Most likely on your alt, you're being matched with other players who say they're really good, but haven't played many matches at that rank either, so the transient effect of your own initial estimate persists for a significant number of games. If you play long enough I'm convinced that your alt rank will crash down to where you are on your original account. When you get matched against established accounts at that higher rank, I think you'll you'll get dominated with a very high degree of certainty.
Chess.com Ratings are a JOKE
Sort:
Jun 28, 2022
0
#23
TheBullyMaguire wrote:
I am 1100 rating.
Or so I thought…
After making a different account than my main one, 15 matches later, I am 1700 rating.
This demonstrates a misunderstanding of how the Glicko rating system works. When you create a new account, it assumes your first rating but gives you a very high rating deviation (RD). The RD will remain high until you have enough recent games to get you in the correct rating band.
So, if you started an account at 800 (at one point you could actually create one at 800, 1200, or 1800 - not sure if that is still the case) and played a game against another 800 and won, your rating would jump to ~1150. If you played a game against an 1100 next, and won, your rating would jump to roughly ~1500. If you played a game against a 1500 next, and won, your rating would jump to ~1750.
Does this mean you are as strong as 1700s? Your highest rated opponent was 1500, so no. If you played a 1700 next, and lost, your rating would drop back to ~1500. When your RD is high, it means the system does not have a high level of confidence in your current rating (your RD will also get higher if you don't play for a while). The goal is to get you into a rating band where your opponents will be roughly your same skill level. As you improve, your rating will go up.
This "second account" experiment is flawed because your second account does not have enough games to stabilize your rating (i.e. get your RD low). That usually starts to happen after you have a directional shift (i.e. if you lost a bunch of games and then started winning some, the same result happens in the opposite direction).
The problem with bots is that they do not make "human" mistakes. The problem a lot of people have with puzzles is, since they know there is a tactic in play, their tactical vision is more acute than it is in a real game. For example, I've seen kids with 3000+ puzzle ratings but 1600 OTB ratings. So you should expect your puzzle rating to be significantly higher than your playing rating.
Also, keep in mind the rating system is purely about sorting the players in that pool. So, an 800 on chess.com means that you are competitive with other 800s on chess.com and would almost always lose to someone in the 1200+ ratings (as you can see from the way the rating system works above, if you play other people with high RDs, this results in wild swings in ratings for a while).
Do new accounts face other new accounts or do they meet the normal pool of players?
Jun 28, 2022
0
#25
Jimemy wrote:
Do new accounts face other new accounts or do they meet the normal pool of players?
They get thrown into the normal pool, so it is possible they can play other new accounts.
It is worth noting that you can see a situation where a group of friends joins at the same time and plays each other for a while. You'll often see it with kids - when one kid (who is say, 1300 OTB) signs up and has all his friends sign up (who are all sub-900) and they just play each other for the first several games, the stronger kid will end up with a rating of close to 2000 very quickly. However, once he jumps into the random pool, it will settle down quickly as well.
Jun 28, 2022
0
#26
CooloutAC wrote:
I hear this myth all the time. But look at some real world examples. When the kid koweranai got banned, he got permission to make another account. He was a rated 2200 players. His new account must of started at 1200 then bumped him to 1400 after evaluation? But it still took him 200 games, with even a 50 game win streak in the middle there. to get back to his 2200 rating. So for almost 200 games he was skewing matches.
It is not a myth. That is literally how the Glicko system (chess.com uses the Glicko 2 system with a median of 1200 - LiChess uses the same system with a median of 1500) works. If someone was banned and created a second chance account, it is possible the staff manually set his RD for the new account, but a normal new account has an RD of 350 (if I recall correctly - they used to display in in the stats page, but it seems they have removed it), which will result in wild swings.
You can see the same thing by not playing for a while. When you come back, you can win/lose a LOT more points for your first few games.
@bdub76 was correct----go get an OTB rating for absolute accuracy...although besides just playing on Chess.com for "fun", you must also play for PRACTICE (i.e., hone your skills in midgame/endgame as well as experiment with different openings/defenses).
Jun 28, 2022
0
#28
CooloutAC wrote:
PawnTsunami wrote:
CooloutAC wrote:
I hear this myth all the time. But look at some real world examples. When the kid koweranai got banned, he got permission to make another account. He was a rated 2200 players. His new account must of started at 1200 then bumped him to 1400 after evaluation? But it still took him 200 games, with even a 50 game win streak in the middle there. to get back to his 2200 rating. So for almost 200 games he was skewing matches.
It is not a myth. That is literally how the Glicko system (chess.com uses the Glicko 2 system with a median of 1200 - LiChess uses the same system with a median of 1500) works. If someone was banned and created a second chance account, it is possible the staff manually set his RD for the new account, but a normal new account has an RD of 350 (if I recall correctly - they used to display in in the stats page, but it seems they have removed it), which will result in wild swings.
You can see the same thing by not playing for a while. When you come back, you can win/lose a LOT more points for your first few games.
On chess.com you can pick a starting rating of 400, or 800 if you want. which is crazy. Lichess starts everyone at 1500 rating no matter what.
And did you ignore what I said about my real life example? Go look at the account Kowarenai. He started a new account some time ago, it started at 1400. Took him 200 games to get back to his prior 2200 rating with a 50 game win streak in between.
AIUI
Machine doesn't care if it was 600 victories in a row, only that they won, or lost their last game.
As for HOW to get an OTB rating, you must play in tournaments at USCF events (United States Chess Federation). There are usually regularly-scheduled events in every state, and these listings can be found in CHESS LIFE magazine (of which getting a subscription is necessary as part of getting a USCF membership, which itself is prerequisite to playing in those tournament events. So go to their website (USCF.com or .org), and go from there.
Jun 28, 2022
0
#30
CooloutAC wrote:
On chess.com you can pick a starting rating of 400, or 800 if you want. which is crazy. Lichess starts everyone at 1500 rating no matter what.
And did you ignore what I said about my real life example? Go look at the account Kowarenai. He started a new account some time ago, it started at 1400. Took him 200 games to get back to his prior 2200 rating with a 50 game win streak in between.
Being able to pick your provisional rating or having it assigned doesn't make much difference. The system is designed to try to get you to the most competitive band fairly quickly.
And no, I did not ignore it, I specifically addressed it. If he got permission to create a second chance account, it is possible the staff set his RD manually. But looking at his account, that does not seem to be the case. The account was created on Nov 16, 2021, and by Nov 30, 2021, he was back up to a 2260 rating (and that was despite losing to a 900, 1200, 1400, and 1500 in his first couple days - which would make me question his 2200 rating ... the 900 looks to be a mouse slip as Qe8 is checkmate and he played Qe7, but the others just look like beginner games, but that is an aside). He literally went from 500 to 1400 in one day - less than 25 games (that is due to the high RD). It would have been faster had he not lost to such low rated players (as those losses drop your RD faster than if he had continued winning).
Watch when Eric, Hikaru, or Danya do speed runs. They start at 400 and usually within 10-15 games they are in the 1500 range. They usually do not start slowing down until they get in the mid-2000s (if I recall correctly, in one speed run, Hikaru went from 400 to 2000 in a less than 30 games). The reason they blow through the lower ratings is because of the high RD combined with the fact that they do not lose until they get to someone 2500+.
Jun 28, 2022
0
#31
The problem with folks starting at 800 is that if you’re 800, you get steamed rolled over by new accounts, making it really difficult to get out since you’re likely playing folks that are actually a much higher level. Throw in cheaters, and the frustration mounts.
Their account might say 800, but they’re really 1500+.
You’ll get much better practice games over at lichess at a lower level because you start so much higher. So the new accounts aren’t steam rolling you over after 200 games.
Their account might say 800, but they’re really 1500+.
You’ll get much better practice games over at lichess at a lower level because you start so much higher. So the new accounts aren’t steam rolling you over after 200 games.
CooloutAC skrev:
PawnTsunami wrote:
CooloutAC wrote:
On chess.com you can pick a starting rating of 400, or 800 if you want. which is crazy. Lichess starts everyone at 1500 rating no matter what.
And did you ignore what I said about my real life example? Go look at the account Kowarenai. He started a new account some time ago, it started at 1400. Took him 200 games to get back to his prior 2200 rating with a 50 game win streak in between.
Being able to pick your provisional rating or having it assigned doesn't make much difference. The system is designed to try to get you to the most competitive band fairly quickly.
And no, I did not ignore it, I specifically addressed it. If he got permission to create a second chance account, it is possible the staff set his RD manually. But looking at his account, that does not seem to be the case. The account was created on Nov 16, 2021, and by Nov 30, 2021, he was back up to a 2260 rating (and that was despite losing to a 900, 1200, 1400, and 1500 in his first couple days - which would make me question his 2200 rating ... the 900 looks to be a mouse slip as Qe8 is checkmate and he played Qe7, but the others just look like beginner games, but that is an aside). He literally went from 500 to 1400 in one day - less than 25 games (that is due to the high RD). It would have been faster had he not lost to such low rated players (as those losses drop your RD faster than if he had continued winning).
Watch when Eric, Hikaru, or Danya do speed runs. They start at 400 and usually within 10-15 games they are in the 1500 range. They usually do not start slowing down until they get in the mid-2000s (if I recall correctly, in one speed run, Hikaru went from 400 to 2000 in a less than 30 games). The reason they blow through the lower ratings is because of the high RD combined with the fact that they do not lose until they get to someone 2500+.
So he lost 4 games out of 200, one you say was a mouse slip? And you are ignoring the fact he had almost a 50 game win streak in between and that it took him 200 games to get to his prior rating. You consider that quick? To you quick is the fact it only took him two weeks? What about the 200 games he undermined in those two weeks? not a big deal? You are in denial.
If he created his new account on lichess, there you could say he was quickly rated. In this case you cannot.
I mean for a normal game you gain like +8 (once rating is moving normaly) for a win. This is a problem only if you are creating a new account. So therefor comes the question, why did he get banned?
Jun 28, 2022
0
#33
CooloutAC wrote:
So he lost 4 games out of 200, one you say was a mouse slip? And you are ignoring the fact he had almost a 50 game win streak in between and that it took him 200 games to get to his prior rating. You consider that quick? To you quick is the fact it only took him two weeks? What about the 200 games he undermined in those two weeks? not a big deal? You are in denial.
If he created his new account on lichess, there you could say he was quickly rated. In this case you cannot.
No, those were just in his first 75 games. He lost a lot of games in his first 200 and drew several more. He got to ~1900 in his first 100 games. He struggled in the 1800-2000 range, and he started playing in Arenas around that time frame (where he was playing players much lower rated at times - you don't get many points for beating a 1000 when you are close to 2000). Additionally, nowhere in his first 200 blitz games do I see a 50-game win streak. In fact, his stat page has his best win streak at 19 (when you consider his first 25 games were against players mostly much lower rated than he claims to be, that is rather unusual). In short, no, I'm not in denial. I explained how the system works and you attempted to use a real world example to refute it (despite the fact that you can literally look up the formulas used in the Glicko system) and even that example follows the proper pattern for the system.
I think the OP and others have a point, there is definitely something funny about lower order ratings. Some sort of unintended Rating RoadBlock Feature exists which effects mainly new or weaker players. You could join CCC as a newbie, get a stable rating in the range 500 to 1000 and never 'improve' after 1000+ games. There is something wrong here, it's almost as if you have to get sufficiently good to leap over a Rating RoadBlock without actually playing games against people! This is completely counter intuitive to pretty much all other forms of human 'improvement'.
Commerical chess platforms will argue that you have to make use of other features to obtain improvement (or enlightenment!). Yet it's highly suspicious - despite their love of toytown stats - that no objective measures of the efficiency for player improvement with usage are ever presented.
Frankly I think the ELO Rating system has had its day. The differences between platforms, formats and activities basically mean you end up with a dozen measures, the total of which can only give very rough overall estimation. A far better system would be one based on categories of average-centipawn-loss-per-move accepting that you can consider computer evaluations as being fixed absolutes - whereas with the ELO system it's possible for everyone to be improving yet without any rating change occurring along with the 'unknown' effects of player pool variations or behaviours, (if all the GMs in the world died tomorrow there would be an equivalent number of GMs - albeit less talented - in the near future).
Afterall, from watching the current Candidate Tournament, one does get the impression all that matters is how well the GMs can ape computers under fixed environmental conditions.
Jun 28, 2022
0
#35
TheBullyMaguire wrote:
CooloutAC wrote:
I believe there are separate hurdles and fences that once you jump over things get easier. It could be for the reasons you state, but also related is the fact chess.com allows new accounts to choose there starting rating. 400, 800, 1200. We all saw koweranai when he made his new account at 1400 take 200 games to get back to 2200 even with a 50 game win streak. So guys stuck at that 1400 with all the other koweranai's making their account at that rating will be struggling. But once they get to 1500 things will probably get a little easier until the next hurdle.
So ya, in short I agree with you although whether quitting, or simply making new accounts the reason is most likely the same. Aand its not just at the 1200 rating. its other various intervals.
Personally I like the rating system on lichess better. Everyone starts at the same rating and matches and ratings feel more consistent. i'm not hung up on the fact its a higher number then chess.com or OTB ratings, because thats irrelevant. The point of a rating system is for competitive matches and fair rankings, and to determine expected skill within the playerbase.
I believe you made a good point. I think it is at the intervals that people get hung up on. For instance I went had a hard time going from 800-1000 but once I broke it I went to 1200 instantly. Then couldn’t get out of that. I think my 1700 will be a more even playing experience because it seems like everyone I face is around the same skill, rather than at 1200 sometimes I will crush or be crushed.
My hurdles when I crossed over 1200 was how many countless cheaters and sand baggers I ran into at the bracket 1200-1500 tournaments.
Some truth to the OP from my personal experience.
CooloutAC wrote:
AlexiZalman wrote:
I think the OP and others have a point, there is definitely something funny about lower order ratings. Some sort of unintended Rating RoadBlock Feature exists which effects mainly new or weaker players. You could join CCC as a newbie, get a stable rating in the range 500 to 1000 and never 'improve' after 1000+ games. There is something wrong here, it's almost as if you have to get sufficiently good to leap over a Rating RoadBlock without actually playing games against people! This is completely counter intuitive to pretty much all other forms of human 'improvement'.
Commerical chess platforms will argue that you have to make use of other features to obtain improvement (or enlightenment!). Yet it's highly suspicious - despite their love of toytown stats - that no objective measures of the efficiency for player improvement with usage are ever presented.
Frankly I think the ELO Rating system has had its day. The differences between platforms, formats and activities basically mean you end up with a dozen measures, the total of which can only give very rough overall estimation. A far better system would be one based on categories of average-centipawn-loss-per-move accepting that you can consider computer evaluations as being fixed absolutes - whereas with the ELO system it's possible for everyone to be improving yet without any rating change occurring along with the 'unknown' effects of player pool variations or behaviours, (if all the GMs in the world died tomorrow there would be an equivalent number of GMs - albeit less talented - in the near future).
Afterall, from watching the current Candidate Tournament, one does get the impression all that matters is how well the GMs can ape computers under fixed environmental conditions.
The problem with that method is that accuracy only depends in relation to your opponent. Its not a real measurement of skill and matters even less in speed chess games Where different strategies and tactics are employed.
The problem is anonymity, period. And for 30 years game companies have refused to address it aggressively because they don't see it as being as profitable for them. The truth is they don't care that much about fair play as long as they are making money. I guess lichess is unique in that they are a non profit open source platform. They aren't perfect, but at least they haven't manipulated their own system trying to attract more players.
We are at the point now though that facebook has demanded people use real names and at times has demanded copies of birth certificates lol. Twitter over the last year has freaked out and have been banning people left and right for no reasons. It took years for game companies to even have 2FA like google pioneered with emails almost 20 years ago. But they are always behind the curve and chess.com is even behind on that.
And if you think the ELO system is bad, imagine all the sites that have been using it in team games to determine a players rating. Even worse! I remember iD software bought a rating system from GaimTheory for Quakelive, . It was the best rating system for a random team fps I have seen, made by stockbrokers, sports team owners and casino owners. but they never had the guts to actually use it and ended up using it halfheartedly and undermined it themselves by catering to the top players. The fatal move of most gaming sites. Cheaters took over that game and drove it into the ground and founder and ceo John Carmack couldn't stomach it and left the company. lol
The problem is we simply don't apply the same principles to the digital realm that we value in the physical realm. RIOT with league of legends was the closest any gaming company has come to being a respected sport. Even shown on cable TV about 7 years ago. But once again, the alt accounts and smurfers undermining the mmr slowly drove the game back into the ground even with all the aggressive measures RIOT has taken. EPIC has started suing parents of kids who cheat, and even that is still not enough.
The only place these games are truly popular is in Korea and Japan. And I bet we would get great matchups if only playing in those countries lol.
most competitive video games use some sort of ELO system, and most of them are fairly accurate, the alts and cheaters are not the norm, but the exception. in competitive league, the ratio of cheaters to legit players is pretty good. of course im not saying that there are none, the system is still flawed, but its pretty good when evaluating someones playing strength
bdub76 wrote:
The problem with folks starting at 800 is that if you’re 800, you get steamed rolled over by new accounts, making it really difficult to get out since you’re likely playing folks that are actually a much higher level. Throw in cheaters, and the frustration mounts.
Their account might say 800, but they’re really 1500+.
You’ll get much better practice games over at lichess at a lower level because you start so much higher. So the new accounts aren’t steam rolling you over after 200 games.
Their account might say 800, but they’re really 1500+.
You’ll get much better practice games over at lichess at a lower level because you start so much higher. So the new accounts aren’t steam rolling you over after 200 games.
I know from playing MMOs it's very common for players to spawn off ALT characters when their main character's progression slows up. Extremely common in fact, people just crave for progression even if you throw in a lot of repetition. From MMOs people just what to win irrespective of the intrinsic value of the win. Also, you get cheats in MMOs as well, I was always puzzled by this as there was no real advantage to be gained, my conclusion was that people cheat because they will if there are no real-life consequences. it's part of the gaming culture.
Maybe gaming culture is ruining on-line chess for beginners!?
On LiChess you can switch the whole rating system off and just play Casual games. Such games can be far more friendly than on CCC, which lacks the feature. Also, you are unlikely to come across toxic behaviours in Casual games, although players are far more likely to abandon lost games - albeit they receive a warning for doing this.
Jun 28, 2022
0
#38
AlexiZalman wrote:
bdub76 wrote:
The problem with folks starting at 800 is that if you’re 800, you get steamed rolled over by new accounts, making it really difficult to get out since you’re likely playing folks that are actually a much higher level. Throw in cheaters, and the frustration mounts.
Their account might say 800, but they’re really 1500+.
You’ll get much better practice games over at lichess at a lower level because you start so much higher. So the new accounts aren’t steam rolling you over after 200 games.
Their account might say 800, but they’re really 1500+.
You’ll get much better practice games over at lichess at a lower level because you start so much higher. So the new accounts aren’t steam rolling you over after 200 games.
I know from playing MMOs it's very common for players to spawn off ALT characters when their main character's progression slows up. Extremely common in fact, people just crave for progression even if you throw in a lot of repetition. From MMOs people just what to win irrespective of the intrinsic value of the win. Also, you get cheats in MMOs as well, I was always puzzled by this as there was no real advantage to be gained, my conclusion was that people cheat because they will if there are no real-life consequences. it's part of the gaming culture.
Maybe gaming culture is ruining on-line chess for beginners!?
On LiChess you can switch the whole rating system off and just play Casual games. Such games can be far more friendly than on CCC, which lacks the feature. Also, you are unlikely to come across toxic behaviours in Casual games, although players are far more likely to abandon lost games - albeit they receive a warning for doing this.
You can't get better when you end up playing folks using engines or significantly better. It's a lot of frustration. I wish chess.com used the ? like lichess, so that I knew I was playing someone that's new.
Online gaming culture is pretty toxic. Gamergate was ridiculous. I really don't care what's someone's online rating if they don't have an OTB rating that matches. it's a number. It's pointless if it doesn't properly match you up with folks at the same level.
CooloutAC:
Don't disagree with much of what you have said, especially your point about anonymity being the key factor.
Regards my Average-Centi-Pawn-Per-Move system. (ACPM). True when you play and win against weaker players you score much lower CPMs, and against stronger players you score much higher CPMs. However, the point of a rating seem is to match similar talented players. I am also sure you would end up with a dozen measures as per the ELO system. However, I believe changes to the ACPM measures would be more believable - and useful - giving that the datum have a fixed reference point. I am sure there would be other more novel problems.
I would however suggest that if a person's chief aim is improvement and associated self-worth, one would be better off only playing computer bots on-line, and perhaps reserving playing real-people only OTB. Just a thought!
There is certainly plenty of scope for improvements to the computer bots which are currently open-loop. The day will come when computer bots will analyse a player's games, coupled with an in-depth database of human failings and draw-up gameplay for player improvement, thereby becoming closed loop. Training Bots do seem the direction the technology is going, in chess and in other areas of human learning.
It is all outcome based. The amount your score fluctuates initially is quite a lot until it after a few games a level is found and you need to climb. So let us say you win a few games by good play or opponents play bad then your rating might have stabalised higher rather than lower before any subsequent losses have a hit to your score. Once a score finds a stable level advancement takes a lot of wins. Also just because your opponent has a level their rating is at doesnt mean they play at a consistent accuracy level. You can beat people a few hundered points above you and lose to those a few hundered bellow and such inconsistent play means you dont move much in Elo. These things happen.
Yeah, that's why it doesn't mean much unless he shows the account...
Need to check that most of the games lasted around 30 moves, were against different opponents and not just rematching 1 guy, etc.