings in such a way. There are already a few criteria for pairings, adding more would make pairing slower and harder to actually find a compatible opponent. There's no upside to trying to create an algorithm to break win streaks, create wins in order to hit reward centers, "insert your favorite conspiracy".
Dec 28, 2021
0
#18
johnco316 wrote:
It seems to me that the opponent matching algorithm draws people into addictive play. 4-5 easy opponents to get your confidence up and then 4-5 difficult opponents with lower ratings. It leads to streaky play and lets you get almost to the top of your rating before plummeting you downward. Has anyone else noticed this? Is there anything you have found that can keep it from happening?
Hi John,
You are correct.
I am an average(may be below) chess player, but I work in data science and I have experience in working with online gaming as well. So I know how these algorithms work to increase user time on application/website. I am 99.99% sure ONLY player rating is not used to decide opponents. Opening strategy and other factors are used to categorise opponents. So Chess.com gives you an opponent to increase probability that you will keep playing.
Some players in this thread disagree, there can be different segments of player on different algorithms for testing/training/improving purposes.
Well, after all it's a business and Chess.com does want you to play more to have more profits from ads etc.
AI is influencing internet interactions a lot. I hope you have heard about Facebook using addictive algorithms (The Social Dilemma on Netflix). Online gaming no different.
I know you can feel the Matrix. 
Cheers,
AB
you probably made a confirmation bias, we should test this for more cases to see if this is true instead of relying on some anecdotal evidence and some claims from someone in the field.
