How Can We Predict Batting Success At Test Level?
Jack Hope takes a look at what we can use to predict whether a player will be successful as a Test match batter.
Jack Hope takes a look at what we can use to predict whether a player will be successful as a Test match batter.
The wisdom of the crowds
Like all good ideas, the genesis for this article comes from social media. Whilst waiting for the number 197 to take me to Drag Bingo at the Signal in Forest Hill (10/10 would recommend), I asked Twitter (@JackHope0 - join the fun) what the best predictors were for players making the step up to Test cricket.
Generally speaking the responses were, I think, well thought through and taking inspiration from them I decided to Test a couple of the ideas. This got rapidly out of hand, so grab a tea or something, it’s a long one!
Let’s start with the most common response: performance in county cricket is a good indicator of future performance in Tests...
Better on average: comparing records between first class and tests
This one seems pretty obvious. The logic goes that if you are good at the level below Test cricket, then there is a decent chance that you will be good at Test cricket, and it is pretty easy to check that hypothesis. Below we look at the performance of every English Test batter to score 500 runs since 2000, and compare their Test average with their First Class average (within the text of this article “First Class” means First Class minus Test performance):
We can say, with some condifence, that there is a link between First Class batting average and Test batting average. In other words, if someone is a consistent performer at First Class level, you could make a fair assumption that their game will translate to Test cricket. However, at this point it would still be an assumption, the link is not so strong that we would be sure this would happen - and it would be wise to gather more data.
Beyond establishing a link between First Class and Test batting averages though,we see something very interesting, though predictable, when we look at the difference between the player’s First Class and Test averages. Basically, it is extremely rare for a player to outperform their First Class batting average.
It is safe to say then, based on the above, that whilst First Class performance may not be a rock solid predictor of Test success, it gives us a very good indicator of a player’s likely ceiling.
A littler intervention, which I am insterting post-publishing, as well as making a small edit to the text below the first graph.
Nathan Leamon, the England Cricket Performance Analyst, pointed out that the first figure may be an example of Simpson’s paradox.
And certainly he has a point, removing those players does change the game. Although, if we’re going to start cleaning up the data, then there’s justification for removing Pope and Ramprakash (the wildest outliers) from the equation. Do that and we still see a link between FC performance and Test performance, albeit a slightly weaker one (with R squared falling from .43 to .32 - (R squared being closer to 1, in basic terms, means the link is stronger)).
However you want to frame it, the second point from this section, that players rarely get better when they go from FC cricket to Test cricket still stands. It also demonstrates how more of these ideas need to be shared and challenged for us to understand the game better, so if you have ideas go and write about them!
Is Tom Abell able?
Knowing the above, the next step is to apply that logic to real life. For this section we’ll take a look at England’s recent selection, Haseeb Hameed, versus the “cricket hipster’s” choice for the England number three spot, Tom Abell. Obviously, this selection was obviously not a straight shoot-out between Hameed and Abell, but we’re going to pretend it was to explore a couple of other indicators: rolling average and performance versus quality opponents.
We start off with a comparison of career batting average. Tom Abell has the edge here, but it is noticeable, and possibly indicative of England's present situation, that both average under 35. As a possible mitigating factor both players played a lot of cricket whilst still very young, so we can assume that those averages may not be totally representative of the player’s talent level.
With that in mind, and to establish a clearer idea of ability, we can take a closer look at the performance of both by just comparing their averages in domestic cricket over the last two years, to generate a rolling average. Also suggested on Twitter was the idea of comparing performance versus quality opposition, we can do that by segmenting their domestic performance versus top and bottom half teams.
This paints an interesting picture and one that suggests that England should have listened to the hipsters. Based on what we know about career averages Abell looks like the best bet, and when we take a more contextual look at the data he also significantly outperforms Hameed.
It probably also shows us the power of an engaging narrative arc. Perhaps Tom Abell should consider maiming himself, or at the least consider donating a kidney, something to get his name out there.
A quick note, as we’re only looking at two players here we shouldn’t leap to any conclusions about Tom Abell being the missing piece, we’d need to apply these methods on a far wider scale. It would also be useful to test the link between rolling averages/performance versus quality opposition and performance in Test cricket - we are only assuming that this is relevant at this stage.
The Vaughan/Trescothick Conundrum
All that said England are not necessarily wrong to go with Hameed. Selection is an inexact science, occasionally players do outperform expectations, and assuming the county game is a perfect preparatory environment for Test cricket would be wrong. Among other reasons, England may, for example, think that Hameed has a high ceiling and would like to develop a potential great player instead of playing a good one now.
This brings us on to the Vaughan/Trescothick Conundrum, which has dogged this debate since they made their debuts 20 years ago. This has been covered in detail elsewhere, but is fundamentally misleading. As with Hameed and Abell, both made debuts at a young age, distorting their First Class averages, and both were on a strong upward trajectory at the time they were picked - heading into their peak years at 25.
It should also go without saying, that if your selection process is heavily influenced by two players from twenty years ago, who bucked a very well established trend, then you don’t have a very good selection process.
A massive diversion into scouting, baseball, and Moneyball
If you were just here for the cricket and the graphs you can skip this section. Likewise if you don’t give a shit about baseball the next 550 words or so are probably going to be really boring for you.
So far in the article we’ve focussed heavily on what we can measure with data. In my opinion data is still way underutilised in selection,but that doesn’t mean it should be used exclusively. The yin to the data yang, is traditional scouting, which is making subjective calls on a player’s talent level, technique, attitude, etc.
A simple example of this in cricket is assessing whether a player is a good fielder. Fundamentally, you need some experience watching the game before you are able to judge how impressive a catch was. We don’t capture any data on it.
There’s a famous scene, in Moneyball, which provides an illustration of the Data v Scouts debate, as Oakland A’s GM Billy Beane rows with his scouting team, portrayed as an array of dribbling idiots. Eventually Billy Beane wins the argument and immediately turns the A’s into a stats powered baseball juggernaut.
Or so we are told. The reality, when the 2002 A’s are analysed in detail is quite different, and they were driven primarily by the talent of three players who go virtually unmentioned in the book and film.
This Big Three and the art of winning an unfair game
The 2002 Oakland A’s were a good baseball team. Over the season they accumulated an impressive 15.4 Wins Above Average (a metric which measures, erm, wins above average by evaluating individual player’s various statistical contributions). This placed them 5th out of the 15 teams in the American League.
Breaking that down further and we can see that the biggest contribution to this total came from the A’s starting pitching, who chipped in with an enormous 12.3 WAA, 1st out of 15 in the American League). The key men in that rotation were the so-called “Big Three” of Tim Hudson, Barry Zito and Mark Mulder. This is what they did:
Tim Hudson was drafted 185th overall from Auburn University in 1997, where he had been their third best pitcher. Why were the A’s interested? Because one of their scouts, John Poloni, or the one described as the “fat scout” in the book, liked his splitter and sinker (two types of pitches), and thought he was athletic.
The other two (Zito and Mulder) were drafted as first round picks, based on the advice of the A’s Scouting Director, Grady Fuson, who is fictitiously fired in the film. These were more obvious picks, particularly in the case of Mudler, who was selected second overall, however they both represent further successes for the traditional “scouting” approach.
In a sport where around 50% of first round picks never make it to the Major Leagues, there was definitely some luck involved in the emergence of the Big Three. However, what is absolutely undeniable is that those players, delivered to the A’s by their fat scouts, were the foundation for the success of the Moneyball era - no other factor made more of a difference.
What does this tell us about projecting who will score Test runs for England? In direct terms, it tells us absolutely nothing. However it does show that even within highly analytically advanced frameworks, having the odd fat bloke around, who can say with confidence that “the way that player does that thing is good,” can add terrific value.
Conclusion - Processing it all
First of all it is worth remembering that selecting a Test team is an inexact science. The cases of Pope and Ramprakash show that it can go spectacularly wrong, whilst Trescothick and Vaughan show that, on rare occasions, it can go very right.
Secondly, and as part of that process, we need to ask more questions. We can see there are links between First Class and Test performances, but what other information can we factor in here and to what extent? As a next step it could be worth investigating whether: age, home venue at domestic level, performance on Lions tours, performance in international white ball cricket, etc.
Thirdly, and probably most importantly, it is important that teams have a process in place. As selection is unpredictable, failure is inevitable, so teams should focus on implementing the best process they can in this area. A good selection process would probably use data to screen for potential talent and then unbiased scouts to augment that process with expert insight. Crucially the same method should be applied for every player, you can’t just use data when it suits you...
Is this what England’s process looks like? You be the judge...