Tactics training I

Aug 6, 2017, 11:01 AM |

Tactics trainer a very good tool !!!

Well the tactics trainer is a good tool to improve. But the rating 'system' that is connected with that trainer isn't. To advance one has to do literally thousands of puzzles at a difficulty that more or less is the same as the skills of the user.
Well my point is that even after 10 or so good solves the rating goes way to much up. As a result the puzzles become more difficult and the average time it takes to solve the puzzles increases. But be honest is ones rating really lets say 60 points higher after solving just 10 puzzles ? I say no , it's luck , better concentration or another factor. But as a result one starts to solve puzzles at a slower pace which is the inherent punishment that's build in the tactics trainer.
As said to advance one has to solve many puzzles. Pattern recognition simply doesn't work with only a few examples. It's like learning a language with only a grammabook. People don't follow rules people follow habits and in a way 'thinking' in chess has to become a automatic. But with the need of doing literraly thousands of puzzles to advance time is on the essence. So I conclude that in my opinion it is better to learn from very easy puzzles (lets say mate in 2) and see that patterns until solving them becomes a 'habit' than solving more advanced tactical positions.

The experiment.

I know a pin , but I also often miss a pin and many other tactical feats. With that in mind I started to solve many (about 1200) mate in 2 puzzles this week. That's on the average below my skill. I tried to solve every puzzle in less than 15sec. The result was very disappointing because even at 4 ply (that's mate in 2) I missed many many solutions. Of cause If took more time I probably could get more correct answers but the whole point of the exercise was it to get familiar with the position and to develop a habit in solving them. Pattern recognition is a fast process and probably the first process that gives results in finding a answer. So in a way I was training pattern recognition..

It worked.

After only 5 days or so I was able to solve many puzzles much faster. The moves started to 'jump of the board'. This is for me the prove that pattern recognition and finding (at least fast finding) a move has a better result with relatively easy puzzles and a large volume.

Tip : Solve many puzzles (100 a day or more) at a difficulty that is (much) lower than your rating and do it fast !!

Associations and language.

For now i'm still at mate in 2 level. There is just too many stuff i'm missing (well hey i'm a 1500). I will do this for weeks making 'hashtags' of boardpositions that is Little stories and concepts that define board positions. After 100 or more puzzles I reread the text and try to remember what I thought (not the position) when playing that position. The point of this is it to develop a vocabulair. Language is the key in playing better but all the concepts like pinning and vorks are just the beginning. for me it's for instance obvious what a shield / happy mate. long distance check. contact check. BFF pieces is. it's my own lingo. Develop your own. In computerterms that Lingo acts as a generalization process and also as a kind of zobrist hash. A hash is nothing more that a key to remember a position. And like a computer cant reconstruct a position from a hash a human doesn't always have to remember the exact position... But that doesnt mean there is a higher chance to find the better move. Well so be short :

Tip : Make associations , invent your own chessterms.


I'm still developing my 'think' system so after a certain amount of time if solving mat in 2 gets really bored I switch to mate in 3 etc... and after that to tactical training. However the problem with tactical training is what to train...


Stockfishes.. (idea)

The idea is it to generate puzzles by letting a strong fish play with a weaker. Don't know how this will work out. but with that method I can control the 'puzzles' more. So I get imperfect moves within a certain opening sequence.

Matpatterns... (idea)

Finding matpatterns isn't a problem. Doing mutations on a pattern , or solving backwards to get a mate is possible.

Classification of problems... (idea)

Developing a generalization of chessmoves notation and using that to search for tactical problem types like Q0+ RR++