Analysis: The Problem with Using Career Averages

(Cross-posted from sportsanalyticsadvantage.com)

25th September, 2017.

It's extremely likely that the majority of those reading this will be familiar with watching on TV as a batsman walks out to the crease, and some data pops up on the screen regarding his batting performances.  

Typically these will be career averages, with possibly some extra detail given as to previous best scores, number of 50s/100s scored, and these are possibly fleshed out further by the addition of career strike rate data if the match is in a limited over format.  In domestic competitions, particularly T20s, these career averages displayed may be further filtered to display career averages in that particular event, such as the IPL or the T20 Blast, as opposed to all competitions.

It is frequent that the commentators will refer to these stats as the batsmen prepares himself, and it is evident from many of these comments that the commentators - usually ex-players - still are a little unaware of the numbers for an average player.  Frequently, a player may be described as being a quick scorer, despite a below-average career strike rate of 120 being displayed, or worse, the commentators may refer to a veteran as being an excellent player (justifying this by his career averages) when his data over recent years illustrates he is mediocre at best.


Strauss and Morgan data likely to show age-related decline


One example of this is the former England captain, Andrew Strauss.  Amassing over 7000 Test runs at a career average of 40.91, it's reasonable to consider that Strauss enjoyed an excellent career from 2004 to 2012. 

However, if we look at the latter stages of Strauss' career, from 2010 onwards, he played 31 Tests, scoring at an average of just 32.74 - an average that would put a batsman's place in the team at risk.  It's extremely dubious that a commentator would have picked this up, and indeed, his 2011 and 2012 combined average was even worse - effectively he declined from around the age of 33.

This decline is clearly apparent in Eoin Morgan as well.  My recent article here described Morgan's recent woes in T20s (both domestic and in internationals), and there is an evident decline in the England T20 captain's numbers.  

In T20 internationals throughout his career, Morgan has played 70 matches and averaged 28.65, but if we look at his data from the start of 2016 onwards, it falls to 22.38 at a below-average strike-rate.  It is highly probable that at 31 years of age, Morgan is experiencing a similar age-related decline to Strauss.

Despite this, T20 franchises around the world appear to be queuing up to take Morgan, decisions which again illustrate that franchises have much to learn when it comes to player recruitment, as well as auction and draft strategy.


Ageing players still favoured by T20 franchises


This interesting article by Freddie Wilde at CricViz gives a good example of this.  When Wilde looked at the average salary by age bracket of those signed in the T20 Global League, he actually found that there was a marked increase in the salaries paid to those over the age of 30, and in particular, over the ages of 35.

Given that players over the age of 30 are likely to be more high profile, it is extremely probable that many players recruited are done so on the basis of reputation, as opposed to their current levels of ability.  When selecting an older player, it is vital that due diligence is performed to avoid selecting players in decline.  Not doing so would be a huge mistake and this is one area that requires addressing in order to add value to a franchise. 


Bowlers also victims of late career decline 


Such age-related decline is also evident in bowlers.  Stuart Broad, this year, has taken 20 Test wickets at 33.90 at home against a non-peak South African team, and a weak West Indies side.  At 31 years of age, if Broad does not improve these numbers, questions will be asked as to whether his best days are in the past.  The forthcoming away series against Australia will be a real acid test of this, as proven by my analysis later in the article.

Age-related declines even beset all-time greats - Muttiah Muralitharan took 800 Test wickets at 22.72 throughout his career, but managed just 34 at a cost of 40.76 from the start of 2009 until his retirement during 2010.  

In addition, even older players who are considered to have improved in their latter years have exhibited weaknesses.  Misbah Ul Haq, for example, had a strike rate barely above 30 versus pace in Tests across recent years, but his record against spin was exceptional.  Team selection, and in-match bowling selection against these type of players is critical for a team to maximise their chances of success.


Opponent quality and home/away bias with a huge impact in career averages


Further flaws for using career averages to justify a player's ability level come in the form of home/away bias and also opponent quality.  The following table, which contains Test data from 2014 onwards, gives some more information, prior to analysis:-

Team

Home Batting Average

Home Runs Per Over

Away Batting Average

Away Runs Per Over

Batting Average Difference

Runs Per Over Difference








Australia

49.19

4.04

31.29

3.28

17.90

0.76

Bangladesh

31.72

3.23

29.10

3.09

2.62

0.14

England

36.57

3.54

31.14

3.08

5.43

0.46

India

39.78

3.27

36.76

3.42

3.02

-0.15

New Zealand

37.91

3.57

35.34

3.43

2.57

0.14

Pakistan *

44.31

3.30

31.36

3.14

12.95

0.16

South Africa

37.66

3.21

26.91

2.82

10.75

0.39

Sri Lanka

30.55

3.25

32.00

3.18

-1.45

0.07

West Indies

28.47

2.96

25.52

3.00

2.95

-0.04

* Pakistan's home matches were played at a neutral venue


This table shows that the Australian Test batting at home is by quite some distance the best in the world, demonstrating why Broad (and all the English bowlers) will have a tough task in improving their career averages in the forthcoming tour.  It also shows that South Africa and West Indies have been the two worst batting away teams in this time period, showing that instead of an England bowler's career averages worsening at home against these two sides, as Broad's did, they should have been able to improve them.

Looking at this, Toby Roland-Jones is an interesting player to discuss briefly.  Roland-Jones has enjoyed an excellent start to his England career (numbers-wise) but this deeper analysis shows that it was against the current weakest two away batting teams in Test cricket.  Unfortunately, Roland-Jones has now picked up an injury which has put his Ashes selection in extreme doubt, although it is highly probable that his career Test bowling average will thank his injury, given the extreme dominance of Australian batsmen at home.

It is also interesting to note that Australia, Pakistan and South Africa are by some distance considerably better batting when at home, as opposed to away.  Is it really fair to measure a player's career average if the majority of his matches are played against the better Test nations, or in away series?   After all, every major Test-playing nation had better batting data at home than away, with the exception of Sri Lanka.


Comparing player performance to expectations critical


It would certainly be important to guard against over-valuing a player based on a number of easier matches, particularly early in their career.  Gary Ballance, for example, boasted a strong career Test average early in his career, averaging 60.75 in his first eight Test matches in 2014, with three centuries to his name.  However, closer analysis shows that he scored 201 runs in three completed innings at home to Sri Lanka, and played seven of his first eight Test matches at home.  

While it's fair to say that his next overseas trip was also strong - he accumulated 341 runs across five completed innings - it was against the West Indies, one of the weaker sides in Test cricket.  Following this West Indies series, Ballance averaged 19.04 in 12 subsequent matches, with just two 50s and was dropped.  

On this basis, it's difficult to give much of a positive spin on the long-term chances of Keaton Jennings, Tom Westley and Dawid Malan in Test cricket, given that they haven't particularly impressed against weaker opposition at home.  Certainly, a trip to Australia for the Ashes will be a harsh examination.  As I've mentioned in previous articles, the selectors are somewhat over-valuing Malan in the wrong format - data shows he's a much better T20 cricketer as opposed to against the red-ball in the longer formats.


Domestic T20 cricket far from a level playing field


Domestic T20 cricket is another area where those who do not dig a little deeper into career average statistics can be misled.  The following table illustrates the overall data for each major domestic T20 league in the world from the 2015 season onwards:-

Country

Tournament Batting Average

Tournament Batting Strike Rate

Tournament Boundary %





Australia

23.97

128.85

15.05

Bangladesh

20.28

112.56

13.54

England

24.34

132.24

16.33

India

25.88

132.25

17.09

New Zealand

23.46

129.20

15.92

Pakistan

21.94

121.42

15.07

South Africa

24.05

125.22

14.77

West Indies

22.69

122.36

14.79


By some distance, India and England have the highest batting average and strike rates, as well as boundaries being hit more often.  I've mentioned that the T20 Blast is one of the easiest leagues to score runs in, and the IPL has even more impressive batting data - although it also possesses much better quality batsmen.  

Assessment of the main reasons as to why a league will have a higher batting average than another league will primarily focus on both pitch conditions, ground dimensions and bowler quality, and certainly it is reasonable to assume that a Bangladesh batsman (the hardest domestic league to score runs in) playing as an overseas player in the T20 Blast will be able to improve on his career average while in England.  

On the subject of Bangladesh, the data clearly indicates that the tournament is rather bowler dominated, with the lowest batting average and considerably the lowest strike rate on display.  The boundary percentage of 13.54% is also over one percentage point worse than any other of the major domestic T20 leagues.  Bowlers should be queueing up to be selected in the BPL draft, but batsmen, less so.


Player overseas contract choices will have impact on career averages and reputation


A player would be well advised to consider this data when attempting to formulate their future plans.  

For example, there is a clash in the calendar between the T20 Blast in England, and the Caribbean Premier League in the West Indies.  Assuming equal financial considerations, and a choice between contracts in the two leagues, a batsman should pick the T20 Blast, while a bowler is likely to achieve greater success in the CPL.

There is also an imminent clash between the Global League in South Africa and the Bangladesh Premier League.  While there is not yet any data for the Global League - this is its inaugural season - it is reasonable to assume that it will be more batting-orientated than the BPL based on the above numbers.  It would be more beneficial for batsmen to travel to South Africa to play in the Global League, while bowlers are more likely to thrive, and by definition, both improve their career averages and reputation, by playing in the BPL.


Narine's numbers much better in bowler-friendly conditions


Furthermore, there are a number of bowlers who have benefited from playing in rather bowler-friendly conditions, and who are perhaps over-rated by many observers.  

Sunil Narine, for example, from 2015 onwards, has a solid bowling average of 24.56 in worldwide T20 domestic and T20 internationals combined with a stellar economy rate of 6.44, and many consider him to be one of the best bowlers in T20 cricket.

However, Narine has bowled 133.4 overs in the IPL from 2015 onwards, conceding 950 runs (economy rate 7.11) and taking just 28 wickets (average 33.93).  The bulk of his success has been in bowler-friendly competitions, such as the BPL, CPL and PSL, where he combines to have bowled 184.3 overs in the same time period, costing 1026 runs and taking 54 wickets.  In these rather easier leagues for bowlers, he's conceding a much better 19.00 runs per wicket and costing a mere 5.57 runs per over.  Opponent quality and conditions has a considerable impact on Narine's numbers.

Rashid Khan, from a smaller sample size, is a similar case.  I have little doubt that the Afghanistan spinner has the potential to be one of the best T20 bowlers of all time, but his numbers in the weaker BPL and CPL combined so far (17.26 runs per wicket, 5.92 economy rate) are considerably better than in the tougher IPL (21.05 runs per wicket, 6.62 economy rate), although unlike Narine, his IPL numbers are clearly still world-class.

The point I am trying to make is this - no two innings or overs are equal, and some are considerably more difficult, and relevant, than others.  A player's past performance, particularly older than three years ago, is of very little relevance when assessing their current level, yet they do influence both career and tournament averages, and thus influence opinions - and often opinions of those with significant voice - incorrectly.

Our unique Sports Analytics Advantage algorithm assesses both the difficulty of a performance, and also gives more weight to recent performances, as opposed to those in distant memory, and therefore is a more accurate measure of a player's level than assessing career averages.  

Please feel free to enquire for further details on player analysis, and bespoke draft or auction strategies at sportsanalyticsadvantage@gmail.com.
Comments