Thursday, January 02, 2014

2013 ChessTempo Stats

With regards to tactics and ChessTempo, my 2013 goal was to do 20 problems (blitz mode) per day.  At 365 days in a year, that works out to 7300 problems.  I did 7560 problems.  Granted I was scrambling to get a few more in on January 1st and 2nd, but it wasn't a marathon session.

Compared to 2012, my overall accuracy improved.  My average monthly accuracy average in 2012 was 77.55%; while my 2013 average monthly accuracy was 81.89%.  The highlight of the year in the accuracy category was hitting 85.00% correct in October (578 correct out of  680 attempted).  My overall accuracy percentage began 2013 at 72.15% and I finished the year with 74.43%

As far as ratings go - it's always up and down.  I try to look at the trends; and the trends say my rating is steadily declining.  For example, in 2012, I hit a max rating of 1752.6 and my highest monthly median rating was 1721.9.  In 2013, the highest I could hit was around 1748, while the best monthly median was only 1710.  2013 also saw me hitting a new low in blitz rating at 1633 (1637 was the lowest I hit in 2012).  In December, my max was only 1699 with the median at 1679.

So, in summary, it looks like I'm taking a bit more time (improvement in accuracy), at the expense of my rating.

For 2014 - my goal is to attempt 7600 problems; keep improving my accuracy, but also see if I can stem the decline in ratings.


  1. The ratingsystem at Chesstempo is not that easy. There is a duplicate reward reduction. Without a duplicate reward reduction your rating would improve because your performance on "problems you have seen before" = duplicates slightly improves every time you see these problems ( memory ). Without and compenstation the negative effect for chesstempo would be a rating drift. All problems would get rated lower and lower. So Richard did impement a rating punishment system: you rating get "punished" if you see duplicates. This punishment was calculated to keep the ratingsystem of chesstempo stable. Now: If your rating improvement on duplictaes is worse than the improvement of the avarage tactician your rating will decrease ( if you are not improving a lot !! on "nonduplicats" ) . This effect starts after about ~~4000 Puzzles because then you will have many duplicates.

    Uri Blass and i did try to convince Richard to use different alternatives to keep the ratingsystem stable but richard did not listen, sigh..

    The ratingsystem is relativly stable to a change of your thinkingtime. If you think longer then your score is higher = you get an increase in rating but you are slower so you get a decrease in rating.

    Empirical rabbit, Uri Blass and i did make several experiments and and calculations about that effect.
    As long as your score = thinkingtime=accuarcy dont grow extreme your chesstempo-rating will not be changing a lot, the ratingsystem does consider it.

    An extreme example is remiem:
    Stats for blitz tactics
    Rating: 1831.3 (RD: 168.84) (Best Active Rating: 1958 Worst Active Rating: 1757)
    Active Rank: Not Active/2210 (Best Active: 68 Worst Active: 320)
    Problems Done: 14510 (Correct: 13154 Failed: 1356)
    Percentage correct: 90.65% <--------------------------
    He was close to 100% for a long time and still his rating did not change a lot.

    My blog is full of analysis of chesstempo statistics. Seemingly no adult improves in tactics after a few thousand puzzles.

    I found only one exeption:. Zwisch is a young adult. he did chose "hard" problems and as a 2000 rated tactician there are not many puzzles of this rating at chesstempo. he does see douplictes very soon again so he can perform extremly good on them ( and he has a good memory for chesspositions anyway! ). But his performance on "nonduplicates" is by far not that good.

    The easiest method to see if you improve in tactics is to watch the "fide rating estimate" at ct. This is based only on your performance nonduplicates ( at least it was ).

  2. Aox - thank you very much for taking the time to explain that! I really appreciate it!

    I'll need to pay more attention to the FIDE rating and track it to see how I'm doing.

    I have consciously decided to try to be more accurate - but alas, my rating is suffering. But your insight explains a lot of that.

  3. just posting a link to this post ... this link summarized what Aox discussed in his comment:

  4. Depends on what your goal is. If you want to play better chess, then accuracy is what you should go for. If you want a better rating on Chesstempo or whichever site, then you need to combine that with beating the clock. In most real-life tournament games, if you can find a winning combination, it doesn't matter so much whether it takes 30 seconds, two minutes or five minutes.

  5. Accuracy and time are related: as more time you take as higher the accuracy and as higher the accuracy should be as more time you need
    Say: there is a set of problems where you accuracy is 50% if you spend 5 min per problem. Then you will have to spend ~ 10min per problem to get a accuracy of ~80% and~~20min per problem to get a accuracy of ~~90%. If you only want to spend 2.5 min per problem your accuracy will decrease to ~20%.

    One Tactician at chesstempo with a OTB rating of about 1700 did have a Standardrating of ~2500 by thinking much !! more than 1 hour per problem.

    Better Player find and check tactics quicker, because they know by experience/patternrecognition which move might be good and which not. Patzer like us need to check almost every move.. that takes time

    At a chessgame you cant search 1 hour for a tactic at each move. You need most of your time to see what does your opponent plan, what should you plan...

    You can easily change your accuracy at chesstempo by choosing "easy" in your preferences. Then you will see lower rated tactics. you will gain many % in accuracy.. but your ct-blitz-rating will remain about the same. If you solve one of these easy tactics you gain almost no points but if you blunder ( which will happen "sometimes" ) you lose dramatic many points.

    The goal is to spot tactical strokes quicker. There is no one near the board who tells you: "pssst, check for 30 min, it's worthid."

  6. Thanks ChessAdmin and Aox for the comments. I've thought a lot and read about about how to approach tactics on CT (blitz vs. standard). My goal is to be accurate more quickly and therefore I do blitz problems. Aox does a great job summarizing why blitz is probably the right approach.

    Now, if you really are trying to improve you calculation muscle (looking at every branch in the logic tree) and you don't care so much about the time-pressured component, then standard is the way to go.

    I seem to recall Heisman saying something along the lines of: don't spend more than 5 minutes on a problem, because if you can't figure it out by then, you probably never will.

    For me, having that time-pressure component helps me go through a lot more problems and it forces me to scan the board quickly, get a feel for the position and then hone in on the tactic.