I broke the problems up into 6 sets of 85 and 1 set of 83. I stuck by my promise to keep each group under 90 problems as anything more than that becomes tedious, especially the first few times through the set.

Set A: Elo 1050 - 1150

Set B: Elo 1150 - 1300

Set C: Elo 1300

Set D: Elo 1300 - 1400

Set E: Elo 1400

Set F: Elo 1400 - 1450

Set G: Elo 1450 - 1500

Here are the charts for my 1st pass through each group and problems completed in <15 seconds for the 1st pass:

The graphs might suggest that there were actually 3 levels of difficulty here: A/B/C, D/E/F and G. Some practice with A+B made me better at C and D+E made me better at F. Or it could mean that their elo ratings for the problems aren't realistic, who knows.

Final chart of all 6 passes completed in < 15 seconds:

And of course I got better as I did the repetitions. My final success rate was 89.54% which puts me about the same as Polgar and Heisman. However I caught myself making a mistake. My 6th pass was supposed to be 13 days after pass 5 and I was mistakenly doing it after 21 days for the first four sets. Once I corrected this, my 6th pass stats generally ended up being close to my 5th pass. That may have cost me a few percentage points, which suggests I might have been slightly better at this than Heisman and Polgar. My retention for problems could be improving over time with practice.

Up next is the final book that Bright Knight used for his speed training: Jeff Coakley - Winning Chess Exercises for Kids. 900 problems and I've divided them into 10 groups. If I stay on schedule, I should finish around the beginning of December... which will mark my 1 year anniversary of my chess training journey. Kind of fitting.

That repetition of the same set of problems will improve your speed at this set is ..eh..not astonishing. So whats about the "transfer" to problems you never have seen before? Did your rating at chesstempo or any other Tacticserver improve?

ReplyDeleteI think your idea of repeating the problem set is good. I am not supportive of your idea of using less than 15 seconds. I understand every player has a method for success and not every player needs/ should follow the same method. But, from my understanding, it helps to check a position twice before submitting it to play. To check the position twice, you need more time OTB. Habits from practice spill into the actual game. If you are using less time for practice, how do you force your brain to use more time OTB? One of the things I force my students is to use a minimum time (of 1 minute or so) for every problem regardless of whether the problem is easy/ solved/ etc. Reason is, I want them to use time to practice to check, recheck, recheck and again recheck.

ReplyDeleteWatch this video of Carlsen:

https://www.youtube.com/watch?v=PZFS0kewLRQ

I have heard him say this multiple times during different times as well. He basically says he gets the answer to every position/ problem right after he looks at it, within a very short time. Ofcourse, he does. But, he still spends lot of time to confirm the move by checking many times. I think he is right. It is all about reducing that small margin of error by checking multiple times. It takes longer than 15 seconds for sure to check multiple times.

May be I am wrong, but offer me a counter reasoning....

That is an idea that worried me time ago, but as Laurent points out, i think both methods are jsut two ways of trainning different skills. Anyway i think that starting with chess with some clear goals, it makes sense to start by storing a los of patterns before you get to develop your calculation abilities. I think it would make little sense in real otb to confront low pattern recognition vs strong calculation ability...the second would be od little use when asisted by a low #of patterns already stored in your brain.

Deletei worked hard with chessimo for about a year, and once i thought i had grown up a little my pattern recognition, i started to strech a biy my claculation powers by doing more complex problems with no time limit etc, basically wuith chesstempo,; there you can work pattern recognition along with just "pure" calculation puzzle, like mate in 8, where the solution is not necesarily linked to the ability of "spoting the right weird move" but just claculating different lines and sub-lines.

To sum up my point: even to be able to improve at calculation, yopu already need a strong set of patterns stored in your brain, as other way you would be calculating over unlimited number of different branches not being able to efficiently order them upo and selecting what your intution (apttern recognition) tells you is worth enough calculating...

Hi SVC,

ReplyDeleteI think the time needed depends on the type of problem you're trying to solve. Of course, OTB you'll very rarely meet a tactic in the 'pure' form of training. Besides, OTB, you're never sure the tactic is going to work.

I think it makes sense to separate pattern recognition training (what SK does), from calculation training which would be solving more complex positions, closer to what you can find in real-life. The second would certainly require more time than the first, and more 'blunder-checking'. As long as you're aware that both situations are different, I think it's quite okay to try and improve your speed in pattern recognition training. From his other blog entries, it's quite clear that SK isn't the type to rush in OTB situations requiring calculation.

I would say I do both. Depending on the problem, I may spend 2-3 passes spending up to 6 minutes trying to arrive at the solution. That's what the 1st chart reflects. My range of times on each set on the 1st pass. You can see that I spend at least a minute on ~50% of problems my first time through.

DeleteThe 2nd chart shows how many I did in under 15 seconds on the 1st pass. Those 20 - 30% of problems reflect my initial pattern recognition on a new set of problems. The % of problems I'm able to complete that fast seems to be pretty static with each tactics set. But the difficulty of Ivaschenko, compared to say Bain is much different. 30% of problems in Ivaschenko would be the equivalent of about 70-80% in Bain.

Chart 3 shows my progress from calculation to recognition over the course of 2 months.

The chess training paper I linked a few months ago didn't recommend training "calculation" as a means to get better at calculating. Their theory was that being able to calculate accurately and deeply required good pattern recognition to eliminate bad moves. You don't get there by brute force.

I think it's possible that if you play lots of slow games, that's generally sufficient for calculation training without the need to train it outside of games.

Very good discussion, guys. The summary of all your posts mean that you play both style, speed and slow practice which means you are looking for pattern recognition as well as accuracy in the moves. It is difficult to do both, but not impossible. Good luck to all of you!

ReplyDeletechesstempo does have a mixed mode. At this mode you have minimum 5 min for your problem . Only if you spend more time than 5 min then your score is reduced in the case you solve the problem correct. Still such a mode can measure your skills, its a timed mode.

ReplyDeleteI have solved Ivashenko's books (1a+1b, in one volume name as "volume 1"). I have done it in two ways:

ReplyDeleteA) to solve all the puzzles as fast as possible (to remind myself what was missing)

B) to solve the puzzles with the average speed and STOP when I see something important and interesting one.

Sometimes the position requires to dedicated just 5-10 seconds, but other time I think over the position even up to 3-5 minutes. The more interesting the position the more time I decidate to this. However I have no problem to solve some positions very fast (up to 5 seconds).