G/x

[Gábor Szõts]

Geschrieben von: / Posted by: Gábor Szõts at 05 June 2000 09:08:42:
Hello everybody,
Now that I have played a few tournaments with a wide variety of engines I hazard the opinion that it is a mistake to use time controls of a fixed amount of time. Exception is perhaps blitz (e.g. G/5).
Why? Initially, the engine doesn't know how many moves it will have to make. If it uses too much time at the beginning phase of the game, it may get into time trouble later, spoiling even a clearly winning position, which position could also have possibly been reached using less time.
On the other hand, if the engine uses little time for the beginning phase, it may reach a bad position which cannot be saved even having a lot of time remaining.
The method may favour some programs, while it can be disadvantageous for others. The longer the allotted time, the greater the danger of distorted results. In a lot of cases the outcome may be completely random.
Why do we need this? Using G/x can only be justified by arguing that we can finish the game within a given time.
But is that necessarily the only way to achieve this goal? I think not. Suppose we have two hours for a game. Instead of choosing G/60, we can assume the game will take 60 moves on average. Now choosing a time control of e.g. 20 moves in 15 minutes, we can play a 80-move game, and in the rare cases where the game is not yet finished we can use 'arbiter's decision' to determine the result. This practice had been used for human games to avoid adjourning games.
Using x moves / y minutes the engine will know how much time it has, and it depends only on its 'skill' how it manages its time.
For the above reasons, I suggest using a time control of 20 moves in 20 minutes for the 'Battle of Crowns' championships. Here the danger of overtiming is negligible since 4 games are played consecutively.
Best wishes,
Gábor