29,90 €
The neural engine Fat Fritz which is supplied with Fritz 17 is based on the Open Source Project LCZero. However the network was not simply trained by playing games against itself, (the “Zero” concept), but also by using strong GM and computer games. This leads to surprising playing strength.
The new engine option “Display win probability” is pre-installed. A neural engine which is playing according to Monte Carlo techniques gives evaluations of positions which are naturally in the form of the probability of a win.
The interpretation of this is illustrated in the following example.
Let us take the following display and a question by a user on the subject.
We find in the notation 1.e4 c5 54% /3 3 2.Nf3 4 Nc6 54% /4 3. What is the significance of the numbers 3 3 4 4 3 over and above the 54% and does the 54% mean that after these two moves in the game White has in each case 54% probability that he will win and thus a 46% probability that he will lose?
If that is the case: how great then is the probability of a draw (drawing ratio)?? This also should be taken into account, since in chess a draw is also an important factor?
Answer: the interpretation of the probability of a win in the present case looks like this:
Yes, White wins with a probability of 18%. White wins with a probability of +10%, there is a draw with the value =16% and there is a probability of -74% of White losing. A specific calculation is not necessary, since the solution is displayed totally clearly with 18%.
As to the question of the values shown with the notation.
In addition to the percentage probability of a win we have displayed – as is the case in the classical evaluation display – the depth of the search and the time taken.
A mathematical formula for recalculation into the classical evaluation display is also unnecessary.
If you rightclick the engine window and deactivate the entry for “Display win probability”, then the evaluation of the chess engine is displayed as is the case with the other engines as a classical evaluation display!
Back to List
to send a comment please fill the form:
Name *
Email Address *
Subject
Feedback *
