The state of any cricket match can be described in terms of six values - runs, wickets, and deliveries for the two teams. A team's rating at a given point in time refers to its strength at that point in time. Calculating ratings requires answering two questions. First, how is each match to be measured? Second, which matches are to be considered relevant at a given point in time?.
1. How is each match to be measured?
There are many ways to answer this question. One simple way (such as the one used by the ICC) is to assign predetermined points for a type of result, regardless of the margin by which the result is achieved. The method described here is designed to account for margin of victory.
For any cricket match between two teams A and B, we have
aruns, awkts, abf
bruns, bwkts, bbf
runs per wicket (rpw) = (aruns+bruns)/(awkts+bwkts).
This provides the cost of each wicket for the match. The points earned by each team is given by:
A points = aruns/abf + rpw*bwkts/bbf
B points = bruns/bbf + rpw*awkts/abf
A win bonus (wb) is available for each match. It is added to the winning team's points tally. This is given by:
wb = (A points + B points)/2
If A wins the game, A's points tally will be (A points + wb). Otherwise, it will be (A points)
This method is used to rate each team's performance in each match.
2. Which matches are to be considered relevant at a given point in time?
This is a vexed question. The ICC includes matches within the previous 2 years, and the matches in the two years prior to that with a 50% weight. The year always begins in May. The method used here does not specify a predetermined period of relevance.
Each player for each of the two teams A and B, has played matches before (unless the player is on debut). The rating for each team at the start of each match is simply the average of the team's rating in all the Test matches played by all the players involved in the upcoming match. Matches featuring all eleven players involved in the current match are weighed most heavily (since these matches are considered eleven times), while matches involving only one of the eleven players are weighed least heavily (these matches are considered exactly once). Matches which do not feature any of the eleven players in the current match are not considered at all. This method provides a natural decay which relies on the idea that teams are selected to win games.
This provides a match rating for each team at the start of each match. It requires knowing the playing elevent for each match and does not involve any other arbitrary assumptions. See the match wise team ratings for India. A twenty match rolling average is also given.