Patricia 4 game_ply calculation incorrect causing perfect play with UCI_Elo set

[kyleaforrester]

Patricia 4 will only play a sub-optimal move if the thread_info.cp_accum_loss variable is filled up. Unfortunately, (thread_info.game_ply > 5) is a condition required to fill up that variable. Depending on how the position is fed to Patricia 4, the game_ply may not be correct. For example, take this position:

position fen rnb5/4k1r1/p1p1p2Q/1q2P1p1/4B3/2P1B2P/1P3PP1/R3R1K1 b - - 0 18 moves e7f8 h6h8 f8f7 a1d1 b8d7

Patricia 4 will not refill the cp_accum_loss variable for this position because the game_ply is '5' in this instance, which is incorrect (should be 23). Patricia 4 needs to be smarter about how to calculate the game_ply so will not play perfectly depending on how the UCI position is fed to the engine.

[Adam-Kulju]

Hi, thank you for your feedback. The idea is that Patricia should only start making mistakes a few moves after you start playing it, to avoid immediate errors and non-human play (if you allowed her to immediately make mistakes, low ELO Patricias would start off with 1.Nh3 which is not really what we want). If you want to play it from a custom position such as the one you gave, it is a bit of a different story but I think you would want her to play the most critical continuation and not immediately give the game away - this holds especially true for analysis.