See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/227470350

Duckworth–Lewis and Twenty20 cricket

Article in Journal of the Operational Research Society · November 2011

DOI: 10.1057/jors.2010.175 · Source: RePEc

CITATIONS READS

19 5,015

3 authors:

Rianka Bhattacharya Paramjit S. Gill

New Jersey Institute of Technology University of British Columbia - Okanagan
2 PUBLICATIONS 21 CITATIONS 37 PUBLICATIONS 616 CITATIONS

SEE PROFILE SEE PROFILE

Tim Swartz
Simon Fraser University
111 PUBLICATIONS 1,725 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Social Networks Analysis View project

Stylometry View project

All content following this page was uploaded by Tim Swartz on 07 April 2015.

The user has requested enhancement of the downloaded file.

Journal of the Operational Research Society (2011) 62, 1951–1957 © 2011 Operational Research Society Ltd. All rights reserved. 0160-5682/11
www.palgrave-journals.com/jors/

Duckworth–Lewis and Twenty20 cricket

1 2 1
R Bhattacharya , PS Gill and TB Swartz
1 2
Simon Fraser University, Burnaby, British Columbia, Canada; and University of British Columbia,
Okanagan, Kelowna, British Columbia, Canada
Originally designed for 1-day cricket, this paper considers the use of the Duckworth–Lewis method as
an approach to resetting targets in interrupted Twenty20 cricket matches. The Duckworth–Lewis table
is reviewed and an alternative resource table is presented. The alternative table is constructed using
observed scoring rates from international Twenty20 matches. A desideratum of a resource table is
monotonicity in both the rows and columns corresponding to wickets and overs respectively.
Consequently, a Gibbs sampling scheme related to isotonic regression is applied to the observed
scoring rates to provide a non-parametric resource table. Taking into account the more aggressive
batting style of Twenty20 compared to 1-day cricket, the resultant resource table is seen to possess
sensible features. A discussion is provided concerning the use of the Duckworth–Lewis method applied
to Twenty20.
Journal of the Operational Research Society (2011) 62, 1951–1957. doi:10.1057/jors.2010.175
Published online 8 December 2010

Keywords: Duckworth–Lewis method; Gibbs sampling; isotonic regression; Twenty20 cricket

1. Introduction determination of the target using resources is known as the

Duckworth–Lewis method. What makes the adoption of
When considering the substantial amount of research that
the Duckworth–Lewis method remarkable is that the
has been directed toward the sporting world from a
method is widely perceived by the public as a black box
mathematical, statistical and operational research perspec-
procedure. Generally, people do not understand how the
tive, the Duckworth–Lewis method (Duckworth and
targets are set but they do agree that the targets are sensible
Lewis, 1998, 2004) perhaps stands alone as the most
or at least preferable to the approach based on run rates.
significant contribution to sport. Prior to the adoption of
Historically, we note that although the Duckworth–
the Duckworth–Lewis method, the resetting of targets in
Lewis method has gained worldwide prominence, there
interrupted 1-day cricket matches was carried out via run
have been other attempts at establishing targets for
rates. The difficulty with run rates is that targets are
interrupted 1-day matches. For example, for a brief period
determined by taking the remaining overs into account,
including the 1992 World Cup, the team batting second
while ignoring the number of lost wickets. As is well
had its target reduced from the first innings total by n runs.
known, batsmen tend to bat less aggressively and score
The quantity n was determined as the number of runs
fewer runs when more wickets have been taken. The
scored in the first innings in the corresponding number of
Duckworth–Lewis method was utilized and gained promi-
lost overs that had the least number of runs scored. This
nence during the 1999 World Cup of Cricket, and since
approach was immediately recognized as unfair and
that time, it has been adopted by every major cricketing
advantageous to the team batting first. Another short-
board and competition. In 1-day cricket, the Duckworth–
lived approach was based on a modification of the previous
Lewis method is based on the recognition that at the
system by further reducing the target by 0.5% for each
beginning of a match, each side has resources available
over lost. It too was generally seen as advantageous to the
(typically 50 overs and 10 wickets). When the match is
team batting first.
shortened, the resources of one or both teams are reduced
There have been other proposals that have never been
and the two teams usually have different resources for
implemented. For example, Clarke (1988) developed a
their innings. In this case, in an attempt to be fair, a
dynamic programming model where a target could be set
revised target for the team batting second is set. The
such that the probability of winning prior to the
interruption is equal to the probability of winning after
 the interruption. Christos (1998) proposed an alternative
Correspondence: TB Swartz, Department of Statistics and Actuarial
Science, Simon Fraser University, 8888 University Drive, Burnaby, BC
method based on run rates but where the number of
V5A1S6, Canada. wickets available is reduced proportional to the number of
E-mail: [email protected] overs made available.
1952 Journal of the Operational Research Society Vol. 62, No. 11

Although the Duckworth–Lewis method was designed that it is easily interpretable for Twenty20. Some comments
for 1-day cricket, it has also been applied to Twenty20 are provided on aspects of the table. In Section 3, an
cricket. Twenty20 is a relatively new version of limited alternative Twenty20 resource table is obtained using a
overs cricket with only 20 overs per side. In contrast to the non-parametric approach based on Gibbs sampling. The
1-day game and first-class cricket (which can take up to data used in the construction of the new table consist of
5 days to complete), Twenty20 matches have completion all international Twenty20 matches to date involving
times that are comparable to other popular team sports. nations from the International Cricket Council (ICC).
With the introduction of the biennial World Twenty20 We conclude with a short discussion in Section 4. A heat
tournament in 2007 and the Indian Premier League in map is provided to facilitate comparisons between the
2008, Twenty20 cricket has gained widespread popularity. two tables.
Although Twenty20 cricket is similar to 1-day cricket,
there exist subtle variations in the rules (eg fielding
restrictions, limits on bowling, etc) between the two 2. The Duckworth–Lewis resource table
versions of cricket. The variations in the rules, and most
importantly, the reduction of overs from 50 to 20 suggest In Table 1, we provide an abbreviated version of the
that scoring patterns in Twenty20 may differ from the Duckworth–Lewis resource table (Standard Edition) taken
1-day game. In particular, Twenty20 is seen as a more from the 2008–2009 ICC Playing Handbook found at
explosive game where the ability to score 4’s and 6’s is more www.icc-cricket.com. Note that in a full innings of 1-day
highly valued than in 1-day cricket. Since the Duckworth– cricket, a team begins batting with 100% of its resources
Lewis method (and its associated resource table) is based available corresponding to 50 overs and zero wickets taken.
on the scoring patterns in 1-day cricket, it is therefore As a simple example of the use of the Duckworth–Lewis
reasonable to ask whether the Duckworth Lewis method is resource table, consider a 1-day match where the team
appropriate for Twenty20. This is the focus of our paper. batting first scores 256 runs upon completion of its innings.
With the rise of Twenty20, an investigation of the It then rains prior to the resumption of the match. Due to
Duckworth–Lewis method applied to Twenty20 is timely. the lost time, suppose that the team batting second receives
Up until this point in time, such an investigation might not only 30 overs for its innings. According to the resource
have been possible due to the dearth of Twenty20 match table, the team batting second has only 75.1% of its
results. We now have at our disposal nearly 100 interna- resources available, and therefore its target for winning
tional matches, and through the use of efficient estimation the match is set at 256(0.751) ¼ 94 runs. We contrast the
procedures, the question may be at least partially Duckworth–Lewis target with the unreasonably low target
addressed. Also, since Twenty20 matches have a shorter of 256(30/50) ¼ 154 runs based on run rates.
duration, to date, very few matches have been interrupted Citing reasons of commerical confidentiality, Duckworth
and resumed according to Duckworth–Lewis. Conse- and Lewis (1998) provide only partial information con-
quently, if there is a problem with Duckworth–Lewis cerning the construction of the resource table. However,
applied to Twenty20, it may not have yet manifested itself. they do disclose that the table entries are based on the
Until a controversial outcome occurs, there may not be estimation of the 20 parameters Z0(w) and b(w), w ¼ 0, . . . , 9
sufficient motivation to adjust the table. corresponding to the function
In Section 2, we review the construction of the
Duckworth–Lewis resource table and scale the table so Zðu; wÞ ¼ Z0 ðwÞ½1  expfbðwÞug ð1Þ

Table 1 Abbreviated version of the Duckworth–Lewis resource table (Standard Edition)

Overs available Wickets lost

0 1 2 3 4 5 6 7 8 9

50 100.0 93.4 85.1 74.9 62.7 49.0 34.9 22.0 11.9 4.7
40 89.3 84.2 77.8 69.6 59.5 47.6 34.6 22.0 11.9 4.7
30 75.1 71.8 67.3 61.6 54.1 44.7 33.6 21.8 11.9 4.7
25 66.5 63.9 60.5 56.0 50.0 42.2 32.6 21.6 11.9 4.7
20 56.6 54.8 52.4 49.1 44.6 38.6 30.8 21.2 11.9 4.7
10 32.1 31.6 30.8 29.8 28.3 26.1 22.8 17.9 11.4 4.7
5 17.2 17.0 16.8 16.5 16.1 15.4 14.3 12.5 9.4 4.6
1 3.6 3.6 3.6 3.6 3.6 3.5 3.5 3.4 3.2 2.5
0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
The table entries indicate the percentage of resources remaining in a match with the specified number of wickets lost and overs available.
 R Bhattacharya et al—Duckworth–Lewis and Twenty20 cricket 1953

Table 2 The Duckworth–Lewis resource table (Standard Edition) scaled for Twenty20
Overs available Wickets lost

0 1 2 3 4 5 6 7 8 9

20 100.0 96.8 92.6 86.7 78.8 68.2 54.4 37.5 21.3 8.3
19 96.1 93.3 89.2 83.9 76.7 66.6 53.5 37.3 21.0 8.3
18 92.2 89.6 85.9 81.1 74.2 65.0 52.7 36.9 21.0 8.3
17 88.2 85.7 82.5 77.9 71.7 63.3 51.6 36.6 21.0 8.3
16 84.1 81.8 79.0 74.7 69.1 61.3 50.4 36.2 20.8 8.3
15 79.9 77.9 75.3 71.6 66.4 59.2 49.1 35.7 20.8 8.3
14 75.4 73.7 71.4 68.0 63.4 56.9 47.7 35.2 20.8 8.3
13 71.0 69.4 67.3 64.5 60.4 54.4 46.1 34.5 20.7 8.3
12 66.4 65.0 63.3 60.6 57.1 51.9 44.3 33.6 20.5 8.3
11 61.7 60.4 59.0 56.7 53.7 49.1 42.4 32.7 20.3 8.3
10 56.7 55.8 54.4 52.7 50.0 46.1 40.3 31.6 20.1 8.3
9 51.8 51.1 49.8 48.4 46.1 42.8 37.8 30.2 19.8 8.3
8 46.6 45.9 45.1 43.8 42.0 39.4 35.2 28.6 19.3 8.3
7 41.3 40.8 40.1 39.2 37.8 35.5 32.2 26.9 18.6 8.3
6 35.9 35.5 35.0 34.3 33.2 31.4 29.0 24.6 17.8 8.1
5 30.4 30.0 29.7 29.2 28.4 27.2 25.3 22.1 16.6 8.1
4 24.6 24.4 24.2 23.9 23.3 22.4 21.2 18.9 14.8 8.0
3 18.7 18.6 18.4 18.2 18.0 17.5 16.8 15.4 12.7 7.4
2 12.7 12.5 12.5 12.4 12.4 12.0 11.7 11.0 9.7 6.5
1 6.4 6.4 6.4 6.4 6.4 6.2 6.2 6.0 5.7 4.4
The table entries indicate the percentage of resources remaining in a match with the specified number of wickets lost and overs available.

where Z(u,w) is the average total score obtained in u overs 20 overs and we scale the entries so that an innings
in an unlimited overs match where w wickets have been beginning with 20 overs and zero wickets corresponds
taken. Although we readily admit the utility of the to 100% resources. Table 2 gives the full Duckworth–
Duckworth–Lewis table in 1-day cricket, a number of Lewis resource table (Standard Edition) for Twenty20
questions arise based on (1) and the estimates found in where the entries are obtained by dividing the correspond-
Table 1: ing entry in Table 1 by 0.566 (the resources remaining
in a 1-day match where 20 overs are available and zero
K There are many parametrics curves that could be fit. Is wickets taken).
(1) the best curve? Is there any advantage to a non-
parametric fit?
K The function (1) is based on 1-day rules but pertains to
3. A new resource table for twenty20
unlimited overs cricket. Since 1-day cricket is limited
overs cricket, is there an advantage in taking the In the construction of a resource table for Twenty20, it is
structure of the 1-day game into account? important to consider the scoring patterns specific to
K How are the parameters estimated? If the 10 curves Twenty20. For that reason, we consider all international
corresponding to w ¼ 0, . . . , 9 are fit separately, there Twenty20 matches involving ICC teams that have taken
are little data available beyond u ¼ 30 for fitting the place from 17 February 2005 through 9 November 2009.
curve with w ¼ 9. Also, the asymptotes for the curves There are 85 such matches and details concerning these
with w ¼ 0, 1, 2 (see Figure 1 of Duckworth and Lewis matches are available from www.cricinfo.com. Note that
(1998)) fall beyond the range of the data. we have excluded the four shortened matches where the
K In Table 1, the last two columns have many identical Duckworth–Lewis method was applied. We have also
entries going down the columns. Although very few excluded Twenty20 matches involving non-ICC nations as
matches occur under these conditions, is it really sensible we prefer to consider matches of a consistently high
for resources to remain constant as the available overs standard.
decrease? This is a consequence of the asymptote In the study of Twenty20 scoring patterns, we consider
imposed by (1). first innings data only, as scoring patterns in the second
innings are influenced by the number of runs scored in the
For ease of discussion, we find it convenient to convert first innings. In the development of a 1-day cricket
the Duckworth–Lewis resource table to the context of simulator, Swartz et al (2009) consider batting behaviour
Twenty20. Specifically, we truncate the resource table to in the second innings. Although match summary results are
1954 Journal of the Operational Research Society Vol. 62, No. 11

Table 3 The matrix R=(row) of estimated resources for Twenty20

Overs available Wickets lost

0 1 2 3 4 5 6 7 8 9

20 100.0
19 93.6 83.0 110.2
18 90.4 85.8 78.3
17 86.7 80.5 82.8 53.7
16 81.7 74.5 81.9 70.7 32.8
15 76.5 71.4 71.5 65.9 59.9
14 68.3 69.1 67.6 66.2 58.4
13 63.8 68.2 62.4 62.9 59.0 24.3
12 62.1 62.3 60.6 57.3 58.8 44.1
11 60.5 56.3 57.0 53.6 61.0 39.7
10 57.6 49.6 52.1 52.8 48.1 38.6 41.0 35.2
9 54.9 52.1 43.6 49.0 44.1 33.8 35.0 29.7
8 51.0 46.4 41.7 42.2 41.2 36.7 27.5 28.7
7 48.6 45.8 38.9 35.9 39.1 34.8 24.1 25.5
6 54.0 37.9 36.6 30.3 36.2 31.3 20.9 21.4 26.7
5 44.0 32.5 25.4 28.7 29.4 23.9 17.1 14.9
4 28.2 23.4 22.5 22.2 20.9 14.3 10.6 6.7
3 20.6 19.9 16.9 17.8 15.8 12.4 7.6 1.2
2 21.2 17.6 11.9 13.4 10.6 11.0 7.2 1.4
1 8.7 5.2 7.3 6.0 5.5 6.0 2.6
Missing entries correspond to match situations where data are unavailable.

readily available from the Cricinfo website, our investiga- down columns. We also observe some conspicuous entries
tion requires ball-by-ball data, and for this, we have coded in Table 3, particularly the entry of 110.2% resources
a Java script to parse the associated commentary log for corresponding to 19 overs available and two wickets taken.
each match. The script extracts the relevant details on a This entry is clearly misleading and should be less than
ball-by-ball basis, and stores the data in a tabular form for 100%. It arises due to the small sample size (two matches)
easy access. corresponding to the given situation. For our non-
For each match, define x(u, w(u)) as the runs scored from parametric resource table (upcoming), we have found that
the stage in the first innings where u overs are available and the estimation procedure is robust to observations based
w(u) wickets have been taken until the end of the first on small sample sizes as the surrounding observations
innings. We calculate x(u, w(u)) for all values of u that based on larger sample sizes have greater influence in the
occur in the first innings for each match beginning with determination of the table. We have therefore retained
u ¼ 20 and w(u) ¼ w(20) ¼ 0. conspicuous observations such as 110.2%. We view our
We then calculate the matrix R ¼ (ruw) where ruw is the investigation of Duckworth/Lewis in Twenty20 as one of
estimated percentage of resources remaining when u overs discovery rather than an attempt to replace the Duck-
are available and w wickets have been taken. We calculate worth/Lewis table.
(100%) ruw by averaging x(u, w(u)) over all matches where To impose the monotonicity constraints in the rows
w(u) ¼ w and dividing by the average of x(20, 0) over all and columns, we refer to the general problem of isotonic
matches. The denominator is simply the average number of regression. For our purposes, we consider the minimiz-
first inning runs over all matches. In the case of u ¼ 0, we ation of
set ruw ¼ r0w ¼ 0.0%. The matrix R is therefore a first X
attempt at a resource table for Twenty20 and is given in F¼ quw ðruw  yuw Þ2 ð2Þ
Table 3. Note that r20,0 ¼ 100% as desired. Although R is a
non-parametric estimate of resources and makes no with respect to the matrix Y ¼ ( yuw) where the double
assumptions concerning the scoring patterns in Twenty20, summation corresponds to u ¼ 1, . . . , 20 and w ¼ 0, . . . , 9,
it is less than ideal. First, there are many table entries the quw are weights and the minimization is subject to the
where there are missing data for the given situation. In constraints yuwXyu,w þ 1 and yu,wXyu1,w. In addition, we
addition, Table 3 does not exhibit the monotonicity that we impose y20,0 ¼ 100, y0,w ¼ 0 for w ¼ 0, . . . , 9 and yu,10 ¼ 0
expect. Logically, we require a resource table that is non- for u ¼ 1, . . . , 20.
decreasing as we go from left to right along rows and we Although the fitting of Y is completely non-parametric,
require a resource table that is non-decreasing as we go there are some arbitrary choices that have been made in the
 R Bhattacharya et al—Duckworth–Lewis and Twenty20 cricket 1955

minimization of (2). First, not only was the choice of adjacent entries in Table 4 that have the same value. Again,
‘squared error’ discrepancy in (2) convenient for computa- it is not sensible for resources to remain constant as
tion, minimization of the function F with squared error available overs decrease or wickets increase. The problem is
discrepancy corresponds to the method of constrained that in the minimization of (2), various fitted y’s occur on
maximum likelihood estimation where the data ruw are the boundaries imposed by the monotonicity constraints.
independently normally distributed with means yuw and Table 4 is also unsatisfactory from the point of view that
variances 1/quw. Second, we choose to consider a matrix Y: it is incomplete; missing values correspond to match
20  10 based on overs. Alternatively, we might have situations where data are unavailable.
considered a larger matrix Y: 120  10 based on balls. We To address the above criticisms, we now take a slightly
prefer the overs formulation as it involves less missing data different approach to estimation. As previously mentioned,
and leads to a less computationally intensive optimization. we recognize that (2) arises from the normal likelihood
With a matrix Y based on overs, it is possible to interpolate  
1X
on a ball-by-ball basis if required. Third, we have made a exp  quw ðruw  yuw Þ2 : ð3Þ
2
simple choice with respect to the weights quw. We set 1/quw
equal to the sample variance used in the calculation of We therefore consider a Bayesian model where the
ruw divided by the sample size. The rationale is that when unknown parameters in (3) are the y’s. A flat default prior
ruw is less variable, there is stronger belief that yuw should is assigned to the y’s subject to the monotonicity
be close to ruw. constraints. It follows that the posterior density takes the
In Table 4, we give a non-parametric resource table form (3) and that Gibbs sampling can be carried out via
based on the minimization of (2). An algorithm for isotonic sampling from the full conditional distributions
regression in two variables was first introduced by Dykstra
½ yuw j    Normal ðruw ; 1=quw Þ ð4Þ
and Robertson (1982). Fortran code was subsequently
developed by Bril et al (1984). We have used an R code subject to the local constraints on yuw in the given iteration
implementation that is available from the Iso package of the algorithm. Sampling from (4) is easily carried out
on the Cran website (www.cran.r-project.org). The pro- using a normal generator and rejection sampling according
gramme requires 27 iterations to achieve convergence. to the constraints. Although in statistical terminology,
What is unsatisfactory about Table 4 is that it suffers from (3) takes a parametric form, we refer to the approach as
the same criticism that was directed at the Duckworth– non-parametric since no functional relationship is imposed
Lewis resource table. There is a considerable number of on the y’s.

Table 4 A non-parametric resource table for Twenty20 based on isotonic regression

Overs available Wicket lost

0 1 2 3 4 5 6 7 8 9

20 100.0
19 93.6 85.5 85.5
18 90.4 85.5 80.8
17 86.7 80.8 80.8 64.7
16 81.7 77.4 77.4 64.7 55.9
15 76.5 71.5 71.5 64.7 55.9
14 68.8 68.8 67.6 64.7 55.9
13 66.6 66.6 62.6 62.6 55.9 38.4
12 62.2 62.2 60.6 57.3 55.9 38.4
11 60.5 56.8 56.8 54.8 54.8 38.4
10 57.6 52.1 52.1 52.1 48.1 38.4 34.1 29.3
9 54.9 52.1 46.5 46.5 44.1 36.3 34.1 29.3
8 51.0 46.4 42.0 42.0 41.2 36.3 28.6 28.6
7 48.6 45.8 38.9 37.3 37.3 34.8 25.3 25.3
6 39.7 39.7 36.6 32.8 32.8 31.3 23.0 21.4 21.4
5 39.7 32.5 28.0 28.0 28.0 23.0 17.1 15.5
4 27.9 23.4 22.5 22.2 20.9 14.3 10.7 10.7
3 20.7 19.9 17.4 17.4 15.8 12.4 7.7 7.7
2 20.7 17.6 12.5 12.5 10.8 10.8 7.2 1.8
1 8.7 6.6 6.6 6.0 5.7 5.7 1.8
The table entries indicate the percentage of resources remaining in a match with the specified number of wickets lost and overs available. Missing
entries correspond to match situations where data are unavailable.
1956 Journal of the Operational Research Society Vol. 62, No. 11

In Table 5, the estimated posterior means of the y’s stabilize after 50 000 iterations. For cases of missing data,
obtained through Gibbs sampling are given, and these we impute the missing r’s with the Duckworth-Lewis table
provide an alternative resource table for Twenty20. The entries. The imputation is in the spirit of a Bayesian
computations pose no difficulties and the estimates approach where prior information is utilized. Unlike
Table 4, note that we have a complete table. Also, we no
longer have adjacent table entries with identical values
20
and this is due to the sampling approach. Finally, we
19 remark that the methodology allows the input of expert
18 opinion. For example, suppose that there is expert
17 consensus that a table entry yij ought to be tied down to
16 a particular value a. To force this table entry, all that is
15 required is to set rij ¼ a and assign a sufficiently small
14 pffiffiffiffiffi
standard deviation 1= qij .
13

overs available
12
11
10 4. Discussion
9
8 Our intention now is to compare the scaled Duckworth–
7 Lewis resource table originally designed for 1-day cricket
6 (Table 2) with the alternative non-parametric resource
5 table based on Twenty20 matches (Table 5). To facilitate
4 the comparison, we take the absolute values of the
3
differences between the two tables, and produce a heat
2
1
map as given in Figure 1. The darker shades of the heat
0 1 2 3 4 5 6 7 8 9
map indicate the greatest disagreement between the two
wickets lost tables. We investigate these areas of disagreement.
From Figure 1, we observe that the greatest absolute
Figure 1 Heat map of the absolute differences between the
Duckworth-Lewis resource table (Table 2) and the non- differences occur in three regions. First, large differences
parametric resource table based on Gibbs sampling (Table 5). occur in the top-right hand corner and bottom-left hand
Darker shades indicate larger differences. corner of the table. These are precisely the regions where

Table 5 A non-parametric resource table for Twenty20 based on Gibbs sampling

Overs available Wickets lost

0 1 2 3 4 5 6 7 8 9

20 100.0 96.9 93.0 87.9 81.3 72.2 59.9 44.8 29.7 17.6
19 95.6 90.9 87.7 83.0 76.9 68.3 56.5 42.0 27.2 15.3
18 91.7 86.7 82.9 78.7 73.2 65.4 54.2 40.2 25.7 13.9
17 87.7 82.3 78.9 73.8 69.7 62.8 52.2 38.7 24.6 12.8
16 83.5 78.2 75.3 70.5 66.4 60.2 50.3 37.4 23.5 12.0
15 79.2 74.3 70.9 66.9 62.6 57.4 48.4 36.2 22.7 11.2
14 75.1 70.7 67.3 63.7 59.3 54.6 46.4 35.0 21.8 10.5
13 71.5 67.4 63.6 60.3 56.2 51.5 44.3 33.8 21.0 9.8
12 68.3 63.7 60.2 56.8 52.9 47.5 41.9 32.6 20.2 9.1
11 65.0 59.9 56.6 53.3 49.7 43.9 39.3 31.3 19.4 8.5
10 61.3 56.0 52.6 50.1 46.0 40.8 36.1 30.0 18.6 7.9
9 57.9 52.3 47.9 46.1 42.5 37.8 33.1 28.3 17.7 7.2
8 54.0 48.3 44.3 41.7 38.9 34.9 30.2 26.1 16.7 6.6
7 49.3 44.2 40.2 37.4 35.4 32.1 27.2 23.4 15.7 5.9
6 41.7 38.5 35.7 33.0 31.7 29.0 24.2 20.0 14.5 5.2
5 36.2 33.4 31.0 28.6 27.3 25.5 21.5 17.0 12.2 4.4
4 30.8 28.0 26.1 24.1 22.4 20.7 18.3 14.2 10.0 3.5
3 25.4 22.8 21.1 19.4 17.7 16.5 14.4 11.6 7.9 2.5
2 19.7 17.2 15.5 14.1 12.7 11.9 10.6 9.3 6.2 1.6
1 13.7 11.3 9.7 8.5 7.3 6.7 6.0 5.2 4.2 0.9
The table entries indicate the percentage of resources remaining in a match with the specified number of wickets lost and overs available.
 R Bhattacharya et al—Duckworth–Lewis and Twenty20 cricket 1957

very little or no data are available. We do not view these scoring an impressive 191 runs at the expense of five
regions as too important as the resetting of targets would wickets. England’s captain Paul Collingwood commented
rarely use these entries. It is interesting however that the ‘Ninety-five percent of the time when you get 191 runs on
non-parametric approach (Table 5) provides more re- the board you are going to win the game’. In this
sources in these regions than the Duckworth/Lewis match, rain interrupted the second innings after 2.2 overs.
approach. Consider a Twenty20 match with a single over When the match resumed, the Professional Edition of
remaining and two wickets lost. In a match that averages Duckworth-Lewis provided a target of 60 runs from 6
150 runs, Duckworth/Lewis suggests that (0.064)150 ¼ 9.6 overs. The West Indies reached the target in 5.5 overs and
runs is expected in the final over for an average of won the match amidst great complaints (http://www
9.6/6 ¼ 1.6 runs per ball. On the other hand, Table 5 .cricinfo.com/world-twenty20-2010/content/current/story/
suggests (0.097)150 ¼ 14.6 runs in the final over for an 458375.html). Had our Table 5 been used, there would
average of 14.6/6 ¼ 2.4 runs per ball. According to our have been 41.7% resources available from 6 overs and
intuition, in this situation with an extremely aggressive and this leads to a higher and more reasonable target of
talented batsmen, 2.4 runs per ball may be reasonable. (0.417)191 ¼ 80 runs.
The more interesting discrepancy occurs in the ‘middle’ It is important to emphasize that we do not endorse our
of an innings (8–13 overs available with 3–6 wickets lost). non-parametric resource table (Table 5) as a replacement
In this stage of an innings, the non-parametric approach for the Duckworth–Lewis resource table in Twenty20.
based on Gibbs sampling (Table 5) suggests that there is up Our resource table is based on only 85 matches, too small
to 5% fewer resources remaining than provided by the a sample to provide confident table entries. However,
Duckworth–Lewis method. This coincides with our intui- we believe that our table does suggest that there may be
tion as we believe that up to this stage in an innings, some meaningful differences between the scoring rates in
batting is more aggressive in Twenty20 than in 1-day 1-day cricket and Twenty20 cricket. As more Twenty20
cricket. Recall that in 1-day cricket, a team needs to protect matches become available, we endorse a review of the use
its wickets over a longer period of overs. Consequently, up of Duckworth–Lewis in Twenty20 and the estimation
until the middle stage, more resources are conserved in the techniques used in the construction of the associated
1-day game than in Twenty20. We remark that a difference resource table.
of 5% resources may be very meaningful as a target of 240
runs diminished by 5% gives 228 runs. Acknowledgements —Gill and Swartz have been partially supported by
grants from the Natural Sciences and Engineering Research Council of
We revisit one of the four applications of Duckworth– Canada. The authors thank a referee whose comments have helped
Lewis in international Twenty20, which occurred in a improve the manuscript.
match between England and the West Indies during the
2009 World Cup. This was the crunch game of the
tournament for a place in the semi-finals. England scored References
161 run at the expense of six wickets in the full first innings. Bril G, Dykstra RL, Pillers C and Robertson T (1984). Isotonic
The second innings was shortened to 9 overs with a target regression in two independent variables. J Roy Stat Soc C 33:
of 80 runs for the West Indies. West Indies scored 82 runs 352–357.
in 8.2 overs to eliminate England from the tournament. Christos GA (1998). It’s just not cricket. In: de Mestre N and
The English fans were upset and the Guardian claimed that Kumar K (eds). Mathematics and Computers in Sport. Bond
University: Queensland, Australia, pp 181–188.
the Duckworth–Lewis system will be reviewed to take Clarke SR (1988). Dynamic programming in one-day cricket—
into account Twenty20 matches (http://www.cricinfo.com/ Optimal scoring rates. J Opl Res Soc 39: 331–337.
ci-icc/content/story/409482.html). In this match, we note Duckworth FC and Lewis AJ (1998). A fair method for resetting
that the Professional Edition of Duckworth-Lewis the target in interrupted one-day cricket matches. J Opl Res Soc
(which is not available from the 2008–2009 ICC Playing 49: 220–227.
Duckworth FC and Lewis AJ (2004). A successful operational
Handbook) was used to set the target at 80. For research intervention in one-day cricket. J Opl Res Soc 55:
comparison, we refer to Table 2 (based on the Standard 749–759.
Edition) that sets a target of (0.518)161 ¼ 84 runs. The non- Dykstra RL and Robertson T (1982). An algorithm for isotonic
parametric resource table (Table 5) gives an even higher regression for two or more independent variables. Ann Stat 10:
target of (0.579)161 ¼ 94 runs. 708–716.
Swartz TB, Gill PS and Muthukumarana S (2009). Modelling and
As a second example of the implementation of Duck- simulation for one-day cricket. Can J Stat 37: 143–160.
worth–Lewis in Twenty20, we consider another contro-
versial match that occurred after our data collection phase,
and again involved England and the West Indies. On 3 Received November 2009;
May, in the 2010 World Cup, England batted first, accepted September 2010

View publication stats