chess.com ratings are deflated against USCF

I don´t think that they can trick the kids and make them believe that they are good players (if they are still rated 800 or 900), but they will think that improvement is easy and is going to be like that forever.

Going up a lot of point every year during your whole chess career.

That is what I was meaning. For me, making vast improvements easily = much talent....

That will just create other problems at the low end.  You have a player who was genuinely rated 600 but they now have a rating of 1000.  They lose almost every game they play, causing those who play against them to get higher ratings, and eventually you get inflation.

If you really really need to combat deflation, just give one extra rating point to every player with a certain amount of activity in a year.  It encourages activity and won't skew anything too badly.

Also... this thread is STILL going?  I posted in it like a year ago.

I think rating floors should have some expiry date. I know its not east to calculate that, since the occasional win raises them briefly above the floor. But its ridiculous to force a player to stay at 1900 for years if his actualy playing strength is now 1700.

It would require a lot of dedication for a sandbagger to keep at it for, say, 3 years.

I personally would hate to be stuck at a floor rating and I will likely quit chess OTB if it ever happens to me.

Ubik, you can appeal to USCF to have your floor lowered.

After much research, I think that both Elo and Glicko have some flawed logic. While I can't sit and tell you I have all of the math worked out to correct it, I can tell you that I have noticed a trend of research that is looking more closely at the flaws.

The flaws in the Glicko system are really brought to light when you look at the way the tactics trainer's rating system is set up. It amplifies the flaws in the Glicko system. Should a 1200 player be dinged for failing a problem that is rated 1000, simply because, five, 1600+ players drove the rating down of a what would actually be a 1400 puzzle ?

Eventually, a system that that uses move difficulty metrics, instead of awarding strictly on the a basis of wins, loses and draws, will finally show the true overall strength of a player. Some might argue against the logic used in my idea but, rating isn't necessarily a reflection of record, it is supposed to be a reflection of strength or likelyhood of performance.

In a study conducted to determine who was the greatest of the world chess champions, move difficulty metrics was used to determine the accuracy of play. It can also be used to determine the difference in the level of competition from one era to the next, far better than t-scores. Thus it gives a more accurate comparison than simply that of ratings, wins, loses and draws, or titles, length of reign, etc.

If such a system was put in place and used universally, it would improve rating accuracy, cheat detection ability, get rid of rating inflation and also it wouldn't require a pseudo starting point. Everyone could simply start where they are at. When matched for tournaments, regardless of experience level, the match ups would pit those of roughly the same raw skill level, against one another.

I am not saying it would negate the experience factor totally, but it would help to prevent much of the problem that was previously discussed about the kids being underrated, unrealistic rating floors for players who are forced to become perpetual door mats, etc. It would even make sand bagging really difficult. The effort it would take to make it look as if one gradually improved, instead of intentionally playing bad and playing their best, would make it very easy to police and not worth the trouble of trying to sand bag.

All that shows is the ratings for puzzles is flawed, not the Elo and Glicko ratings, which were designed for games.

Ratings are an estimate, not an exact measure. They may be close and they might not be, depending on various factors. I propose an actual system of measuring something that is static, instead of using estimates, to estimate other things, allowing for the compounding of error.

Nameno1had, that system has some issues:

1) It requires every game be properly recorded, submitted, and imported into the system.  For chess.com that part is not a problem; it already has a copy of the games.  For the USCF it would be almost impossible.  You should see some of the notation I took when I was learning.  As black, for some reason I often wrote my first move as P E7-E8.  I'm fairly sure I did not actually make a habit of capturing my own king to start the game.  Inevitibly, you will have illegibile scoresheets, incorrect scoresheets, incomplete scoresheets due to time pressure, typos from whoever gets the unenviable job of transcribing the scoresheets, actual illegal moves that are played during the game, games where the players started with the kings on the D-file or with the bishops and knights accidentally swapped, and probably more issues that I can't even think of.

2) It requires every move of every game be analyzed in-depth.  This would put a strain on the computing resources.  It's one thing to analyze the moves of a single grandmaster; it's another to analyze the moves of every game in every tournament in the country.  And the more you skimp on the computation, the more likely it is that your analysis will be inaccurate, so you can't cut corners here either.

3) Grandmaster games typically don't have positions where one side is winning by a queen.  Have you ever seen the computer analysis of a lopsided position?  It's constantly saying that you should have played THIS move for a mate in 10, and not THAT move which turns out to be a mate in 11.  A player might go for an obvious win instead of the shortest win.  Or they might decide to promote three pawns just because they can.  Such things really do not reflect on the player's ability.  And you can't just stop the analysis when one side is up by a queen, unless you want to miss the move where the losing side could have forced a stalemate.

Don't compare online ratings with OTB ratings....

There was a time when  9 out of 10 people didn't cheat. Now 99 out of 100 use opening books(allowed, which is a joke) and 9 out of 10 do cheat. Therefor the old statement isn't true anymore. What a waste of time.

I am glad you brought these issues into the open. I had given them some thought previous to our conversation. The first issue in particular is the one that stands to be the biggest obstacle, for this system to ever become universal, in my opinion. Money and politics would hamper this change. You are right about the attempts to convert all of the old records. It would be a daunting task.

However, my solution isn't to try converting all of it over night. In fact, it would actually require that old system's rankings would have to be used, while stats were compiled under the new system, until one year had passed. That way, everyone would know where they would stand under the new system, so that the old one could be done away with.

In the case of completely retired players, until it is more convenient or in the case that it is impossible, a good old asterisk should do the trick. This is happening in other competitive sports. It really shouldn't be that big of a deal. As for the sake of accurate comparisons between players of different eras, I have already mentioned that experts in this area stipulate that, special mathematics and research are required anyway, so that should be a moot point.

As for the available resources to continually calculate this extra data, there are a few things you should know about it. One is that the advent of improving computer technologies are making this sort of thing more and more feasible everyday. Another that is probably overlooked is that, engines and computers are already calculating this data, but aren't calculating it exactly the same as I propose. This minimizes the extra resources needed. It isn't as if additional data need be reviewed for game analysis. It is the same data. It will simply be looked at a bit differently.

As for your last point, you make my case for me in a sense. The fact that you mention GM's rarely obtain a queen advantage over their opponent's is quite telling. It simply states that every choice of the GM is generally harder to make than that of lower rated players, whom after making obvious errors, leave an easier time for their opponent's to choose the best move.

When looking at your statement regarding the mate in 10 and 11 move scenarios, it becomes even more clear to me that you still don't fully comprehend what I am proposing and it will further bolster my case that my system will work and is most accurate. The system I propose won't force players to always chose the best moves, most complex and shortest mating sequences, etc, so as to be rewarded properly. After all, if it requires a 3300 Elo computer to see the mate in 15 and the 2800 GM is still not seeing it, it is a clear indicator that the difficulty in finding a best move in one position, isn't the same as in others and requires a certain skill level to find those moves.

If it takes the 3300 elo rated engine to see the mate in 10 and the GM sees the mate in 11, he isn't going to be dinged for chosing an inferior line compared to the computer. In fact, this is also a huge advantage of this system, when it comes to cheat detection.

I'll admit, I don't have all of the math worked out in front of me to show you what Elo or Glicko rating it would take to accurately pick the best move in positions with a particular difficulty factor but, then again as I have stated previously, it isn't something I'd want to use for the finished product, if I am claiming it is flawed. I was simply stating it the way I did, to try to make a relative comparison at this point, just to try to give you a better understanding.

BTW, in using this system over the board, everything could and should be analyzed. Nothing would get thrown out. Even in correspondence, everything could be analyzed to see if perhaps someone was using an engine during the opening, then a database if possible, after the opening ran out of moves.

Yes, because I always check the MCO during blitz games.

Long explaination nameno1had :)

There was a time when  9 out of 10 people didn't cheat. Now 99 out of 100 use opening books(allowed, which is a joke) and 9 out of 10 do cheat.

Is this true? Or only for turn-based games?

I'm relatively new here. I've only played in the Dan Heisman Slow Chess League. I hit the opening books before the game, but I put them away once the clock starts ticking.

My impression of all my opponents so far is that they too are playing straight-up and honest.

I've played through many games I see in players' profiles and noticed no signs of cheating.

Which is not to say there is no cheating at chess.com, just that it is not nearly so prevalent as ghms claims.

If you don't like people using books, play only live chess. 

You can argue with AdamRinkleff but the fact is that I totally agree with him. The blitz players I play 1200-1300 are not 1200-1300 USCF rated players, they are much better. I came across this page searching for chess.com blitz deflation. If the Elo rating system has no objective standards, for example, there are no standards that you can say about an 1800 level player, then it is not a good rating system. Experience and Adam's mathematical evidence supports that chess.com blitz ratings are deflated as compared to chess.com standard ratings and USCF ratings.

Sometimes, some apples are shaped a bit more like oranges. I guess it is fair to say that apples and oranges are shaped about the same...

Since there is no such accredited university, in which either of our nations would accept in a court of law, recognizable by one another as proven authorities on the various subject matter, I can see no way to prove legally, my right to claim possessing enough of such knowledge, to rightly make or prove it in such areas. However, this also disqualifies you from the right to voir dire my education and dismiss it as not up to the legal standard, this of course, if we had some legal standard to adhere to.

Since a standard would have to be first established, by which we can base our rights to make claims or deny those of others. Based on the lack of such precedence to prove our rights and the lack of any recognizable authority, for you to try disproving my claims without one, is as foolish as p!$$!ng into the wind but, to each their own.

Since I have established the aforementioned ideas, I am willing to play by those rules to make my point. Since I can't make claims, I will simply ask a set of questions.

1) Are you an expert in the area of the mathematical working components of both Glicko and Elo ratings systems, that are used to in the two systems of play compared here that are used respectively, by Chess.com and USCF ?

2) Do you have the proper specimens from both santioning bodies to allow for proper comparisons of time proportion considering the length of blitz games is more standardized for rated USCF games and the specimens here could have played several lengths, not all of which, are acceptable to USCF ?     

3) Do you honestly think comparing the results of games played on line where cheating can't be ruled out, can accurately be compared with OTB play, in which, cheating is next to impossible in short blitz games, even if you had the same rating system for both entities and specimens that only participated in the exact game length settings, in every case ?

4) There is a serious sale on fruit going here, are you buying it ? Every variety of apples and oranges that you like...

Elo is designed to rate people within a population, it's not designed to rate people who are playing in different pools. It is possible to estimate ratings between different populations, but when the very method of making moves is different (online vs OTB), the results will also be different.

That's not a "flaw", that's how it works. Even this site recognises there are differences, which is why it offers different ratings for different types of chess.