Linear vs nonlinear regression doubt


I am currently working on a study about how much time a chess engine should think per move in a chess game. The inputs (known data) would be how much time is left (in seconds), the evaluation of the engine (centipawns) and the move that is being made. My question is: is it a multiple linear regression case or a nonlinear one? In each case, what type of function/approach would be more appropiate?

Sure. Here is an example: the chess engine evaluates a position as 1.54. The remaining time in the game is 347s. And the move to be made is number 34. How much time should the engine spend thinking? Similar if evaluation is 0.4, 8.7... Of course if the evaluation is too low or too high the engine should not have any problem with the game; and thus the time to think could be significantly be reduced. However, if there is time left, it is the beginning of the game and the evaluation is uncertain, the engine should spend more time figuring it out the right move to make.
Hope this helps,


No cake for spunky
This seems almost like a linear programming question. But wouldn't you want other criteria such as if results with different amounts of time lead to better or worse moves? I would think you could run a simulation to determine if a move was better or worse although I guess playing style of an opponent would influence this.