In the last cricket season, I have been working on better ways to analyse cricket team performance beyond the results, scores and averages, especially at club and school level where the scorecards are often the only source of information.
The motivation behind this project is the too basic traditional way of analysing performance. It doesn’t allow for spotting strengths and weaknesses in a team. The result might tell you how close the game was, but not where it was won and lost. Averages only make sense if you know the context of the matches, but these contexts change every game from rain affected over reductions, to pitch quality and opposition ability. If there is video analysis – which is rare anyway – it often focuses purely on the technical and not on the outcomes.
So, my mission was to come up with a way of overcoming these obstacles to provide a reliable measure of performance in cricket batting and bowling. The goal was to give players, coaches and captains something to work with when planning practice and tactics.
In the next few articles I will show you this method and provide some real-world examples from teams I have analysed in 2019. The first step is a more useful context.
Par, Target and Expected Scores
Context is important in matches. It’s unhelpful comparing the average scores in ODI cricket to, say, U16 regional cricket in Scotland. Scores are too different despite the matches both being 50 overs.
That’s where Par scores become useful.
I examined historical scores in several competitions including the WDCU Premier Division (WDCU) and came up with an average score for the competition, this is the Par score. For WDCU this is 189. This provided a touchstone context for scores: We know 189 is the average score between two average teams.
To create an even more granular context, we can weight Par by the ground the match was played.
For example, games at Clydesdale average 225, while games at Uddingston average 145. After weighting the Par scores are 197 and 159. So, we can be confident – based on scores from 187 matches over three seasons – where the average lies on any given ground, no matter which two teams are playing on it. If a team bats first at Ayr, for example, we know they have an even chance of winning by scoring 188.
There also will be occasions to know the Par wickets, which uses the same method but for wickets instead of runs. Here’s what that looks like:
Naturally, this only applies to batting first in limited over matches. When batting second, the score is known. This score plus one becomes the Target and is no different to usual.
Expected Runs and Wickets
At this point we know what a team needs to do, but what about bowlers and batsmen?
This depends on the information we have available, but even at the crudest level we can calculate accurately what an average player is expected to do. This is Expected Runs (xR) - conceded for bowlers and scored for batsmen - and Expected Wickets (xW).
The more runs a batsman scores above xR, the better they have done. The fewer below xR the worse they have done. The same applies for bowlers runs and wickets. This is more useful as a measure than batting average, which needs a few games to be reliable and even then, is only “the more the better”.
Calculating xR and xW is from scorecards alone has issues, but can be done to create a ball-park figure in the following areas:
Batting first xR using Par score and average wickets.
Batting second xR using Target score
Bowling first xR using Par score and overs bowled
Bowling second xR using Target score and overs bowled
Bowling first xW using Par wickets and overs bowled
Bowling second xW using Target wickets and overs bowled
For the bowlers, we can create Expected Average, Economy and Strike Rate (xA, xE xSR) and compare it to actual performance.
So for example, the xR for batsmen batting first at Ayr is 23.5. While a bowler who bowls 10 overs would expect figures of 2-38 (actually 1.6 wickets and 37.6 runs but rounded).
If we have access to balls faced by the batsmen we can also create Expected Strike Rate for the batsmen (xSR). Although not all cards offer the information so it can’t always be factored in.
There there is a problem with this limited method. It does not account for the stage of the game, such as batting or bowling at the death with little time left. So a bowler at the death is less likely to reach xR and more likely to reach xW. A batsman coming in in the last two overs has little hope of reaching xR (although this latter point can be overcome if we know balls faced).
These are limitations of the method enforced by only having scorecards and not scorebook information.
To deal with this, we can produce Expected scores using the DLS calculator, but it requires more information than the average scorecard gives us. However, if you have access to both the DLS calculator and over by over score updates - like in a scorebook - you can get a more accurate Expected score.
To do this we need to compare the score at the end of any over with the DLS Par for rain affected matches. When batting second, the DLS calculator can be used as normal to generate the table (regardless of the weather). When batting first, the table can be produced using the Par score. From here, we can easily see how far the batting team total, and individual bowlers and batsmen all are ahead or behind the expected score. It is both more accurate, and takes into account the stage of the game and the wickets fallen; solving the above issue.
Once we have this basic framework information, we can start to create a more useful metric for players and coaches that the averages. Click the links below for articles on each one: