Quote:

Original post by LonelyStar ... Next Window: 0 to 1 Alpha beta returns 1 Next window 1 to 2 Alpha beta retusn 2 Next window 2 to 3 ... you get the Idea. I know, that the correct value at that point is 0!

Quote:

Original post by LonelyStar The AlphaBeta function never returns anything outside the boundings of alpha-beta, in case of MTD(f, this means that everytime either beta-1 or beta is returned. I know, that the firstguess should be s accurate as possible, but of course I can not know an absolute accurate MiniMax value at that time! But what if my first guess is off by about 500 (500 means a small capture for my evaluation function). Than, (assuming the correct value is 0) MTD(f) first uses the window: 499-500 than 498-499 e.t.c. So it takes 500 passes! On http://www.cs.vu.nl/~aske/mtdf.html it says, it should take about 5 to 15 passes. I seem to have totaly missed something! Can somebody explain a little to me? Thanks!

// ... for every move do ...{    // do the move    val=-NegaMax(-beta,-alpha);    // undo move    if (val>alpha)    {	alpha=val;            // a leaf value > beta is obviously > alpha                              // alpha will take this greater value	if (alpha>=beta)      	{	    break;            // the loop broken	}    }} // ... end of loopreturn alpha;                 // and this gretaer value returned to main caller