#10 It's worth noting that right now 22411 has 178 karma: downvotes are absorbed & overwhelmed by upvotes. If players correctly solve puzzles, how can they know what alternative moves would have been rejected? They can't, and yet they upvote anyway.
I always downvote a problem which i did not solve. I get 2 points for solving a problem and i lose 30 points for not solving one. I want to prevent others from experiencing this frustration.
The problem is that i lost to an algorithm and i can not argue with the algorithm. I consider myself smarter than algorithms.
Another problem is that many of these puzzles are boring and sometimes also ambigous, if not incorrect.
Puzzles should be selected/created and peer reviewed for correctness by humans before they are published. Puzzles should also not contain ambiguities. Puzzles should also not contain redundant pieces (I dispute if redundancy and ambiguity has any training effect, i believe it is just visual noise). And Puzzles should be inspiring and not boring.
Thats all stuff algorithms cant handle. Making good puzzles is an art.
---
ICC has a good human generated puzzles collection. I remember having lots of fun solving these. I liked most a puzzle collection with double rooks which was very simple, no redundancies, and where the position just changed a bit and then there was a completely new solution.
The problem is that i lost to an algorithm and i can not argue with the algorithm. I consider myself smarter than algorithms.
Another problem is that many of these puzzles are boring and sometimes also ambigous, if not incorrect.
Puzzles should be selected/created and peer reviewed for correctness by humans before they are published. Puzzles should also not contain ambiguities. Puzzles should also not contain redundant pieces (I dispute if redundancy and ambiguity has any training effect, i believe it is just visual noise). And Puzzles should be inspiring and not boring.
Thats all stuff algorithms cant handle. Making good puzzles is an art.
---
ICC has a good human generated puzzles collection. I remember having lots of fun solving these. I liked most a puzzle collection with double rooks which was very simple, no redundancies, and where the position just changed a bit and then there was a completely new solution.
#12 You're correct that many of the best puzzles are artistic and algorithms can't handle that (since an automated test for art does not exist).
To the extent it is possible to define rules for testing whether a puzzle is good, it is possible to automatically generate possible. However, algorithms such as this are computationally expensive:
1. Randomly create a puzzle
2. Test if it is a correct and good puzzle. If so, submit it; else, repeat steps 1-2.
IMHO good puzzles have multiple solutions (or "ambiguities") since real chess games have multiple good moves, although some solutions may be better than others.
To the extent it is possible to define rules for testing whether a puzzle is good, it is possible to automatically generate possible. However, algorithms such as this are computationally expensive:
1. Randomly create a puzzle
2. Test if it is a correct and good puzzle. If so, submit it; else, repeat steps 1-2.
IMHO good puzzles have multiple solutions (or "ambiguities") since real chess games have multiple good moves, although some solutions may be better than others.
In the same category as puzzle 22411 is puzzle 26729. I failed the puzzle with the same moves I used to easily (even with some inaccuracies) crush the level 8 AI:
http://en.lichess.org/UfUsmNl4
http://en.lichess.org/UfUsmNl4
This topic has been archived and can no longer be replied to.