There’s a lot to be done to improve metrics for players in Twenty20 cricket. This summer I’ve mostly focused on team-level metrics that give you a sense of the whole game, not about the performance of individual players.
There’s a lot of focus in Twenty20 cricket on a player’s strike rate, and that is undoubtedly important. But it’s also important that a batsman is able to last a reasonable amount of time. Players who hit a six and then get out off their second ball on a regular basis will have an extremely high strike rate, but won’t be of much value to their team. So the ability to stay also has value.
When you consider this point, the old-fashioned batting average (the number of runs scored over the number of times your wicket has been taken) has value. Ideally we’d come up with another metric which can mix together these two simple measures to give a sense of the ability of a batsman to score fast but also stick around long enough to make an impact.
In this post, I’m going to focus on a particular datapoint which I think has value when making assessments of players: when they come into the match.
Not all balls are the same, and not all overs are the same. Generally matches follow a pattern where the number of runs scored speeds up as you head towards the end of the match (barring the loss of significant numbers of wickets). You can see that in this graph:
The powerplay covers the first six overs in Twenty20 matches. During this time teams may only place two fielders outside of the circle which marks out 30 yards from the pitch. Following the powerplay, teams may have up to five fielders outside the circle. This clearly has an impact on the game. While the first over is the lowest-scoring over of the match, runs are scored quite quickly in overs 3-6, before collapsing in over 7. It takes until around over 15 before the batting team usually surpasses the scoring rate of the powerplay, but the average number of runs scored per over exceeds nine runs for the final overs.
Obviously you would expect different performances from batsmen depending on when they enter the match. If a batsman enters earlier in the match, they are not expected to score as quickly, but they have more time to play, while those coming in later are expected to score at a faster rate but may not have time to score as many runs.
The next few charts show the various scoring statistics broken down by the over in which the batsman entered the match. Opening batsmen are judged to have entered in over 0, while those who have come in later have been categorised by matching up their position in the batting order to the fall of wickets of preceding batsmen. I have included the metric for all men’s Twenty20 since mid-2011, as well as just for matches played in the Big Bash League since the first season in 2011/12.
Firstly, median runs scored per batsmen:
The curves are similar, although in the BBL we see more erratic data (likely due to a much smaller sample) and higher scoring for batsmen entering earlier.
Next, the number of balls faced:
Again, the curves are similar although BBL batsmen entering the match between over 5 and over 9 face more balls than the global average.
Next, this shows how often these batsmen get out:
The trend is very clear – batsmen entering in the first fifteen overs have a very high chance of losing their wicket, while those entering in the last five have a better show of surviving.
Finally, let’s take a look at the strike rate. This one surprised me. While, as expected, the average strike rate gradually increases for games played globally, you see a much weaker effect in the BBL.
So what does this all tell us?
Firstly, it is clear that the point in the match that a batsman comes in matters – it’s not reasonable to judge a batsman by the same standards if they open or if they come in late in the innings. Those coming in earlier have longer to bat and score more runs, but those coming in late score faster. Those coming in early are also more likely to lose their wickets. Indeed, you’d expect them to lose their wicket, as not losing their wicket may suggest they have wasted too much of their time at the crease with defensive play. Because of this, it may not be ideal to use a traditional batting average, which rewards batsmen for holding on to their wicket.
My first idea for a simple metric is one which simply takes the median number of runs scored by batsmen entering in that over, and measures the batsmen’s performance as a percentage of that score.
At the moment, I’m thinking I would measure that performance against all cricket played in the previous five years, back to 2011, but that will need to be finessed.
I’m going to come back tomorrow with some tables of the best and worst performing batsmen in the BBL this season according to this measure.
What other metrics would you like to see which adjust based on when a batsman arrives at the crease? I’m open to all suggestions.