Monday, November 14, 2022

Frequency of occurrence

 In order to maximize the effect of my training, I look for the tactical motifs with the highest frequency of occurrence. I make use of ChessTempo, but its label system is rather dodgy. Sorry for that. Dutch and English labels are mixed, for instance.

I use problems of just 3 plies deep, since they show the clearest motifs.

Mate:

  1. edge
  2. lawn mower
  3. Damiano bishop
  4. swallow's tail
  5. dovetail
  6. pawn mate
  7. opera
  8. suffocation
  9. back rank
These 9 mates cover 70% of the mates that occur in practice.

Motifs:
  1. fork
  2. multiple attack
  3. tellen (counting)
  4. verdedigende zet (defensive move)
  5. aftrekaanval (discovered attack)
  6. removing the guard
  7. tussenzet (in between move)
  8. verre pion (remote passer)
  9. wegjagen (chasing away)
  10. gain of tempo
  11. offer (sacrifice)
  12. penning (pin)
  13. discovery
  14. ingesloten stuk (trapped piece)
  15. overbelasting (overloading)
  16. röntgen
  17. veld vrijmaken (square clearance)
  18. verdediger slaan (capture the guard)
  19. zwakke achterste rij (weak back rank)
  20. desperado
  21. lokken (luring)
  22. onderbreken (interposition)
  23. stille zet (quiet move)
These motifs cover 80% of the tactical motifs that occur in practice.

2 comments:

  1. I'm curious: are these two lists ranked in order of frequency of occurrence?

    I collected data from lichess and ChessTempo on the number of puzzles tagged, using the specific tags officially supported on each site. Prior to reaching any conclusions, I realized that the number of puzzles chosen for each tagged category MAY NOT be indicative of the frequency of occurrence in games.

    (1) Do the tags represent some objective (independent) finding based on a search of games or do the tags just reflect the subjective categorization by those who tagged them? That question came to mind when I saw the comments regarding justification for the tag(s) chosen rather than clear-cut evidence of a small concise set of tags.

    (2) Do the tags reflect the reality of different rating classes? I strongly suspect that the tactics and checkmates that occur most frequently varies widely across rating classes. I would expect that explicit examples of basic tactics would occur on the lower end of the rating continuum, whereas those same tactics would NOT appear explicitly (occurring only as possibilities that were avoided rather than as moves actually played) with the same frequency at the higher end of the rating scale.

    (3) Are the programs used to search and find tactics and mates capable of finding all instances of the various tagged categories? Given the difficulty in distinguishing between actual tactics/mates played and potential tactics/mates on a "pieces-on-squares" basis, as a former programmer/analyst, I am somewhat skeptical.

    Obviously, prioritizing learning of the statistically important (most frequently occurring) tactics and mates is a laudable and "common sense" approach toward focusing and reducing the effort required to reach mastery of these essential elements.

    Some less-than-perfect selection criteria (SWAG) seems significantly better than random guessing (WAG).

    ReplyDelete
    Replies
    1. The lists are ranked in frequency of occurrence of the tags. Motifs with less than 700 tags are not included.

      Ad 1. The label system is subjective. What to make of a problem with 15 upvoters for a suffocation mate and 14 down voters? I don't care too much.

      Ad 2. The whole rating spectrum is comprised. But since I select only problems of 3 plies deep, the average rating is low. Usually somewhere between 1000 and 1300. I don't care too much. Anything of that rating level that I don't see a tempo is worth learning. It is about absorbing the endpoints.

      Ad 3. I don't know. Chess Tempo used as a criterium a change of 2 pawn points by a single move to judge where there is a combination involved, if I'm not mistaken. Maybe Aox knows more about this.

      The system rapports 8304 problems labeled as "fork" and 4 problems labeled as "reloader". So it helps to decide on which to spend my time.

      Delete