Sunday, 4 August 2013

Is Brendan Taylor doing too much?

Happier times for Brendan Taylor: captain, but not keeper
Zimbabwe have just come off a convincing 5-0 series defeat to India. While that result was not particularly surprising, the lack of runs from Brendan Taylor, normally their best batsman, was a surprise.

Some of that may have been due to India having done their homework on him. Duncan Fletcher is a clever coach, and will have been aware that Taylor is capable of winning games almost single handedly with a big innings. In 128 innings against good opponents (test teams or Ireland) he has 6 ODI hundreds. To put that in context, at a similar point in his career, Jayasuriya had 5 ODI hundreds. It would have been, therefore, negligent of India to have not had specific plans for Taylor.

But as much as that may have been a factor in this series (and also potentially in the last two against West Indies and Bangladesh where Taylor was also ineffective) there was possibly something else going on. In the last few matches he has been keeping wickets and captaining the team. Some players thrive when given the gloves and the captaincy, but most players find their batting suffers when they are both captain and keeper.

Throughout Taylor's career when he hasn't been both keeper and captain, he averages 32.66. However, when he is doing both his average drops to 14. Below are some tables showing his respective statistics as keeper/captain. I've included the two most familiar (average and strike rate) and also the batting index, which combines the two. (Some context for batting index, Kallis has an ODI index of 33.03, Tendulkar 38.66 and Kohli 43.08)

AverageNot KeeperKeeper
Not Captain28.4231.48
Strike RateNot KeeperKeeper
Not Captain66.3365.81
Batting IndexNot KeeperKeeper
Not Captain18.8520.72

We can see that when he was not captain, keeping didn't impact on his batting much at all. The difference between an index of 18.85 and 20.72 is negligible. However, when he is captain, the difference is astounding. Tempering that, however was that the number of matches where he was captain wasn't really enough to be confident about the statistics just based on average and strike rate.

I really wanted to know if the difference was significant enough to to say that Taylor should give up the gloves. To do this I used a technique called boot-strapping. This is a modern statistical method, that's useful for situations where there is not large samples, and/or they are not normally distributed. It was first developed in the late 1970's and has only really obtained prominence in the past 20 years. As a result bootstrap analysis has not been regularly used in cricket statistics, despite it actually being a very appropriate method. For more info on bootstrapping see the Wikipedia article.

I first used a statistical package that is designed for use in NZ schools, and that does most of the work for me (iNZight), and compared Taylor as captain but not keeper with all of Taylor's other innings against the other sides in the top 11. As it was a general stats program, I had to ignore not outs, and just find the difference in means. I thought about looking at medians, but the mean gives a bonus for making the most of a start, and that's a significant thing in cricket. The graph of the analysis looks like this:

The numbers in red are the confidence interval of the difference. In other words this is saying that if Taylor didn't have the gloves, we could expect him to have a mean score of between 1.84 and 48.97 higher than he has with the gloves.

This is a very wide range, because we do not have a large sample of innings where he has been captain and not keeper, but despite this, we can see that we can be confident that there will be an improvement in his batting, even if that improvement is only small. The middle number is the best estimate based on the data that we have. In this case we can expect that if he didn't keep wickets, that he'd score roughly 25 runs more per innings.

However there's something not quite right about using a mean to analyse cricket scores, so I then quickly wrote a macro in Excel to generate something similar, but this time comparing batting averages. Here the interval went from 0 to 81, with a best estimate of 29. This was wider because not outs impact significantly on the variability of the data. However again we see that there is no negative numbers in the interval, meaning that we can be confident that Taylor's batting would improve if he took off the gloves.

The only reason left for giving Taylor the gloves would be if there was no other keepers that were up to standard. At that point the decision could be that it was fine to lose their best batsman, because his role as a keeper was more important. However Zimbabwe have Regis Chakabva waiting in the wings. If he's good enough to keep in a test match, he's good enough to keep in an ODI.

It might not have made a difference in this series, but I think it's still in Zimbabwe's best interest to bring in Chakabva as keeper, and take the gloves off Brendan Taylor.


  1. Brilliant. Absolutely brilliant analysis, Michael.

    1. Thanks Nicko! I was surprised by how dramatic the difference was.

  2. Great article. To me there are some good keepers in Zimbabwe, not just Chakabva. Mutizwa is good as is Charles Coventry, but it's unlikely we'll see Coventry back in Zimbabwean colours anytime soon.

  3. This article is very educative, I hope Waller will come across it and take something out of it.