05 August 2012

Training Log

My focus the past two weeks has been drifting away from chess. Although the U.S. Open began yesterday only a few hours from my home, I am traveling in the other direction. When I teach a six week history course, as I am doing beginning 14 August, my attention shifts almost wholly to it for a few weeks before it begins. Nevertheless, I keep up on daily chess: tactics training, correspondence games, occasional rapid and blitz games online.

After an eight hour session working on my history course on Friday morning, I played ten 3 0 games on Chess.com. Losing five of the first six games was frustrating, but then I finished with four consecutive wins. The past few weeks, I have played very little 3 0 chess, using most of my online playing time for 15 0 games.  Reducing junk is important to chess improvement.

As for tactics training, I solved a handful of problems in the Shredder iPad app, the Tactic Trainer app, and attempted 113 problems through four sessions on Chess Tempo. My two posts the past week, "Failing at Tactics" and "Tracking Failures," detail some of the challenges through which I struggled. My last session on Wednesday was ten problems with an 80% correct rate.

The graph shows steady progress upward through the end of July until I scored a mere 41% correct through 80 problems on Monday. I regained over half of the rating reduction through a mere ten problems. My average recent time per problem went up from ~75 seconds to 98.

I waffle a bit over how low I should aim to keep my average time per problem. Is speed important in training, or do I gain more from seeking a high degree of accuracy?

Stats for Standard Tactics
Problems Done: 2481 (Correct: 1316 Failed: 1165)
Percentage correct: 53.04%
Average recent per problem time spent 98 seconds

I spent a bit of time going through game three from the Spokane City Championship, but have not yet completed my annotations for posting.


  1. Hi James,

    If you use standard mode in CT, you don't need to rush. Imagine it's a normal OTB game and take as much time as you would in a critical situation OTB.

    If you want to test your tactical speed, you can switch to 'blitz mode' (Options/) where you can test how fast you are compared to other players. However, be warned that the 'blitz' pool in CT is stronger than the 'standard' pool (which is already stronger than the average USCF...)

  2. A related question is how do I measure my performance, given my time and score? I discuss that in my latest article:


    That is not what you asked, however. There appears to be good evidence that solving simple problems quickly using spaced repetition improves your performance at solving those problems. The same technique probably works for harder problems too, but you would need to tackle a very large number of them. Coakley recommends writing solutions down before looking up the solution in the book. He has trained a string of Canadian Junior Champions, so perhaps he knows what he is talking about. This training certainly makes you slower in the short term, however. There is also lot to be said for a happy medium.

  3. For me its not the speed but the accuracy that matters. Afterall, the more you do solving puzzles and look only for accuracy the more some positions and ways of solving things will be ingrained better in your brain and therefor will eventually speed up your solving.

  4. The argument against excessive training for accuracy is that you get slower and slower and that the number mental faults identified and fixed per unit time is low. It is easier to add extra checking than it is to speed up. If you do too much checking in training you also risk masking your faults.

    Another issue is that of the optimum level of difficulty. I believe the answer is the easiest problems that show up plenty of faults.

  5. I think that problem "speed against accuracy" should be discussed a little bit different. In my opinion the easier the puzzles (tasks) the faster you should solve it. And the second item is behaving the same way: the the easier the puzzles (tasks) the more error-free you should solve it

    That way:
    1) The easy tasks should not be solved longer than 45-60 seconds.
    2) The difficult ones should not be solved fast, but accurate.

    Although you should remeber that if the puzzle is so hard that you cannot solve it in 15 to 20 minutes, you should go to the next one. After some time (1 or 2 days) you should come back to that "devilish" and try once again.

    And what about accuracy level?
    I think that it depends on the set of puzzles: how difficult are they and how many time did you solve (or more precise: see!) them.

    Lets divide puzzles on 5 levels. You should have:
    Puzzles #1: the easiest - minimum 95% of correct solutions
    Puzzles #1: easy - 90% of correct solutions
    Puzzles #1: medium - 85% of correct solutions
    Puzzles #1: hard (difficult) - 80% of correct solutions
    Puzzles #1: very hard (difficult) - 75% of correct solutions

    And after finishing some parts of the puzzles (like 100 or chapter or a whole workbook) you should come back to the puzzles that gives you problems and/or you made them wrong. Just try to solve them once again and try to find out why you did not succeed for the first time.

    I think it should help. What do you think about this "system"?