Quote:
Originally Posted by davmcg
Think that is rather implied odds?
Overall shouldn't a model of the game be based on defence and attack?
Defence broadly being assessed on the % that when scoring zero, the opponent does not score at their visit.
Attack based on scoring average when in a break.
A final stat could look at whether the opponent scores when the break ends (ie identifying how a player manages risk).
The following stats are compiled by the TV production team:
-Safety %: a non-potting shot which results in no pot success for the opponent;
-Short pot success %: A shot less than 6 feet resulting in a pot;
-Long pot success %: A shot more than 6 feet resulting in a pot.
I don't think these stats are archived anywhere, unfortunately.
I don't believe there is a stat collected for pot attempted and failed yielding opponent's pot success.
Markov chain model can be represented by a state matrix and a transition matrix. The model would then use different transition matrices (based on player skills and available options) to yield the next state. The state matrix should describe the opportunities for potting a red/color, and safeties.
Typically players look at least a few shots ahead, so the state matrix should account for that to some degree.