I'm surprised no one mentioned @boris-trapsky which tries to predict the player and set traps based on that, iirc
@cosmon said in #29:
> Nice article! How do you control Stockfished depth?
I run Stockfish locally and then you can limit the search to a specific depth or just print out the evaluations at the previous depths.
this was the best blog post so far in the world
@Toscani said in #30:
>
how can you be so sure. can you make a definition that would correspond to all those that you consider traps and non-traps.
is it not dependent on appropriate player side limited learned state (or chess vision state). I would think that one might consider a pure board defition that would be independent of players, that would be assuming perfect players (they know all of chess, better than masters or even best engine ecosystem gladiator, the "42" engine).
a sacrifice of a piece, to be called sacrifice in human cultural radar means that it would appear interesting to capture on the side capturing it, but then committing that move enables another stinging tactics from the side offering it, with possibly variable depth to witness such sting. often most visible is when matierial loss is opening positional combination onto a "big-whale" mate that trumps all couting shenanigans.
a piece being geomtrically trapped is not what this concept of trap is, I forgot about that more general language use of the word as adjective. It is about immediate static vision, direct applicatiohn of the mobiltiy ruleset.
I think that the traps on this blog radar are about dynamical ones.. (in which the poisoned pawn, could, leading to a board-trapped queen, become a trap, but if we already have experienced such cases and learned to generalize it bettter and better, would those still be traps or just chess logic internalized, and now part of the evovled chess vision).
isn't it obvious that a complete begginner thre are more traps to be perceived from that unskilled and ignorant leanring initial condition (of their chess world internal represenation I would posiit, respecting my weirdness).
And sure one can end up squeezing an opponent, into no good moves left. how is that necessarily a trap. If one looks at NAMED traps in knowledge bases, as far as I have be exposed to, and with my fallible memory and internal representation of language and boards so far, traps have propreties that seem to me to be about human chess vision foresight limitations.
so you say there are not hidden traps and their are hidden traps. In perfect chess there are no hidden traps I say. or the players are not perfect. Am I mistaken? about the concept of traps.
When in endgame problems we can narrow down all the moves and their continuations that is exactly what perfect chess is.. for those mini-games humans can be perfect chess players. for example we can define in absolute what a zugzwang is.
That is why I was asking if there were NAMED traps (worthy of cultural sharing so far in recorded history of human population sampling of chess world) in endgame territory.
Local optimum versus global optimum search:
it does depend on what we consider local and what we consider global, but those can be parametrized as in the blog as modeled by chess search iterative deepening Lichess black box being opening on local blogger PC.
And then we, non engine development specialists but chess board critical thinking able debaters can then have an augmented discussion about chess concepts making those apparent black box tools palatable and accessible to our language communication, without needing ourselves to become engine developers.
This is because that technology is based on mathematical models having been posited about how to explore chess world.
There are other such models efforts that are still hidden under the hood.. We do not all have the resources to go local PC though. Thanks for the blog author for opening our user side blinders.
@jk_182 said in #32:
> I run Stockfish locally and then you can limit the search to a specific depth or just print out the evaluations at the previous depths.
Can you explain me how do you do it please? :)
Wow
@cosmon said in #36:
> Can you explain me how do you do it please? :)
you would need to download stockfish executable, and run it command line or if some GUI allows you to control depth in human range than use that GUI. It would also need that executable. Then using your OS command line.
search google "stockfish". or wait for someone here to post the links.
You could also ask lichess to change the user parameters from "time to compute" to godepth which is a command line UCI command. And manually upon chess position of yours look at certain amount of multiple PVs (we can do 5 max here, but I would consider even 3), and see how playing with the godepth (instead of 4s minimym time to compute) from low to high, how the ranking of those change. But we can't do that.
Very interesting post! It's interesting to think about a trap as sort of the opposite of a brilliant move based on your model: Brilliant moves look bad at low depth, but improve with a deeper look, while traps are the converse. I don't know if there is much work about how we subjectively label certain moves as trap-like, but I think this could be a neat question.
@dboing said in #35:
> Local optimum versus global optimum search:
> it does depend on what we consider local and what we consider global, but those can be parametrized as in the blog as modeled by chess search iterative deepening Lichess black box being opening on local blogger PC.
Erratum: opened, not "opening" on local PC of the blog author. And Lichess black box for on Lichess, we can't do that.