T10 Analysis - Just Use The Data!

25th November, 2019.

Email: Sportsanalyticsadvantage@gmail.com

Yesterday, the 2019 season of the T10 League concluded with Maratha Arabians winning the title, convincingly defeating Deccan Gladiators in the final.  I've been vocal on social media several times during the tournament with my thoughts extremely strong towards teams failing to understand the format with regards to their recruitment, selection and strategy decisions, and I want to illustrate some of this in greater detail here.

Firstly, it is my belief that the T10 format is extremely likely to grow throughout the world.  It has uniqueness and brevity - something which is seemingly in demand among society in general currently - while I also like the fact that there aren't any demands on using local players as the majority of the squad, which is unlike the various short-format competitions worldwide.  In theory, this concept would increase the quality of the majority of leagues worldwide, with some leagues in particular using locals as almost 'filler' places in teams and squads.  Furthermore, it is interesting to compare the merits of England fringe/county players against West Indies players, as well as a number of other players from the likes of Afghanistan and Sri Lanka.   

In fact, I must admit that on a personal note I was disappointed not to be involved in the tournament, and for several years now I have had three specific tournaments in mind that I would be extremely keen on working in - these are the T10 competition, as well as the IPL and The Hundred.  Having secured a role in The Hundred with Birmingham Phoenix, the two other leagues still remain on my to-do list.

As well as feeling that the T10 format is ripe for expansion, one of the other reasons why I am keen to work in the format is that I believe that the teams are far from running in an efficient manner.  Captains seemed unsure of what would be a good score batting first, and there has been a huge chasing bias.  Recruitment was, in my view, extremely bizarre by many teams - for various reasons - and in this piece, I'll be analysing some of the numbers which focus on the above areas.

Establishing a Par Score

Throughout the first two T10 tournaments - the 2017 event was a rather shorter tournament - it is fair to suggest that there was a strong chasing bias.  In fact, 70.73% of chasing teams won in the first two years of T10 (2017 and 2018 combined) but these matches were played in Sharjah which, interestingly, also has a chasing bias in T20 as well and perhaps this had a considerable influence in those historical T10 matches in Sharjah.  The 2019 season was played in the Sheikh Zayed Stadium in Abu Dhabi, and I think this had some bearing in captains finding it difficult to establish a par score, although in the 2019 T10 event, batting first teams won one more match than chasing teams batting first, so the chasing bias disappeared this year.  

However, I said that establishing a par score batting first is difficult, but not impossible.  Throughout 2017 and 2018 the mean score batting first was 111.37 for 5.05 wickets lost, but let's not forget that teams batting first won fewer than 30% of these matches.  It's probably reasonable to dial this up to around the 125 mark to generate a 50-50 chase at Sharjah.

Based on the scoring this year, and general commentator's thoughts, it was clear that run-scoring was tougher in Abu Dhabi than Sharjah, and historical T20i data from the two venues would be in line with this rationale.  In Sharjah, players scored at a strike rate of 123.49 from 2015 onwards, but this fell to 109.76 in Abu Dhabi (a ratio of 0.89).  Batting averages were very similar at both venues during this time period.  

An extremely basic ratio adjustment, for example, would reduce this par score at Sharjah of 125 to 111 at Abu Dhabi, which would have given captains and coaches a figure to work with in advance of the tournament.

Of the 10 instances of teams scoring 111 or greater at Abu Dhabi in 2019, nine of them were won by the team batting first while of those teams scoring 110 or fewer this year in T10 batting first, just three of 14 non-tied matches went the way of the team batting first.  So, it seems evident that scoring greater than this ratio-driven par score - I'll re-iterate, a figure that could have been established using data in advance of the tournament - should have been a clear and realistic target for teams looking to put a competitive score on the board batting first.

Batting Recruitment

I have so many thoughts and data-driven evidence on how to recruit batsmen efficiently in T10 that I could probably write a book on the subject, but I'll try and summarise the discussion as best as possible. 

Firstly, historical data from previous T10 editions is very useful indeed with regards to working out drivers for format success.  Prior to this year's T10 event, teams would primarily have needed to focus on the 2018 event which had a considerably bigger sample size of data than the 2017 tournament, and here are some charts which illustrate some drivers for tournament success from the 2018 event:-

This first chart looks at team batting average vs team batting strike rate, with teams in red qualifying for the knock-out stages.  It isn't difficult to establish that the top three batting teams (by some distance) qualified for the knock-out stages.  While some readers may think this is obvious, it actually isn't as obvious as it would appear at first glance - the best batting does not nearly guarantee T20 teams qualification in this way, with bowling quality being a considerable driver with regards to team success in T20.

Given that team batting strike rate was a clear driver for success in 2018, it is logical to consider that boundary percentage would also be a huge driver of success as well.  There is a clear direct relationship between player boundary percentage and their strike rate (for example, it is extremely rare for a player to strike at 145+ in T20 if they have a below-average boundary percentage) and the chart below rubber-stamps the above assertion that boundary percentage would be a huge driver of team success:-

The teams with the highest boundary percentages (furthest right on the x-axis) all qualified for the knock-out stages, and the four teams who had the lowest failed to do so.  We can also see that Rajputs and Sindhis in 2018 had the second and third highest non-boundary strike rates - effectively they were good rotators of strike as a team unit - but failed to qualify.  In such a condensed format, rotation of strike is hugely over-rated and batting recruitment should focus on boundary-hitters.

This is particularly the case for the top four batsmen for each team, who in 2018 faced on average 41.30 balls per innings (out of a maximum of 60).  If you have a batsman in the top four of a team who isn't an above-average boundary hitter, and capable of striking at 180+ over the long-term, teams aren't scoring with close to maximum efficiency.

The table below illustrates the percentage of balls faced and balls per innings for each batting position in T10 in 2018:-

Batting Position

1

2

3

4

5

6

7

8

9

10

11

%

21.19

24.00

15.61

13.02

8.01

6.42

5.52

3.51

1.79

0.57

0.35

Balls per Innings

11.86

13.43

8.73

7.29

4.48

3.59

3.09

1.96

1.00

0.32

0.20


It is absolutely evident from that 2018 historical data that teams must prioritise their batting resources towards above-average boundary-hitters.  However, in 2019, this wasn't nearly the case...

Here we can see the relationship between a player's T20 boundary percentage from the end of the 2016-17 Big Bash onwards, and their 2019 T10 boundary percentage.  While the T10 balls faced for each batsman isn't of a huge sample size (this sample covers those who faced 50+ balls), it is already possible to draw several conclusions.

Firstly, if a player hits below 20% boundaries in T20, it is extremely difficult for them to be a considerably above-average boundary hitter in T10, and it is a straightforward process to create an algorithm to predict a player's boundary % and strike rate in T10 based on their performances in T20, taking into account factors such as recency and opposition quality.  Of the sample above, only Johnson Charles, Chadwick Walton and Lendl Simmonds managed to achieve this, and it is interesting to note that all three of these players are West Indian.  There is a reasonable theory worth exploring that West Indian domestic batsmen - typically known for being poor rotators but strong boundary-hitters - struggle to fulfil their full boundary-hitting potential in the CPL and in home T20 internationals given conditions but are able to express themselves much more in more batter-friendly conditions in T10.  This looks to be the case with this trio, although Walton (51.28) and Simmons' (42.22) non-boundary strike rates were so poor that they actually had a sub 170 strike rate despite such strong boundary-hitting.

What we can also see is that only just nine of the 26 batsmen who faced 50+ balls in this year's T10 event hit over 20% of balls faced for boundaries in T20 across the last three years or so, and with this in mind, teams evidently didn't recruit to give themselves the best chances of success.  Certainly, there were several very strange signings with veterans Hashim Amla and Angelo Mathews - who have enjoyed fine careers, although mainly in Test matches and 50 over cricket - predictably delivering below-average performance relatively in line with expectations in the bottom-left hand corner of this chart (low T10 boundary %, low T20 boundary %).

Mathews played all six of Delhi Bulls' matches but it would be a statistical struggle to make a case for him over the likes of team-mates Will Jacks or Sherfane Rutherford in this format, while Amla was also an ever-present for Karnakata Tuskers, whose batting recruitment I found particularly curious.  Both Mathews and Amla managed non-boundary strike rates in excess of 80, which is elite-level for this competition, although as discussed, this isn't the skill-set required for success in this particular format.

So, it seems clear from historical data that a strong batsman in T10 needs to be able to strike at 180+ (managed by 10 of these 26 players) with a T10 boundary percentage in excess of 25% (managed by 15 of these players).  As mentioned, it would be possible to create an algorithm to generate a database of players expected to achieve these targets, and this is a service that I will be able to offer T10 teams moving forward.

Did the 2019 T10 tournament continue in line with requiring high strike rate and boundary percentage as a driver for success, as it did in 2018?  Pretty much, yes.  Here are the same graphs for 2019, as I detailed above in 2018 - again with teams in red qualifying:-

In 2019, four of the five best teams from a batting average perspective qualified (also a solid success driver in T20 cricket as well), while the two teams with batting averages of 20 or below won a total of four from 11 completed matches.  

In addition, the three teams with the highest boundary percentages all qualified.  Arabians' boundary percentage was off the charts.  The two best rotation-orientated teams, Abu Dhabi and Tuskers, failed to qualify - again indicating that the ability to rotate the strike in this format is completely over-valued by teams.

Specialist Bowlers are Essential

While we saw in 2018 that boundary hitting and general batting quality was a big driver to T10 success, the bowling ability of teams was also critical - as it is in T20 cricket.

So, the chart above highlights that three of the four teams closest to the bottom left-hand corner (low team bowling average and low team bowling economy) qualified for the knockout stages - again highlighted in red.  Punjabi Legends' batting data was so poor last year that even a solid bowling effort was not enough to see them qualify from the group stages.

In a recurring theme from last year, the same thing happened with T10 bowling attacks.  Four of the five teams closest to the ideal bottom-left hand corner qualified from the group stages while Warriors went far off the equivalent last year's Legends - poor batting stability couldn't bail out a decent bowling effort.  

However, it's easy to sit here and write about how a teams bowling attack needs to be strong in order for them to have a strong chance of qualification.  Ideally we'd also be able to ascertain how to create that strong bowling attack.

The first and obvious way to do this would be to try and work out player expected data based on their weighted T20 numbers, and the algorithm which I have developed can do just this.  Because of commercial sensitivity, I won't go into this in great detail here.

Despite this, what I have done is create a chart below which illustrates the economy rates of the T10 bowlers bowling 10+ overs in the 2019 tournament and relate it to their bowling economy rates in T20 from the 2016-17 Big Bash onwards, for bowlers with a decent T20 league bowling sample size:-


There are several conclusions we can draw from the above.  Firstly, that low T20 economy specialist spinners had predictable success in the competition, with Fawad Ahmed, Qais Ahmad and Chris Green being closest towards the ideal bottom-left hand corner (low T10 and low T20 economy).  In short, pick a top-level spinner and you should get rewarded.  However, Sandeep Lamichhane didn't have such a good time of things (although his T20 economy is also worse than these three players) and perhaps T20 coach Dean Jones, on commentary, may have been onto something when he suggested this isn't the best format for the Nepalese leg spinner.  

What we can also see is that the players with T10 economy below 10 are all regular T20 bowlers.  With the exception of Andre Russell, all are specialist bowlers and not all-rounders.  Conversely, as we look towards the disaster top-right hand corner (high T10 and high T20 economy) we see that the group of players in this area are exclusively all-rounders (David Wiese, Jordan Clark, Thisara Perera, DJ Bravo and Ben Cutting). 

Ditch the All-Rounders

As you might imagine from the above paragraph, my data perceives all-rounders to be over-rated in this format compared to regular bowlers.  Below is a list of the bowling figures for all-rounders in the competition in 2019, and the combined economy rate in particular for these 21 all-rounders does not make for pretty reading:-

Player

Team

Balls Bowled

Runs Conceded

Wickets

Average

Economy








OVERALL


817

1600

57

28.07

11.75

Andre Russell

Warriors

60

100

6



Asela Gunaratne

Warriors

18

28

0



Ben Cutting

Gladiators

60

128

3



Dan Lawrence

Gladiators

6

2

0



Dasun Shanaka

Arabians

42

97

1



David Wiese

Tigers

85

156

7



David Willey

Bulls

42

98

1



DJ Bravo

Arabians

84

162

8



James Fuller

Arabians

12

23

0



Jordan Clark

Qalanders

66

123

9



Kieron Pollard

Gladiators

42

92

1



Lewis Gregory

Abu Dhabi

6

18

0



Moeen Ali

Abu Dhabi

12

30

1



Mohammad Nabi

Bulls

48

88

3



Paras Khadka

Abu Dhabi

12

30

1



Peter Trego

Qalanders

10

23

1



Robbie Frylinck

Tigers

42

113

1



Samit Patel

Qalanders

30

56

5



Seekugge Prasanna

Qalanders

36

33

4



Thisara Perera

Tigers

86

164

4



Wanindu Hasaranga

Arabians

18

36

1




If we referred to the chart earlier which showed team bowling average and team bowling economy, these group of all-rounders would be pretty close to the top-right hand corner - essentially as far away as it gets from being ideal.

The thing is, teams don't really need all-rounders in this format.  I'll re-post the chart I mentioned earlier, which shows the average balls faced by batting position in T10 in 2018:-


Batting Position

1

2

3

4

5

6

7

8

9

10

11

%

21.19

24.00

15.61

13.02

8.01

6.42

5.52

3.51

1.79

0.57

0.35

Balls per Innings

11.86

13.43

8.73

7.29

4.48

3.59

3.09

1.96

1.00

0.32

0.20


With the top four taking up 41.30 balls on average per innings (69% of balls faced) there really is no need for as much batting depth as in T20.  With numbers six and seven facing fewer than seven balls on average between them, and eight and nine facing fewer than three balls per innings combined, why would you waste these spots on all-rounders?  All you really need at seven really is a bowler who is capable of decent boundary-hitting, of which there are plenty out there.

Ideal Team Balance

Considering the above, I perceive that the ideal balance for a T10 team would be a top five of specialist boundary-hitting orientated batsmen including a wicket-keeper (if one of these is an above-average bowler, all the better for team options and flexibility) with an all-rounder at six (who must also be an above-average boundary-hitter).  Then five specialist bowlers at 7-11.  These positions combined face on average around 6.5 balls per innings, so there really is no need for batting depth - if number seven, as detailed above, can be a decent boundary-hitter and score something like 6(3) he's really done all that can be asked of him.  Teams in T10 should not relying on positions 7-11 to set a par score or chase a tough target.

This type of team balance covers the vast majority of expected balls faced with theoretical high expectation, while providing a minimum of six viable bowling options.  All it has done is eradicate the unnecessary plethora of all-rounders which teams still appear to want in such a shortened format - this approach from teams is utterly illogical and flies in the face of all statistical evidence.

As regular readers will be aware, I've mentioned numerous times that I feel that the usage of data is vital with regards to maximising expectation from cricket teams, and in an ideal world this would be in conjunction with the knowledge and ideas of a top-level coaching team.  However, I'm also strong on the fact that a squad recruited solely via data would perform much better than a squad recruited solely by the potentially biased 'eye test' in the long term, and the data I've discussed in this piece illustrates as such.


If this article has given you insight into the data that Sports Analytics Advantage can offer cricket teams around the world in formulating team strategies, selection, draft or auction plans, or any other work, please feel free to enquire at sportsanalyticsadvantage@gmail.com.
Comments