| |
BGonline.org Forums
Petanque
Posted By: Michael Petch In Response To: Petanque (Tom Keith)
Date: Friday, 6 May 2011, at 4:42 p.m.
"I think it is possible to program a neural net to learn curling strategy. The input is basically the position of the rocks in play, the number of rocks yet to be played, and the score of the game."
Having curled I'd say it isn't so simple. Ice surface, conditions, humidity, frosting, type of brooms be used (Some brooms like corn can be used to drop more debris increasing the chances of a rock being picked - something Kevin Martin was notorious for at the world championships when he was young). The ice at a championship is far different from that of typical "club ice".
As well the ice conditions do not remain the same from the beginning to end, and because of this curlers have to take varying amounts of ice at different times. This means that a shot that may be possible in end 1 is not possible in end 10, or can be made but the chances of success change.
Of course the current score, how many ends are left, and who has hammer will play into strategy within a particular end as well.
I'm not saying curling can't be modeled but the inputs into a neural net I believe are more complex than suggested because ice isn't an entirely predictable element.
One other thing, it can also depend on the physical strength of the players. Women players generally can't throw the same up weight as men do (Not a sexist remark). This alters strategy. Some men can make take out shots with up weight that women can't so the strategy between men and women is often different. Watching the Mens and Womens world championships or even looking at some of the best Women's and Men's teams in Canada at the Scotties or Brier events can give you an idea just how different the styles can be.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.