Version: 3.5 (using KDE 4.6.0) OS: Linux In the attached game spider_solver_broken is winnable but spider_solver_broken_unwinnable according to the solver. There is a legal move to move between these two positions. Reproducible: Always Steps to Reproduce: Load these games - the solver predicts one is solvable and one isn't but there is a legal move between them (and the game is indeed solvable from either position). Actual Results: Solver gets it wrong Expected Results: Solver should get it right OS: Linux (x86_64) release 2.6.37.6-0.7-desktop Compiler: gcc
Created attachment 63395 [details] kpat "unwinnable" position (which is winnable)
Created attachment 63764 [details] kpat spider position winnable
Created attachment 63765 [details] unwinnable position one reversible move from the winnable position
Created attachment 63766 [details] the orignal winnable position for go with the orignal unwinnable example
Created attachment 63793 [details] unwinnable position
Created attachment 63794 [details] winnable position one reversible move from unwinnable2
Thanks for all the example save games. I can reproduce with the latest version of KPat, with the exception of the second save game which is labelled unwinnable here.
Created attachment 67141 [details] another unwinnable game convertible to winnable. A game labeled "unwinnable"; moving the 4spades to 5clubs makes it marked winnable.
Created attachment 82456 [details] savegame 897321213 with problem demonstration This game marked as winnable. If 2 lowest cards from 2nd from right column moved to 4th from right colum, it changes to no longer winable. But this move is reversible. Therefore some of them is wrong. I don't know which one.
Spider is a very painful game to solve - way too many possible moves at any given time. So the solver cuts out some of these possibilities - and sometimes it's wrong.
Don't take my word for it :) https://arxiv.org/ftp/arxiv/papers/1110/1110.1052.pdf
https://invent.kde.org/games/kpat/-/merge_requests/43 - I replaced the spider solver now, but I need better spider experts than I am to try it. Because a lot of the save games attached in here aren't solvable and I don't know if I'm just too stupid to find the solution or if the solver is real (it's very possible that the savegame of 2013 is interpreted differently now). Any volunteers?
I'm sure there now other situations where the solver will pick something wrong, but I need new data to work on this. So this bug is fixed