If I may hazard a non-definitive answer - It sort of depends what you are looking for -
Early in the game I think a lower number of plies is more usable. For one thing, I don't know how it can help me to simulate more plies forward than I would hope to envision over the board, so I tend to go with 2 or 3 plies. At the very beginning of a game it's sometimes interesting to look at a 2 ply sim, and also 3 or 4 plies, to understand what the consequences are of a given play (like finding out whether it's very bad to open a hot spot, immediately or long-term). I usually find out, if I run a large number of plies, that it's not worth it to look so hard - the differences between plays are small. (If it's hard to decide between two moves, it often doesn't matter that much which one you choose.) Other factors (like the random tiles you draw next, and whether you know the bingo that is playable) far, far outweigh the little bit of difference you get by going a huge number of iterations or plies. I have found Quackle useful for understanding which parts of the game I'm more competent at, and which parts I could improve on better. If you use more plies then the simulation will tend to have more 'noise' in it and it will take more iterations before the noise settles down. However as it gets towards the endgame the technicalities of the pool (like if it is unbalanced in some particular way, or if certain tiles are crucial because of particular hooks on the board, or a Q stick etc.) then I find it more reasonable to use many plies, to the end of the game. If I understand Quackle right then the simulation always uses the same leave values regardless of the pool, for the last iteration, on a simulation before the bag is empty. I'm not sure how Quackle evaluates leaves near the end of the game, so I have been meaning to ask that here or to look in the code to see it. I also am wondering how Championship Player does its sims to determine plays. Kevin Leeds. On Mon, Nov 24, 2008 at 10:29 AM, Phil Robertshaw <[EMAIL PROTECTED]> wrote: > I don't sim games as often as I should. I've been experimenting with 4- > ply sims, between 500 and 1000 iterations. Wondering, though, if I > should use 'many' plies, which takes considerably longer. What is the > general combination of plies and iterations which most people use? I > find 4 plies generally gives a pretty good indication of the best > moves, but often disagree with the results. > > Phil R > >
