31 May 2009

Rating Estimation

It seems that a lot of chess players new to or unacquainted with tournament chess want to know their rating. Others who have played online, but not OTB (over the board) frequently ask how one rating compares to another. There are several ways to estimate one's rating, but all should be treated with skepticism.

The series of ten problems at Chess Maniac might be better than most. I tried it a few days ago and scored 1830, which is reasonably close to my recently gained USCF rating of 1819. Less reliable are the Chessmaster personalities. Vlad, for example, allegedly plays at 1846, but the software does not adjust for processor speed. Few human A class players commit the egregious positional errors in the opening that are Vlad's trademark, nor are they as adept at tactical calculation. When I entered B class, I played a sequence of training games against Vlad, finding him surprisingly easy to outplay positionally.

Rating History

In the early 1990s, I was coming back into chess after more than a decade away. I had played in some team matches in high school, at least one club event at the Spokane Chess Club, and had joined the US Chess Federation and played postal chess. All these activities ended by the time I had been in college one year. Among my circle of acquaintances, the second best player could beat me occasionally. I did not renew my USCF membership and I stopped playing postal.

I bought my first personal computer in 1989, learned that Chessmaster 2100 was available, and that it allegedly played at a master level. Even though I was too busy with graduate school for play--aside from a fanatical obsession with golf during the summer,--I started learning a bit about computer chess. Finding the right level for playing against the box pulled me into efforts to estimate my own level of skill.

Through CM 2100, 3000, 4500, and 5000, I used several features of the program to approximate my skill level, which also increased slightly as I played against successive versions. I also spent some time in the library reading Chess Life, especially Bruce Pandolfini's "Solitaire Chess." In "Solitaire Chess," students guess the moves of a master level game and earn scores for right answers and some alternatives. The total score at the end of each game corresponds to an approximate rating.

When I returned to the Spokane Chess Club in fall 1995, I told the club's contact person that I thought I might be an A Class player. But, at my first club meeting, I played a series of pick-up games against an E class player with an approximately even score. In March 1996, I entered my first USCF rated tournament, scored 2.0 of 5.0, and earned a provisional rating of 1232.

Competitive Edge

My first published non-provisional rating nine months later was 1425. My estimate based on various exercises turned out to be high by 400 points. But, my initial published rating very nearly could have been much higher. In the 1996 Washington Class Championships, I had Black and move on board one in round four in the position below. A win in this game would have guaranteed me a tie for first, and likely would have given me an initial non-provisional rating over 1500.

I spent quite a bit of time assessing the consequences of 19...b4 with the idea of driving the white queen completely out of play on a1. I played a better move, 19...Ng5, but not the best move. However, as my opponent improved his position move after move, and his queen became active, I lost my way.

The win slipped away, and I found myself in a drawn king and pawn endgame.

Black to move

Inexplicably, I played 45...Kf4?? After this loss, I also lost round five and ended the event sharing places 8-15.

I returned from the event and told a close friend that I no longer had the competitive edge that had been my norm in high school cross country. My loss demonstrated that chess skill development included emotional preparation and mental focus if I was to have the success I sought. My friend played chess casually, but well understood competition. He has been a state champion in cross country, and had run for two college teams, including a leading Pac-10 school. Last fall, the high school team he coaches became the first team west of the Mississippi to win the national championship.

My rating bounced around in the 1400s for the next five and one-half years, finally jumping above 1500 in February 2002. It dipped back into the 1400s for most of 2004, and in 2005 started a steady climb up to 1601. Once I reached B class, I did not drop below, and after my performance last weekend, reached A class. To get to A class, I've needed to learn tactics and endgames, and expanded my opening repertoire. But, more important, I've needed to reduce the obvious blunders made in haste, and I've needed to become far more consistent against lower rated players. I've also increased my drawing ratio.

In the USCF rating system, a rating of 1800 creates a floor of 1600. If I never win again, I will never drop below B class. But, I still have ambitions. My current established rating remains a crude underestimate of my capabilities.


  1. Nice historical blog entry. :-)

    One thing i dont agree with and this is this floor rating thing. Why did they put this in place? What if a player drops down in strenght and plays worse (lets say at 1500 level) then his floor but thanks to this floor stays at lets say 1600? Is the rating of his opponent not bogus aswell since their rating gets calculated with that 1600 while it should have been 1500?

  2. Interesting excursion into the past.
    chesstiger - the floor exists to avoid sandbagging, if you played well, you can't later lower your rating by much. In Canada, we don't have it. Some people can play worse during the time, as I noticed, due to their age. Still, in US there are lot of tournaments with big prizes, so I can understand it.

  3. Don't quote me on this, but I think they changed the rating floor thing within the last 10 years. Before that it seemed like the bottom was endless or maybe 400 points lower than your top, now it's 200 down from your highest hundred point level.

    I googled and see a lot of chatter about it stemming from around 2006. Keep in mind that one of the problems back in the day is that people were leaving chess (and taking their rating with them) so that chess ratings were actually deflating (points leaving the system as it were). Now I think we are staring at the inflationary side of the coin with this floors thing.

    My FICS is around 1800 and my OTB is around 1750, so it's not all that far off. My OTB rating is still rising, too.

  4. linuxguy,

    The floors have changed, but not in the manner you describe. The USCF rating floors changed from the 4th edition (1993) to the 5th edition (2003) of the U.S. Chess Federation's Official Rules of Chess in a manner that should be deflationary, not inflationary. To wit, the 4th edition established floors at 1400-2300 when a player achieved a rating 100 points above each point. Those rated below 1500 had a floor at 100. Now, one must achieve a rating of 1600 to establish the 1400 floor.

    Under the old rules, my current floor would be 1700, rather than 1600.

    I've heard many players remark that 1600 players today are much stronger that 1600 players of the 1970s, even though the percentage of players over 1600 seem to have grown. Although it is difficult to credit such impressionistic judgments, the causal explanation seems credible: computers are often credited with stimulating a general increase in the playing strength of average tournament players.

  5. There are floors at 100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 2000 and 2100. The point difference to determine your floor is 200 points. So you need a peak rating at or above 1400 (but below 1500) to achieve the 1200 floor. Once you get a peak rating of >=1500 your floor will increase to 1300....etc

    Anybody with a peak rating below 1400 has a floor of 100.

  6. Has anybody tried www.ChessZen.com Looks like their ELO estimate by analyzing the games is accurate. They also have estimated ELO ratings of various world champions including Morphy, Fischer and Kasparov.

  7. Well, I've tried the chessmaniac ten problems, and it gave me a 1950 elo rating (spent an hour on it, but still). I have never done OTB tournament, but I have been playing seriously and studying for only four months, and unless I am some kind of new Bobby Fischer which is unfortunately extremely unlikely, there is very little chance that I have a 1950 elo or anything approaching.

    I think the only way to get an idea of one's elo rating is to play OTB and do things properly. FICS elos is completely broken since you can abuse low level computers so easily once you have found their weakness.

  8. Also try www.elometer.net