Debutants in Away Tests have shorter careers

Would you expect players to be disadvantaged by making their debut overseas? Surely the best players get picked and have a decent run in the side until there’s sufficient data to disprove the analysis that got them selected in the first place?

Afraid not. Away Debutants are discriminated against! Debut at home you can expect a nine Test career. If your first game is an away match, that drops to six.

Fig 1 – Average (Mean) and Median Tests played by location of Debut. Includes top nine Test teams, since 2005.

A reminder – home advantage in Test Cricket is big. Somewhere around 17%, depending on how you cut the data. If your expected batting average is 35, that’s 38 at home and 32 away. A player who starts their career overseas is likely to underperform, and is at greater risk of being dropped when the naïve assertion is made “OK, they have a decent First Class Average, but they are only averaging 29 in Tests.”

Half of Away Debutants don’t make it to seven Tests. And yet the mean number of Tests played by Home Debutants is only 1.1 matches more than Away Debutants. For some reason the early benefit to Home Debutants doesn’t persist. What happens after seven Tests to explain that?

Fig 2 – Frequencies of Number of Tests played. Includes top nine Test teams, since 2005.

The behaviour flips – from Tests 7-20 more Home Debutants are discarded than Away Debutants. I expect that this is because some players who had an easy home series to get into Test Cricket then get caught out when away from home.

After 20 Tests, a player has generally played a similar numbers of home and away Tests, so there’s no great difference between the two curves.

So What?

  1. Some Away Debutants play fewer Tests than they deserve. Conversely, some Home Debutants are kept in the team longer than they should be as a result of the stats boost they get from playing more home Tests than away.
  2. It’s time to move on from raw averages. Adjusted averages are the future. Not just adjusted for home/away, but also the ground they are playing on (think Headingley vs The Oval), the quality of opposition and the innings number. This is not a complicated task, and I’d be very surprised if it isn’t already happening behind closed doors. Admittedly I haven’t yet done this when rating Test players. But then, this is a hobby for me. Also, until a player has played 20 matches, I use their First Class average to appraise them. Which is coincidentally the point at which Debut Location ceases to matter as an input.
  3. Don’t make your Test Debut in an away game if you can help it. I appreciate this is not practical advice, so instead, if anyone reading this has made their Debut in an away game, make sure you quote your home/away adjusted average whenever possible! Ebadot Hossain, am looking at you.

It’s almost the same story for ODIs

A quick calculation says Home Advantage in ODIs is c.11%, so we would expect ODI debutants to have similar trends to Tests. Which is true for matches 0-20: Away Debutants are more readily discarded after a handful of games, then Home Debutants are in the firing line from 4-20 matches.

Fig 3 – Frequencies of Number of ODIs played. Includes top nine ODI teams, since 2005.

After 20 matches it gets more interesting. Overall, Away Debutants have greater longevity on both a mean and median basis. Of the Post-2005 players with more than 100 ODI Caps, 16 began at Home, 22 began Away.

Fig 4 – Average (Mean) and Median ODIs played by location of Debut. Includes top nine teams, since 2005.

What the dickens? I can’t confidently explain this. Could have hidden it from you, but it’s interesting and therefore worth sharing, even if I don’t understand it. I’ll offer one possibility: ODI series are often tacked onto Test series, so in an away series the star Test players stay on for the ODIs, meaning that only highly regarded red ball players make the team. At home, the top Test players can more easily be rested, so lesser known players might get a go.

The Short List: Away Test Debutants

Below is the list of players that played fewer than seven Tests, and started away from home. Have a read, see if you can pick out some players who might have had 20 Tests if given the benefit of a home debut. Luke Ronchi and Owais Shah jump out at me.

Fig 5 – Players to Debut away from home since 2005 and play fewer than seven Tests. Data implies 20 of these players would have played 20 Tests if they had debuted at home.

A review of England’s bowling options

When England fans are nervous, hits to my summary of their Test batting options spike. This is the companion piece for bowling, allowing me to monitor a nation’s worries about replacements for Broad and Anderson.

We’ll start by looking at how performances since 2016 translate to expected Test averages, then discuss the implications of that.

Here’s my view of the expected batting and bowling averages of the leading contenders:

Fig 1- Expected Test averages of England’s leading bowlers, based on data since 2016. Note the reversed x-axis: an ideal player would be in the top-right, a weaker player bottom-left. Anderson stands head and shoulders above the other bowling options. For the second and third Ashes Tests, the attack of Broad, Archer, Leach, Woakes, Stokes is pleasing in that all five bowlers are from the best eleven available.
Fig 2 – England’s bowling options – those with expected average below 30 and selected others.
Note that Archer’s white ball record implies he will be more successful than recent red ball data indicates.
County Cricket performances won’t necessarily translate to Test Cricket – where pitches are flatter and games aren’t played in April/May/September in England. Stevens probably wouldn’t average 30 in Tests, but one should start with the data and adjust rather than the other way around.

Discussion

1. Older players & Succession

Five of the top 17 players are aged over 33. That means England need clear succession plans. Conversely, it also suggests Woakes and Broad might have more Tests in them than we think: Stevens, Anderson and Clarke have not diminished with age.

2. Ben Coad

Coad has consistently performed well in Division 1 for Yorkshire. Last three years: 50 wickets at 21 (2017), 48 wickets at 16 (2018), 36 wickets at 25 (2019). You know how Simon Harmer has been tearing up Division 1 and winning games for Essex? He has 156 wickets at 20 since 2016; comparable with Coad’s 135 wickets at 21.

It was a surprise that Coad came out so much better than all other bowlers bar Anderson. Consistency is key – for instance Broad and Woakes had a bad year in 2017 (averaging 36 and 51 in Tests respectively).

The next red ball Lions activity should feature Coad. It’s astonishing that he hasn’t played yet. England weren’t far off with the Lions attack of S.Curran, Gregory, Robinson, Leach, Porter- but they’ve got to find a way to look at Coad.

3. Division 2: Ben Sanderson and Ryan Higgins

I’d like to see Gloucestershire and Northamptonshire get promoted to Division 1, mainly as the neatest way to get these two playing the best standard of Cricket available. There’s a significant leap in standard between Division 2 and Test Cricket, so without ball-by-ball data it’s hard to be sure how good Sanderson and Higgins are.

If Gloucestershire don’t get promoted this year, I wonder if someone will have a quiet word with Mr Higgins and suggest he seek a Division 1 employer. Higgins is very good. I wrote about him here.

Sanderson is the wrong side of 30, so if he were to get a Test callup it would be following a lot of injuries to younger alternatives. Like James Hildreth he’ll be someone who could have made the step up from Under 19s to the full England side, but never got the chance.

4. Spin options

There’s only one viable spinner- Jack Leach. Even adjusting for the advantage he gets from playing at Taunton, he’s the best England have got. His batting’s not great, so in non-spinning conditions England should consider a batting all rounder instead. Maybe that’s harsh on Moeen Ali, but I think the “most wickets for England in the last 12 months” statistic flatters Ali – taking the longer view, his Test bowling average of 37 is nothing much to shout about.

5. Replacements

If Woakes or Stokes were unavailable: Gregory or Higgins are the best batting bowlers on the list, capable of slotting in at number eight.

If Broad or Archer were injured (and Anderson still out), Coad would be the logical replacement.

I don’t see Sam Curran as being ready for Test Cricket. His bowling average of 30 flatters him when his first class average is 29: expect it to go up if he plays more Tests. He’s only 21 – for now there are better bowlers out there.

Post-script: Methodology

To calculate expected Test averages, I took performances over the last three-and-a-half years in Second XI, County Championship, and Test Cricket adjusted for the relative difficulty of playing at each level.

I’m aware of two extra elements to add: weighting towards more recent performances and adjusting for age (young players should be getting better). These will take time to calculate, so will have to wait for the Autumn.

There’s a third factor I’d like to look at – the link between ODI and Test performance. Since not all players will perform equally well in red and white ball Cricket, I’m at present unsure how I’d quantify such a measure (eg. X averages 26 in ODIs, therefore is expected to average 32 bowling in Test Cricket).

Further reading

Wisden tipping Coad for greater things: https://www.wisden.com/stories/county-cricket/ben-coad-yorkshires-late-bloomer-englands-potential-wildcard – no doubt I’m not the first to notice that Coad is rather good.

Mythbusting: Vaughan and Trescothick selected for England despite modest First Class Records

In the perennial debate “Selection: Art or Science”, one of Art’s arguments is that Michael Vaughan and Marcus Trescothick had Modest Records* before succeeding in Test Cricket. In this piece, I’ll demonstrate that there was nothing in their early international red ball careers that couldn’t have been predicted by the right data.

Let’s start by looking at young Michael (born 1974) and Marcus (born 1975) developing in County Cricket.

Fig 1 – After an exceptional first few seasons note the unusual blip in Vaughan’s 1999 performance. This meant that rather than showing improvement with age, his career average hovered around 36.
Fig 2 – A more pronounced improvement with age. Three year rolling average consistently exceeds career average.

Using the three year rolling averages as a measure of current ability, and applying the current performance difference between County Championship Division 1 and Test Cricket (28%, see link), we would expect both players to average 28 in Tests. Adjusting for age (a 25-year-old is better than they were at 23, so would be expected to outperform their three year rolling average), the data driven approach says they had expected Test averages of 29.

I am interested in the predictive power of this data. The decision to elevate them to Test level in 1999 (Vaughan) and 2000 (Trescothick), should be judged on what they delivered by the end of 2000**. How did they get on? Between them 712 runs at 31***. That’s at the lower end of an acceptable average for a top order batsman, albeit two runs per wicket better than expected. If the “Art” camp would like to claim credit for left-field selections, I’m willing to give them credit for two runs per wicket.

Fig 3 – Vaughan and Trescothick 1995-2000 batting data. Limited to County Championship matches where possible (1995 and 1996 were the early days of the internet, so I could only find First Class information)

Discussion

If you judged a player on career County averages you would have balked at Trescothick’s selection in 2000.  We have seen that career averages are misleading: Trescothick played County Cricket aged 20; his career average took a long time to recover from playing while so young.

Career average is a blunt measure of ability. If the sample size is big enough, a better method is to use last three years’ average**** and adjust for age and the level they played at to give an expected Test average.

Conclusion

Vaughan and Trescothick had modest career records before their Test debuts. They also had modest starts to their Test careers!

Taking nuanced view based on age and recent averages, the Test career starts of Vaughan and Trescothick were entirely predictable- no hunches necessary!

Further Reading

Here’s the excellent Vic Marks discussing Vaughan and Trescothick’s records at the time. Gives a bit more context around other factors behind their selection (England U19 and ODI performances): https://www.theguardian.com/sport/2000/jul/30/cricket

*Google is full of classic examples of Modest Records. I like this one because it features one protagonist describing the other: Michael Vaughan, 2014: We have picked players in the past with poor county records, such as Marcus Trescothick, and they have thrived on the international stage.

** They averaged 52 and 41 in domestic Cricket in 2001, so my contention is that they would definitely have been ready by late summer 2001.

*** I’m keen not to be accused of cherry picking the data. By the end of 2000 they only had 23 completed innings between them. Extend the analysis to the end of 2001 and they averaged 36 after 32 Matches.

The other factor to ensure I’ve not rigged this is to note who they played against and where. Vaughan had four tricky Tests away in South Africa, then four easier games hosting the West Indies. England went undefeated in Trescothick’s six Tests against West Indies/Pakistan. Three were at home, three away.

**** Three year rolling average is a bit simplistic. An even better methodology would be to weight towards more recent innings. That’s a “nice-to-have”. Three year rolling average is good enough for our purposes.

Sibley or Roy?

Sibley or Roy

For the third Test I’d like to see Dominic Sibley open the batting. To subscribe to this line of reasoning, you’ll need to be persuaded of two things: firstly, it is not necessary to gain experience in “lesser” Test series to ensure peak performance in the Ashes. Secondly, that Sibley is one of the two best opening batsmen England have available.

Opening Up

There’s a school of thought that new players need to be “blooded” to succeed without first playing against weaker teams in the Test format. The data indicates that this is a fallacy.

The 96 openers to make their debut since 2005 scored at an average of 32. That is poor when compared to the average for all openers (36). However, that’s not the right comparative. Weaker players will play fewer Tests, so debutants are of lower ability than the average Test player.

A better way of assessing openers on debut is to compare performance with that player’s career average, adjusted for whether the debut was at home or away. Looking at it that way, players on debut scored three runs more per match than expected.

Why should openers do well on debut?

This is unexpected – often players will debut before their peak, their average will improve as they get better with age. It’s possible that openers are not thrown in at such a young age as middle order batsmen (because it’s a specialist position and no one wants to be 0-1).

Another option is that batsmen had an advantage when the bowler has to played against them before, and is yet to develop a plan. This may have been the case last decade, but is unlikely in modern cricket.

Note that debuts were evenly spread across opponents- it’s not like selectors wait for the weaker opponents before trying new players.

Sibley the Best?

Before the summer, I rated Dominic Sibley as a decent opening batsmen, impressive for a 23 year old, but some way short of Test standard. His expected 2019 First Class average (based on Championship and 2nd XI matches from 2016-18) was 36. That made him at best seventh on the list of possible Test openers. For fans of lists, Burns (51), Stoneman (44), Jennings (42), Mitchell (40), Hales (39) were ahead of him on merit, and Hameed (expected average 36) was also ahead because of his fame.

Fast forward to August 2019. Reflecting Sibley’s spring/summer return of 940 runs in my ratings, his expected Division 1 average jumps to 42. Tied for second place with Stoneman.

Adjusting for age, Sibley would expect to average 34 in Tests.

If one limits the search to red ball cricketers, there could be few complaints with Sibley opening the batting for England.

However, there’s this Roy chap. Top ODI player- averages 43. Can he make it as a Test opener? Leaving his white ball record to one side (because I’ve not looked at the predictive power of white ball results on red ball expectations), in First Class Cricket he averages 38. He has performed better recently: averaging 43 in Division 1 over the last three seasons. However, in those three-and-a-half-seasons he barely played: 32 completed Test/Championship/2nd XI innings while Sibley has 73. Roy didn’t open in either of his games last year.

Conclusion

Don’t be afraid to give an opener a debut if their record says they are capable.

A reasonable scenario is that Denly (expected Test average 31) picks up 60-80 runs over the course of the Lord’s Test this week, and is dropped on the back of averaging 24 after ten innings.

Sibley to open, Roy slots in at four? Could work.

Do Hundreds Matter?

Ambivalence: having two opposing feelings at the same time, or being uncertain about how you feel.

Cambridge Dictionary (Cambridge University Press)

Yesterday (2nd August 2019) Rory Burns scored his first Test century.  It was a struggle. It was error strewn. But he got there. In the evening the twitter consensus was that the important thing was that he did it, not how he did it.

Fig 1 – Twitter Screenshot.

What didn’t I like about Butcher’s point*? I’ve always thought that you can learn more from a lucky long innings than a duck. Someone gives a chance first ball, gets out – so what? Another player hacks their way to thirty, regularly playing and missing or chopping the ball past the stumps – well there you have some useful data – the bowler is dominating the batsman with a decent sample size.

But a hundred? I’m conflicted. You can’t use a hundred as evidence that a player is no good. Can you?

You are surprisingly unlikely to get more than a hundred if you average under 25

Fig 2 – Theoretical Probability of scoring a century in a single innings vs average.
Fig 3 – Theoretical Probability of scoring at least one century over fifteen innings

Over his first seven Tests, Burns averaged 22. A batsman with that ability only has a 1% chance of getting a hundred in any innings. Does that mean that Burns was lucky? That depends what you believe: either he is a twenty-something averaging batsman who had some fortune, or a thirty-something averaging player who is performing as expected. County data says he’s the latter.

One can get lots of runs at a decent average without a hundred.

In the appendix I’ve listed the top run scorers in Test Cricket who didn’t get a century. Was Chetan Chauhan any less of a batsman because he scored 2,000 runs averaging 32 but never made more than 97? No, the data says there’s a 5% probability that it was just chance that meant he didn’t get a hundred, with no psychological flaws or lack of stamina at play.

That’s all lovely – but DO HUNDREDS MATTER?

Joe Root’s hundreds don’t matter. He has played a lot of innings, so his average is the one metric you need. 49.03.

Rory Burns’ hundred doesn’t matter to me (other than to say there’s no reason to over-rule my analysis– expect him to average 39 over the long term in Test Cricket). Burns scoring a century matters only if Test scores are all you use to appraise batsmen. You would rate him significantly more highly after scoring a hundred than before (though, knowing it was an ugly hundred, you would probably rank his expected average somewhere from 25-35).

One example where I think hundreds matter is when there’s very little to go on. Dan Douthwaite scored a hundred against Sussex when playing for Cardiff University. As he wasn’t on my radar at that point, that 100* meant that I could safely assume he would average more than 25 in Division 2.

I’m thus ambivalent about rating players based on scoring hundreds: most of the time I’ll take averages over hundreds. When there’s noting else to go on, a big score can tell you that someone’s not bad.

Appendix:

Fig 4 – Test records of batsmen with no hundreds. Final column shows the probability of a player with that average not scoring a hundred in the number of innings they played.

*I should add that Control Percentage isn’t a metric I value as highly as False Shot Percentage. To my embarrassment, I’ll admit I misread Butcher’s original tweet as criticising False Shot Percentages. Still, serendipity – if I could read, I’d never have worked out whether I care about hundreds.

The Ashes: A tale of two spinners

I wrote an Ashes preview. It was boring. You won’t be subjected to it. Fortunately, when researching that I noticed a strange feature of Nathan Lyon’s bowling: he is great in the first innings of a Test.

At the time of writing it’s unclear whether we’ll see Moeen Ali vs Nathan Lyon as the opposing spinners in the 2019 Ashes – Ali’s batting has been poor of late, so it’s hard to justify his selection. Easier to make seam-friendly wickets and neutralise Lyon. Career averages show why that’s tempting:

Fig 1 – Nathan Lyon and Moeen Ali’s Test bowling records (as at 30/7/19)

That data masks two things – firstly, since 2017 both bowlers average 29. Secondly, and interestingly, how they perform through a match.

Fig 2 – Lyon (Yellow Triangle) and Ali (Green Square) by Innings of the match. Axes are the same in all four charts.

Let’s walk through that quartet of charts. In the first Innings, Nathan Lyon is about as good as it gets. An average of 32 is 11 runs per wicket better than the average for all spinners. He’s right up there with Ashwin & Jadeja. Moeen Ali is, frankly, awful. Averaging 16 more runs per wicket than Shane Shillingford. That Green Square is poles apart from Lyon’s Yellow Triangle.

Through the second and third innings, Nathan Lyon stubbornly refuses to improve. The chasing pack catches him, then outshines him by the third innings. Ali is comparable with him at that point (and within touching distance of the rest).

Now it gets weird. If anything, Lyon is worse in the fourth innings. A bowling average of 34 is now ten runs worse than that for all spinners since 2010. The control is still there, as his economy rate is unaffected. The sample size is fine (58 wickets in the fourth innings). Odd.

Meanwhile, the fourth innings is Ali’s playground. 59 wickets at 22, he’s right up there with the big boys. Go Green Square, go!

Let’s end with some practical uses for this, before it becomes pub trivia.

  • Nathan Lyon can be part of a four man attack for Australia – he can bowl effectively in the first innings, so Australia don’t need to play four quicks to have sufficient firepower early in the match.
  • Moeen Ali shouldn’t bowl in the first innings for England. Stokes can play the role of fourth bowler, and Ali should bowl no more than ten overs per day. Save him for later in the game.

Further Reading

Cricinfo independently noticed this back in 2017 (ie. I haven’t copied them, honest!) Unfortunately for them, they attributed the difference to the Asian continent. That quirk has now been ironed out.

Could Woakes open the batting in Tests?

My thinking was akin to some shambolic calling when running between the wickets: No. Yes. Wait – No!

No – Jonathan Agnew suggested that Chris Woakes was a potential opener a few years ago. To paraphrase, Woakes was someone that could do a job, perhaps on tour if an opener got injured.

This sounded ridiculous- fast bowling all rounders don’t open the batting, especially when they normally bat at six or lower. Presumably this was just a case of a commentator getting carried away during a long stint at the microphone.

Yes – And yet, as time went on, with Cook appearing to struggle and Strauss’ shoes unfilled, maybe there was some sense to it. With Root refusing to move up from number four, England had two top three places to fill (soon to be three once Cook retired).

Wait – Hang on. Easy to have wild ideas from the sidelines, would anyone really pick an untested all rounder to open the batting when they could pick from an assortment of county openers? One would have to be pretty desperate. Best to give Ali, Burns, Buttler, Compton, Denly, Duckett, Hales, Hameed, Jennings, Root, Roy, Stoneman a go before doing anything rash!

And maybe England do see something in Woakes’ batting – he batted at three against West Indies in the World Cup. Time to look at this properly.

No – While I can’t see any record of Woakes opening the batting, we can see performances against the second new ball. Not a bad proxy for performance at the top of the innings. Soon I’ll build a County Championship ball by ball database, for now his Test record will have to suffice:

Chris Woakes’ overall batting average is 29 in Tests, 35 in First Class. While he has only been dismissed nine times when batting during overs 80-100, averaging 26 in that period says that if anything Woakes would average less than his career average if he opened the batting.

Also, moving a weaker batsman up to open moves everyone else down one place. That increases the chance a batsman runs out of partners. Not ideal.

Not an especially interesting conclusion: Woakes could but shouldn’t open the batting for England.