Chess.com Ratings are a JOKE

You can click on their names, it not like chess.com hides if they are new. It will tell you when the account is created and how many games they have.

> Let’s say all the people in that rating are improving in skill at the same rate you are. 

^^^ that's the flawed assumption.

Accepting this crude mental model, in reality, people don't improve at the same rate, and they are matched up against each other.   That little bit of noise is enough to ensure they get sorted properly.

Now, if you add smurfs / sandbaggers and what we're not allowed to mention but will instead call "advantage players" to the mix, then that statistical model does not predict ratings will "stay the same", but instead will fluctuate wildly and randomly.   Hm, kind of like they tend to do here....

Yes there is something to be said for having a Chill-Out Account.  I only really noticed this on LiChess as the feature is very obvious. Not sure how you play unrated on Chess.com as I never noticed this and all my games were rated - this actually put me off playing as I didn't want to risk-my-rating and frankly didn't always need the stress of playing. I have a tendency to play late in the day so the last thing I wanted was to lose 50 or 100 rating points just before bedtime. 

With this account I intend playing only the Bots, starting at the bottom and working my way up till I find Bots that beat me at an acceptable level. I also play them with unlimited time which gives me a very significant advantage but should saturate my talents. Aside from the lack of stress, eventually I should always be playing chess against more talented opponents and hopefully learning faster - this is not easy to achieve on chess.com with player games. Also the Bots have far more consistent gaming strength than players, which frankly at lower levels the ratings are all over the place. 

This is one of the reasons I think the ELO system at lower levels is very dodgy - it just doesn't work as intended and is an extremely poor predictor gaming strength.  There is just too much randomness. You could easily set aside an hour to play 5 or so Blitz games, lose every single game dropping say 100 rating points. This is just silly when you are in the rating range 400 to 800.  There is no way you have become that less talented in the space of an hour.  The other feature is that you can drop much faster than you can climb - you can see this in the low level profiles. You lose 100 to 200 rating points in a few hours and take weeks to recover but you never gain 100 to 200 rating point in a few days and take weeks to drop.  This is why there is talk of Walls and Roadblocks. The ELO system is just not doing it's job for whatever reason. The system is behaving as if there is a significant flood of fully talented 1000+players into the the sub-1000 player pool, day after day, week after week, month after month. Heck maybe this is exactly what is happening, who knows. Clearly I would expect the situation to be less severe the higher the ratings.

Anyway, my plan for the next year is to buy into the consistency of the Bots. I have no idea whether improvement will be better but I am certain there will be a lot less frustration. I am using playing against real players on LiChess to objectively determine the outcome.

For what it's worth, after a month and a half of this 'experiment' my LiChess rating has smoothly increased by 100+ to 1600+ after 70 games played, although I am sure this is more to do with my starting chess talent being recognised than anything else as I have started with the very lowest level Bots on Chess.com, which are not much of a challenge as yet. 

Allowing for rating differences between platforms the 1600+ is about 100 to 400 higher than the maximum I ever achieved on chess.com over a period of three years., where my rating randomly bounced between 800 and 1000.  Of far more importance is that the LiChess feels solid and less random. If I put the effort in I would hope to see rating improvement, this was never the case on chess.com, no variety of efforts yielded any 'feelings' of improvement, yet clearly improvement of some sort must have been occurring giving the duration if nothing else. Ironically one thing that has been consistent across the platforms has been my ELO puzzle rating, 2300+.  

Time will tell, but I have no regards about giving up playing real people on Chess.com. So far, I have a much more stable gameplay experience to build on and measurably improve from, imo. 

One thing I would HIGHLY RECOMMEND to chess.com is to fix the starting rating as per LiChess. It may not be a cure but would offer some degree of protection to beginners and lower level players. The ELO rating match-up system in lower level games is just not working!

ELO does not measure skill or talent, just past performance.  The notion of skill comes from the fact that, in order to climb the rating ladder, you must be performing consistently.

The problem at the lower levels, and why it seems almost random, is the players at those levels make mistakes quite often and it is random if their opponent sees it.  Not long ago, I saw a game between 2 ~400-rated players where White blundered his queen on move 8.  At the 1500+ level, the game ends here; however, they kept playing and Black blundered his own queen, and a rook within about 10 more moves and White ended up winning the game.  It is these kinds of situations you see at the lower rating levels that makes their performances seem random.

Hello

You only throw in the towel on a blunder at your level because you believe the opponent's rating is an accurate reflection of their talent, which is highly likely to be the case. But what would happen if you didn't have such faith in the ELO system, would you throw in the towel then? Could it be that the behaviour of lowbie players is merely a reflection of their experience of the inaccuracy of ELO system at these levels?  And yet the accuracy of the ELO system is invariant. Being weak at chess doesn't make you a foolish person.

Have you ever, at say 1900, played against someone of the same level, and timed-out with your opponent having more time than they started with? I have played lower level players that would put Magus to shame. I didn't think they were cheating they were just far better players, yet had a similar ELO rating.  It's the frequency of such events that lead you to consider the ELO system dodgy. You will never experience this at higher levels, sure you lose games, but you're never going to experience the RoadKills with anything like the frequency that occur at lower levels. And yet the whole point of the ELO system is to produce 'fair' match-ups, there shouldn't be any RoadKills if the system was working as intended.

What's random is not the skill of the players, - I doubt this changes much day to day, if not week to week - it's the randomness of the rating being a true reflection of the skill of the players. At your level the ELO rating is highly accurate, and you depend on this being the case in order to achieve wins and avoid loses as you would alter your gameplay accord to the strength of the opponent, at low levels the system is highly inaccurate. Whatever the variation in randomness of gameplay it gets trumped but the randomness of the match-up inaccuracies of the ELO system in low-level games. Sure it shouldn't work this way, but at the enter-point of chess.com it does.

To quote "The Queen's Gambit":  "When you lose your queen in this manner, you resign."

Players that can play at that level or higher know that if they drop a piece and have no compensation for it, the game is over.  Their opponents will consolidate and simplify.  So yes, they "throw in the towel" because they know they have no compensation for the lost material and the game should progress in a single direction.  To give you an idea, people in the 1800-2000 rating range (on this site) convert winning advantages of +3 or more over 88% of the time.  People in the 2000-2200 range convert those advantages over 90% of the time.  The same advantage is converted only 70% of the time in the sub-1000 ranges.

The behavior of the players at the lower levels is not a reflection on the ELO/Glicko rating system at all.  It is a reflection on their inexperience and lack of knowledge of the game.  You see the same thing when you to go a local kids' club.  The beginners play chaotically and end up with random results.  You can see some masters (i.e. Dan Heisman) refer to it as "hope chess".  I never said being a weaker chess player made someone foolish - just that they lack the understanding, knowledge, and skills of stronger players.  When I play an 800, they look foolish.  When I play a GM, I look foolish.  When a GM plays an engine, they look foolish.  That is all relative, but it simply a demonstration of skills that have been developed (or in the case of engines, programmed).

You will often see low level players claiming "I've played against people that would put Magnus to shame".  I'll answer this simply by saying that Magnus has forgotten more about Chess than all of the sub-1500s in the world know, combined.

Regarding the increment games, that is simply a matter of playing quickly and premoving.  While mouse speed and premoving are skills that can help you with online play, they are not skills that help you get better at chess (even if they help you get better results online).  Who is really the better player:  the person who wins a blitz game on time from a completely busted position, or the guy who lost on time from a completely winning position?

Again, rating is not a reflection of the skill of the player, but of past performance.  An 1800 is a better player than an 800 simply because they have demonstrated the ability to consistently win against lower rated players.  The main reasons you see people stuck in the sub-1200 range is their lack of tactical vision (i.e. they blunder often).  Look at the games of any random sub-1000 rated player (someone who has been in that range for a while, not a new account that has only been there for 2-3 games) and you will likely see they have a ton of mistakes in their games regularly.  That is why almost every coach I've ever talked to has said one of the first things aspiring players should focus on after learning how the pieces move is basic tactics.  That is why programs like the Steps Method focus on tactics for the majority of the program designed to take a complete beginner to ~2200 FIDE.  The same thing can be seen in the 9-book Yusupov series (Build Up Your Chess/Boost Your Chess/Chess Evolution).  The problem with players getting stuck in the sub-1000 (really, sub-1500) rating ranges has almost nothing to do with ELO being a bad system and everything to do with the players in those ranges not having the tactical skills.  There is nothing wrong with that - we all were there at some point.  You simply cannot blame "the system" when the problem is your lack of study, or simply studying the wrong things to help you improve.

I am one of those "low level" players. I've been at a high of 1127 and am now just above 1000. I've played almost 400 games. What PawnTsunami says is accurate as it applies to me....except for the part about knowing anyone who could beat Magnus. I make mistakes regularly and while blunders are diminishing, still there in most games. I recognize my own lack of tactical skill. Once in a while I'll play brilliantly, but it's anything but consistent. I had an opportunity to fork king and queen with a knight yesterday. That opportunity was there for two moves, and I never saw it.  Not seeing it cost me the game. I scan the board looking for check moves, discovered check moves, hanging pieces, forks, weakly defended pieces, etc.... and I see lots of these during the course of a game......but then at some point, I will miss a big one..... or my opponent will and it will determine the outcome. 

But I love the game and if I never get above 1200, I'll continue to enjoy it.

And I don't find the term "low level" insulting or demeaning. Compared to a 2000+, it's just exactly what we are......kind of like a "high handicap" golfer.  Truth is truth. 

"People in the 2000-2200 range convert those advantages over 90% of the time"

I'm in that range (somehow??) And in longer games that is more than true. But if you're a blitz player at say 2000 who hangs a piece in 10 moves, don't resign. I mark that I've stabilized at my current rating when I can beat someone around me after hanging a piece in the opening about 25% of the time...

Keep in mind I've "stabilized" at 2100+ blitz.

In rapid, you have more of a point. But 10 min games are really more like blitz...

Just know I didn't kill myself when it happens.  

Giving the format you play, 1 hour per player, I doubt anything I have said would apply to your experience. I would expect the ELO system to be pretty solid.

If you want to get a rating above 1200 may I suggest you mix things up a bit and play some Daily Games.  You can very easily bias the games in your flavour by taking much longer than opponents when considering moves. 1400+ should be possible after a few dozen games, which will take a few months to complete. 

Thanks for that great suggestion, but the "rapid" format really works well for me. I'm retired and can devote a few solid hours to playing, and if I blunder a game, I can just start another and try to do better.  When I think about the problems in my game, I don't think about rating, I think about why I didn't see that fork or that bishop lined up on my rook. I want to see those things consistently.....within the hour. 

But again, thanks for taking the time....I do appreciate that. 

I have learned in this thread that @kowarenai is a great example of how the rating system works - how he reached a reasonably accurate rating after a few days and much less than the 200 people who are in denial of the facts claim. It doesn't matter where people start with their rating: once they've played a few games, it will get to a level that reflects reasonably accurately where they sit relative to others in the same rating pool.

@TheBullyMaguire Thank you for making it easy for the Support team to action your multiple accounts. Probably they'll just give you a warning and let you choose which one to keep, but I expect your "other" account that is currently rated 1700 will drop shortly back to the 1100 level once you lose some more games and the RD on that account settles down.

Leagues have encouraged people to stick with the one account and play more games with it, so the ratings of those folks in the higher leagues should be super trustworthy.

Verification is a move towards allowing people to identify themselves for the purposes of entereing a cash prize tournament - unfortunately, it looks like they're not actually collecting much personal details, but presumably Chess.com will refine this as they go.

It would be great if Chess.com offered seek filters for these attributes, as that'd knock out the vast majority of new accounts and potential cheaters.

oh i used to hate that crap at FICS too! my rating at 15m was about what i've crawled back to now... 1500, but it would be 1300 in 10 minute! it makes no sense because i always had 5 extra minutes in 15m! in my mind, i'm spotting my opponents free minutes and watching my ratings sink anyways.

that, and i REALLY made a "2000" mad when i beat him TWICE in a row in blitz.

one rating regardless of time limits! i'm with ya on this all the way OP

oh... i read this wrong... it's 2 different accounts. if you play the wrong opponents, your rating can get artificially high or low when you start out. i had a run from 1450 to 1620 a couple weeks ago, but now i'm back around 1550.

i see ratings as merely a measure of improvement or lack thereof myself. people lose their freakin' minds over ratings!

I guess someone should get a 2200+ rating for beating 100 straight sub-1000 rated players.  Makes sense ...

You have already admitted to being bad at math.  You may as well add reading comprehension to that list.

Accuracy of play is what matters but depending on the position.  2 players can play accurately because they played a rehearsed opening, then traded the pieces down which engines tend to not penalize heavily for inaccurate piece trades as long as the game comes out pretty even after the trades.  Now on the other hand if both players are playing accurately in difficult positions where the only good move is extremely hard to find, and are doing this all through the middlegame, then you are dealing with a very very good player.  This is about a 1300 on this site, and 2400+ OTB.  There is a reason why a 1300 is in the 90th percentile.  When in FIDE a 1300 should be a bottom feeder noob who just learned on how the horsey moves a couple weeks ago.