The scientific approach to the game of backgammon. Part 5.

The non-deterministic same games appear more the factor that we do not know exactly how the rest of the game in the future, that is, the game does not affect some of our best reasons. It is a testimony of cubes (when we can not predict what will fall out in advance), or some other random factors. In our case, these random factors, namely the testimony of cubes - a finite number of options, a few combinations. We review each combination in the game and see how the game will develop if we dropped something such glasses or other glasses, for each combination that ... AG. But this increases the number of variations in the progression ... AR. Yes, there appear more ... BM. And not only increase the number of variants of the game, moreover, it is not clear what is to use algorithms, and algorithms exist for this (I'm going back to Berliner) different approaches.The first approach - it's just a random simulation of several branches of the position, more precisely - the development of the game thread. This is one version of the program. But it's been a long time, it is the first program to play, dated about 80 mi, maybe 90 meters a year, but not later. And after that all the programs - everything, I do not know of any exceptions among the good programs for games, except ours - are written in so-called neural network technology.And then at the same time I became myself an expert in the game of backgammon. I realized that the kind of backgammon games, no less than the left and right brain is used, then there are some things quite impossible to explain - why one item in the game is better than another, that is only possible, as I jokingly say, right hemisphere explanation. But now, it may be wise to go to what is the right hemisphere algoritmizuetsya work, and that is reflected in the programs and games, and other tasks deskretnoy optimization - a dynamic assessment of the position, even better, use a dynamically generated risk functions. About static evaluation of the game, I have already spoken briefly. In nondeterministic games, thanks to this indeterminacy, we do not know exactly what we would get, and we sort out all sorts of random outcomes. We fell testimony cubes are something we get such a forecast, the next - the next forecast. So, we have for each outcome of a random event, we have some fun collection, some built of static estimates. And on this we selected different types of risk functions. Already here we have almost come close to "Geli-Fishu". If we now win it (I repeat, our Russian version) somewhere 55 percent, maybe a little more, then we lost the same amount. But that was good. Winning at decreasing the risk functions. Decreasing - it means that we have at least a little, but the pessimists.