Is a 100 point difference actually significant?

100 points difference is time referant and your ability is more like a graph, goes not only up , but up AND down.  like this S

Yes windows, I believe skill fluctuates often. This one time, I played 10 1-minute games in a row. And after which I played 5-minute and 10-minute games, I had like a 7 loss-streak, some to really bad players too! Once I got the hang of longer games again, and slowed my playing down, started being cautious, now I have a 14-win streak.

What I am wondering is, if you see some average joe play a game. Can you guess what his rating would be at chess.com?

If you're 1600, and you see some guy who is truly 1400 player, can you guess he's a 1400 player more easily than a Grandmaster who's suppose to be more knowledgable about chess but may not be able to tell the difference between 1300 and 1500, because to the GM, everyone is noob??

Is this scientific or perspective/relative? IF we were to take an average of mistakes, blunders, and inaccuracies in chess games of certain rated players, would be come up with a consistent pattern?

Would you be able to assign traits to a 1400 player and a 1600 player, and therefore, by attacking and demolishing these bad habits and enforcing these traits, we can instantly turn a 1400 player into a 1600 player?????

These are the important questions...

On chess.com I don't think 100 pts is very significant.

In USCF or FIDE, I do.

Fischer lost points for winning the World Championship.

'nuff said.

I'm more about using the rating system to look at my progress. I've been at around 1300-1400 for years, and I'm wondering what I have to change in order to reach 1500 or 1600... This is why I was looking for patterns.

Certainly not if people are implementing artificial rating floors. Really, there are better ways of dealing with sandbagging than injecting clearly false data into the system.

My one problem with Chess.com's ratings is the way they deal with closed cheater accounts. But that at least is a temporary problem, which gets resolved over time as ratings recalculate to their correct level. A floored rating will mess up a system as long as that player is active.

I've managed quite a few upsets otb.

But I don't think I've ever beat a player on chess.com ranked more than 200 points above me at the time of the game.

 That's probably true of 70% of the players on the site...while 100 points is not alot on this site, 200 points is more or less a wall and if you were betting on a match with a 200 points rating difference you would do well to bet on the guy with 200 points more, it would be putting your money in the bank. 

I honestly thought due to the K factor (or whatever it's called) that ratings above (2000 or 2200 whenever they implement it) do increase or decrease slower so are more stabilized and are more accurate.  Or is that not a correct conclusion, to think that they're more accurate because the points come and go more slowly?

However even if this is the case, it's still the same formula with the same expected score, so 100 points at any level (amateur vs GMs) it will be the same right?

BTW ichabod you're right, I always think of it is a chance of winning but it's an expected score.

Within the Elo system, it's the same expected score. What changes is how fast your rating changes to reflect the actual score being different from the expected score.

Yeah, that's theoretically correct I think.

But in practice, at lower levels, there is a randomness in the frequency and severity of mistakes that influences the randomness of the result.

I guess what I'm saying is, that at the very top level 100 rating points should convert to your 64% score after 10 games, but at 1200 level you may possibly need many more games to reliably get this type of consistency.

This makes sense in my head, although I dont feel I explained it very well.....

I don't know if 100 pts on chess.com is significant...assuming we're talking about players who've all played 500+ games, I guess it might be? 

I beat players 200 pts more than me at least once a week, so not sure. Maybe it's harder to do that as you get higher? 

Is that FIDE or chessdotcom? Your rating theoretically is a relative strength between you and the rest of the players. 1600 today is a different skill than 1600 a hundred years ago or the next 10 years. So if you want to improve your rating, you just have to improve your skill quickly. Find your weakness (thing that determines your losses) whatever it is, fix it, and your rating will improve. Forget about new skill (or learning about a few more positions), because you can add more skill but still lose because of one same weakness.

As to the "mistery" behind the rating difference between lower rated players and higher rated players: ELO rating system is not perfect. It is a chart of a inverted exponential distribution where if the skill/rating difference is too big the accuracy will suffer. The biggest problem is of course in the top level where top players will have to play lower rated players with huge gaps in rating. This way top players must perform better than expected. They can win a tournament but they may still loose rating point, so they must unnecessarily (with the risk of losing the tournament prize) avoid draw.

The "fact" that chess is a theoretical drawish game, which becomes truer in GM level, makes it difficult to win against each other. Plus the fact that once you are at the top you have to play more lower rated players than higher rated players makes 100 ELO rating difference is really a huge gap in GM playing skill!

The lower the level, the less importance the 100 rating difference is. And if you follow the same logic when we discuss the ELO inverted exponential chart, you will find out that lower rated players have the "advantage" from the ELO rating system (You can play for a few tournaments and get a huge leap in rating).

And then you have to know that standard rating is different with blitz rating. (Even the ELO system is not accurate for blitz games) or turn-based rating.

In blitz rating I can go above 1600 and I can go below 1000. Turn base is also different because some players play the games in blitz style but some others use engine or think harder than "usual".

For a simple idea you can go to www.exeterchessclub.org.uk where you can find grouping of knowledge, skill and training material based on USCF class level.

May be it is not about ridiculousness, but, if you don't change the system, less high rated players will want to participate in "lowly" tournaments.

Imo, the problem with current ELO rating system is not a big issue as long as FIDE doesn't rely heavily on those ratings to determines who should play in the candidate events.

Don't know why you say ELO isn't accurate for blitz... it's accurate for any head to head match chess or otherwise.  I see you say your blitz rating can fluctuate many hundreds of points... that's not common, I would think that has more to do with your consistency than a flaw in the statistics formula.

Yes, it is important to observe phenomena, and more importantly is to draw the conclusion or take lessons from them. And I think you can do it fairly well.

There's a key point in your post that I want to address. My blitz rating can go above 1600 and also can go down below 1000 (in other site). The question is why? The answer is the same with the point in your post about playing 5-minute games after playing many 1-minute games.

Remember the advice: try to find the best move in every move you make. This is a very very important advice!

My play is better if I try to think in every move that I make. But sometimes I'm too tired to think but I still want to play games (and have fun). Sometimes I just dont care what my opponent reply will be I just be ready with my next move.

In blitz you have to take into account the clock. But if you want to fix your reflex after playing faster games, you just need to think on every move you make and forget about the clock. Then you will see that you will "win" the game but lost on time. Then the next step is easier. You just need to do faster "what" you can do slowlier.

The "playing always the best move" advice is important for me because I know that there are situations where my physical fitness just doesn't allow me to do so. I can easily be tired, especially in tournaments where I have to wait for a long time before the game start. Sometimes I know it if I cannot win any game. The physical fitness issue is the most important issue with me.

I think here 100 points aren't anythink becouse often we play just for fun and in un-serious mode, I find myself many times in a game in wich I made a crazy attacking move and after a week returning to the game with the idea of playing more positionally, this ruins the rating  of course and all of us maybe in the same game can play very good moves for some time, then somethink non related to chess changes our humor and you play bad moves, all depends. And as other pointed there are the players who reach 2000 and more by playing low opposition and that inflates too, non talking of cheaters.

I’m aware that this is a complicated issue. The idea is simple but the discussion may be complicated, just as you can see from the previous discussion (And also my mistake in grammar).

It is important to note that when someone says “elo rating” he might means “FIDE” rating system, which also changes over time. This is confusing, but let me assume that we all know that we can use a brand new formula, or use the elo formula but implement it differently.

If we use the Elo “scheme” it is then important how to determine the K-factor, or how to get an establihed data to based the calculation from. Elo original idea is based on a few simplification, a fixed K-factor (32 if I’m not mistaken), and a normal distribution which is not true.

Note about K-factor: If based on elo table you should perform 8.0 but you perform 8.5 then you are 0.5 better than expectation and your rating will be increased by 0.5K, or 5 points if the K-factor is 10, or 10 points if K=20.

So the elo system is based on data. You should have enough data to have an “established” rating. Once your rating has been established (i.e. in GM level), you can reduce the K-factor so to increase the “accuracy” of the system (If the rating fluctuates easily, how can you predict the outcome of next matches??)

Blitz, turn-base and standard games are almost 3 different games. Players ability to play turn-based chess is different with his ability to play blitz and his ability to play Go (or any other head-to-head games). So we should have a separate data for blitz rating, those are games played in 15/0.

So we must have G/15 established data for generating an established blitz rating. The ability to generate an established rating (or data) to predict future result is the accuracy of the system. And we know now that Elo formula is less accurate in low level chess than in high level chess because we cannot easily get an established data in low level. If we want to say that Elo system is not accurate, that is absolutely valid because that what statistics (and probability) is, but we have to give a comparator to be more responsible.

So to say that Elo is not accurate for blitz, I should use standard games as comparator.

Now the discussion should be easier I guess, as we can focus on why it is easier to see performance consistency in standard games than in blitz games (e.g. level of injected blunder or level of uncertainty in those two games)