You are on page 1of 12

Journal of the Operational Research Society (2005) 56, 804–815 r 2005 Operational Research Society Ltd.

td. All rights reserved. 0160-5682/05 $30.00

www.palgrave-journals.com/jors

Towards fairer measures of player performance


in one-day cricket
AJ Lewis*
Oxford Brookes University, UK
Traditionally, evaluation of individual player performances in one-day cricket has been based on measures such as
batting and bowling averages, and strike and economy rates. It is recognized, within the game of cricket, that such
measures have severe limitations in assessing the true performances and abilities of players. Whenever a player’s
performance statistics are quoted, there is nearly always some verbal qualification that is necessary in order to place the
statistics into context. Such comments include the stage of an innings a player usually bats or bowls, as this significantly
affects a player’s opportunity for scoring runs or taking wickets. Further, traditional measures do not allow the
comparison of the skills of batting and bowling as they are based on incompatible scales. This paper uses the well-
established Duckworth/Lewis methodology to create alternative measures of player performance. These measures, it is
argued, have the potential to reflect better the performances of players having regard for the stages of innings that runs
are scored or conceded and wickets taken or lost. The proposed measures could impact in evaluating player
performance for a particular match, a series of matches, or in the longer term over players’ careers. Further, the
methodology enables an objective measure to be created that compares and combines the performances of batsmen and
bowlers and of a player’s all-round ability. Data from international matches are used to illustrate the methodology and
how it compares with existing measures.
Journal of the Operational Research Society (2005) 56, 804–815. doi:10.1057/palgrave.jors.2601876
Published online 10 November 2004

Keywords: sports statistics; cricket; modelling

Introduction To illustrate the issues to be addressed in this paper, the


current bowling measure of economy rate uses the runs
The game of cricket is a sporting statistician’s dream. Over
conceded per over delivered (which consists of six balls). The
the decades, summary statistics have been produced and
bowling average uses runs conceded per wicket taken. But
published in many sporting chronicles, the most famous of
these measures make no allowance for the stage of the
which1 is known as ‘Wisden’. This publication, among many
innings that balls are delivered, wickets are taken or of runs
facts and figures, produces summary statistics, such as the
that are conceded. Whether a wicket is taken with the first or
number of runs, wickets and catches in the season just past,
last ball of the innings is not taken into account. Similarly,
for players achieving first-class status. It is widely recog-
the stage of the innings at which runs are earned and
nized, however, that such statistics only tell part of the story
conceded is not taken into consideration. Similar comments
of a player’s performance. One-day cricket is particularly
apply to the batting measures of performance such as strike
prone to distortions in the value of players’ contributions.
rate and batting average.
This is because of its shorter duration compared with
The second Victoria Bitter (VB) Series final on 25 January
traditional cricket and because the nature of the game 2003 between Australia and England, a typical example of a
requires that the good of the team be placed ahead of one-day international (ODI) of 50 overs per side, will be
individual performance. used to illustrate some of the issues. The scorecard for this
It will be assumed in this paper that the reader is familiar match can be seen in Engels1 and on the Internet. (http://
with the game of cricket and its terminology. A brief uk.cricinfo.com/link_to_database/ARCHIVE/2002-03/OD_
explanation is provided in Duckworth and Lewis.2 More TOURNEYS/VBS/SCORECARDS/AUS_ENG_VBS_OD
extensive explanations can be seen in Engels1 and on the I-FINAL2_25JAN2003.html accessed 20 July 2004.) Some
Internet. (http://www.cricket.org/link_to_database/SOCIE selected details, and figures relevant to later discussion, are
TIES/ENG/ACS/CRICKET_HISTORY/LAWS/index.html provided in Table 1. In this match, Hayden scored 69 runs in
accessed 20 July 2004.) 91 balls, which is a strike rate of 75.82 per hundred balls.
Hogg scored 71 runs in 77 balls and was not out. His strike
rate was 92.20. (Note that cricketing measures of perfor-
*Correspondence: AJ Lewis, The Business School, Oxford Brookes
University, Wheatley, Oxford OX33 1HX, UK. mance are traditionally rounded down to two decimal
E-mail: ajlewis@brookes.ac.uk places.) Hayden’s innings lasted for 40 overs and three balls
AJ Lewis—Player performance in one-day cricket 805

Table 1 Australia’s innings—batting measures


Batsman Runs Out? Balls faced Strike rate Net batting contribution Batting resources Batting resource average

Hogg 71 0 77 92.20 þ 16.90 23.4 3.03


Hayden 69 1 91 75.82 þ 21.39 20.6 3.34
Gilchrist 26 1 32 81.25 þ 0.99 10.8 2.40
Lee 18 1 17 105.88 þ 2.70 6.6 2.72
Martyn 11 1 14 78.57 10.45 9.3 1.18
Bichel 11 0 10 110.00 þ 0.81 4.4 2.50
Bevan 10 0 21 47.61 þ 2.78 3.1 3.22
Symonds 8 1 29 27.58 16.95 10.8 0.74
Ponting 1 1 8 12.50 16.78 7.7 0.12
Warne 0 1 1 0.00 7.38 3.2 0.00
Williams 0 0
Byes & Leg Byes 3
Wides & No balls 1

Totals 229 7 6.0 99.9

(written as 40.3) of the 50 overs. Hogg batted during the last Data (Australia) and the Press Association (Sports) in the
21.3 overs, received the very last ball of the 50 overs and did United Kingdom.
not lose his wicket. Of the two batsmen who has performed
the better? Based on runs for their wicket and strike rate,
Hogg comes out on top. But this ranking ignores the stage of Use of ball-by-ball information
the innings that the runs were scored. Hayden was one of the
opening batsmen and needed to ensure that wickets were not Ball-by-ball information can be utilized in two ways through
given away cheaply at the beginning of the innings thereby the D/L methodology. One way, as first outlined by the
avoiding pressure on the later batsmen. On the other hand, author in 2003,4 is to assess the contribution in terms of the
Hogg had the ‘luxury’ of knowing that, particularly in the net runs gained by batsmen, and in terms of net runs
last few overs, he could take inordinate risks to boost the conceded by bowlers. More extensive details of this
Australian total, which became 229. procedure are provided below. This utilization captures the
It can be argued that although Hayden’s two measures of zero-sum nature of the game of cricket. This measure will
performance (strike rate and runs scored per wicket lost) also provide the mechanism by which performances by
were below those of Hogg, in fact Hayden’s performance is batsmen and bowlers can be compared or combined, a
the more meritorious having regard for the relative match characteristic absent from existing measures of performance.
circumstances of their two performances. A second way of using D/L methodology, as previously
Such issues as these are seen as shortcomings in the outlined by the author,4 is to assess runs earned or conceded
traditional performance measures in evaluating the worth of in relation to the resources consumed by batsmen or
a player’s contribution to his team’s results. contributed by bowlers. By coincidence, Swartz and
The Duckworth/Lewis (D/L) methodology2,3 provides the Beaudoin5 presented their outline of a very similar procedure
ability to compile mechanisms that evaluate a player’s to that of the author at the same conference in 2003 and
performance in relation to runs expected for the stage of the upon which they subsequently expanded in Beaudoin and
innings and in relation to the resources that are consumed Swartz.6
when batting, or contributed when bowling. In order to use These separate measures are expected to provide a
the D/L methodology more detailed information is required counter to the criticisms of the standard performance
than is supplied by a standard summary scorecard as measures of batting and bowling performance of averages,
published in Engels;1 the details on the outcome of every ball and strike and economy rates. These proposed measures will
that is bowled are required. This detail is now becoming recognize that, in the later stages of an innings in particular,
available initially through verbal summaries such as is seen the resources consumed or contributed by the loss of a
for the ODI previously cited. (http://uk.cricinfo.com/link_ wicket are small and that the number of runs earned or
to_database/ARCHIVE/2002-03/OD_TOURNEYS/VBS/S conceded are adjusted in relative importance for the stage of
CORECARDS/AUS_ENG_VBS_ODI-FINAL2_25JAN the innings. They will recognize, also, that to concede no
2003_BBB-COMMS.html accessed 20 July 2004.) More runs off the final ball is equivalently meritorious to taking a
recently these details, in electronic form, are now becoming wicket; there is no credit earned or lost for a ‘sacrificed’
available through commercial companies such as Champion wicket off the last ball.
806 Journal of the Operational Research Society Vol. 56, No. 7

Net contribution of batsmen and bowlers Early in a team’s innings, because expectation of run
7 scoring is less than a run a ball, a single run will make a
Johnston et al proposed a player evaluation mechanism
positive contribution. But towards the end of an innings,
using Clarke’s dynamic programming model in one-day
with wickets in hand, a single run will result in a negative
cricket.8 There has been no subsequent development of these
contribution as more than a run a ball is then expected from
proposals. The proposal in this paper, which expands
the model in (1).
considerably on the author’s earlier outline,4 is that the
At first sight, the bowler’s contribution to the team effort
Duckworth/Lewis model2,3 can be used to evaluate the
would seem to be able to be evaluated as the negative of the
contribution a player makes compared with that expected
batsman’s net contribution since cricket is a zero-sum game.
for the stage of the innings the contribution is made.
However, because ‘extras’ are treated unequally between
batsman and bowler, there needs to be a slight adjustment to
The Duckworth/Lewis (D/L) average runs model the contributions of both players in order that the total net
contributions of completed innings is indeed zero.
The D/L model of one-day cricket2,3 is
There are several forms of extras. The bye and leg bye
Zðu; wÞ ¼ Zo FðwÞf1  exp½bu=FðwÞg ð1Þ count neither for the batsman nor against the bowler yet are
aggregated with other runs to form the total for the team.
In this model, Z(u,w) represents the average further runs The wide ball and no ball are debited against the bowler,
obtained in the u remaining overs when w wickets have been added to the batting side’s score but not credited to the
lost, where Zo and b are positive constants. F(w) is a positive batsman facing at the time. It is possible that a batsman also
decreasing step function with F(0) ¼ 1. This function is scores runs at a no ball but not at a wide ball. Such balls do
interpreted as the proportion of runs that are scored with w not count in the total of number of balls bowled in the
wickets lost compared with that with no wickets lost and, innings to date. Consequently, the ball is bowled again until
hypothetically, infinitely many overs available. That is it is legitimate during which time several runs may be added
F(w) ¼ limu-NZ(u, w)/Z(u,0). As subsequent discussion will to the team total and to the batsmen’s individual totals. In
focus on the ball-by-ball contribution of players then, since the following analysis, the expected runs contribution by the
there are six balls in every over, writing i ¼ 6u, (0pip300) batsmen is zero for the wide and no ball. Similarly the
will represent the ith remaining ball of the innings. expected number of runs conceded by the bowler is zero for
At any stage of an innings, the worth of a player’s such deliveries. The corresponding expectations of batsmen
contribution per ball can be evaluated using (1). If there are i and bowlers are applicable only at the delivery that makes
balls remaining and w wickets have been lost then the the ball legitimate.
expected runs, ri, from the next ball will be either

ri ¼ Zði; wÞ  Zði  1; wÞ ð2aÞ Adjusting for extras


or Let ei be the number of extras conceded by the ith ball
remaining (1pip300). Of these, let gi be the combined
ri ¼ Zði; wÞ  Zði  1; w þ 1Þ ð2bÞ number of byes and leg byes and hi be the combined number
of wides and no balls. Thus, ei ¼ gi þ hi and the team’s total
depending, respectively, on whether the batsman survives the
is S ¼ Ssi þ Sei ¼ Ssi þ Sgi þ Shi, the summations being over
ball or loses his wicket.
all 300 balls of the innings. For an average ODI, the total
When i is not much less than 300, which is the standard
score of Team 1 is Z(50,0) ¼ 235. (http://www.cricinfo.com/
length of an ODI innings, and w ¼ 0, then (2a) will be small
db/ABOUT_CRICKET/RAIN_RULES/DUCKWORTH_
and (2b) will be large. These relative orders of magnitude
LEWIS_2002.html accessed 29 January 2004.) If S4Z(50, 0)
reflect the high value of dismissing a top-order batsman early
then the batting team have exceeded expectations from the
in an innings. Conversely as i approaches zero (2a) will be
D/L model and conversely if SoZ(50,0). If there were Sei
relatively large and (2b) will be relatively small provided w is
extras in a particular innings then the batsmen have
not approaching 9, which is the situation when the last men
performed above average if the sum of the batsmen’s
are then at the wicket.
contributions exceeds Z(50,0)Sei. Thus, the contributions
If the batsman scores si runs from that ball then, using
of batsmen, adjusted for the effect of extras, for the ith ball
(2a) and (2b), his net contribution ci, for that ball, is either P
remaining are achieved from ci ¼ siri {Z(50,0) ei}/
P
ci ¼ si  fZði; wÞ  Zði  1; wÞg ð3aÞ Z(50, 0). Since the average of the total runs, ri, is
P
Z(50, 0) it is easily seen that ci ¼ SZ(50,0).
or In the main, the net contribution for the bowler is the
ci ¼ si  fZði; wÞ  Zði  1; w þ 1Þg ð3bÞ negative of that for the batsman facing the same ball, except
that extras weigh more heavily against bowlers. The extras
depending on whether or not the batsman survives that ball. not counting against the bowler, in a particular innings, are
AJ Lewis—Player performance in one-day cricket 807

P P
represented by Sgi. Thus, the net contribution by a bowler resource average) can be assessed by si/ pi where the
on any ball, adjusted for the effect of extras, is represented sums are over the number of balls that the batsman faced.
P
by ci ¼ ri fZð50; 0Þ  gi g=Zð50; 0Þ  ðsi þ hi Þ. It is again The bowlers’ performances are evaluated using the total of
easily seen that the total of the net contributions of the runs, no balls and wides that are conceded. A bowler’s
the bowlers over the 300 balls of the innings is average run contribution per unit of resource contributed
P P P
ci ¼ Zð50; 0Þ  S. Consequently, ci þ ci ¼ 0 so (shortened later to [bowling] resource average) is measured
P P
that, after allowance for the different treatment of extras by (si þ hi)/ pi where the sums are over the number of
between batsmen and bowlers, the zero-sum nature of the balls the bowler has bowled.
game is captured. For the batsman the higher the resource average the better
If an innings is not completed, because the team batting the performance. For the bowler the lower the resource
second has won the match within their 300 balls, then average the better the performance.
P P
ci þ ci may differ slightly from zero. This discrepancy Beaudoin and Swartz5,6 use the D/L methodology in the
is ignored in subsequent discussion but the effect on same way but multiply the two ratios obtained by 100 and
conclusions is likely to be negligible. call the results ‘runs per match’ (RM). They interpret their
RM measures as the average performance of the batsman or
Average runs for resources consumed and resources bowler if he were to do all of the team’s batting or bowling
contributed himself. A batting RM above 235 would indicate a batting
performance above average whereas, for a bowler, a RM
The D/L methodology2,3 converts the Z(u, w) in (1) into the below 235 would represent an above average performance.
proportion of the average total in 50 overs, Z(50,0) to
produce
Comparing and combining batting and bowling
Pðu; wÞ ¼ Zðu; wÞ=Zð50; 0Þ ð4Þ
performance measures
This function is tabulated in percentages and used for Traditional measures of performance of batting and bowling
target resetting purposes. (http://www.cricket.org/link_ are incompatible. These offer no objective way to evaluate
to_database/NATIONAL/ICC/RULES/ODI_PLAYING_ the worth of, say, a batsman scoring 100 runs against the
CONDITIONS_D-L_METHOD.pdf accessed 20 July worth of a bowler taking, say, four or five wickets for not
2004.) The function is interpreted as the percentage of very many runs in an innings. Similarly, over a series of
resources (relative to those for a 50-over innings) that a team matches comparing a batting average of, say, 55 with a
has remaining for the u overs it has left when w wickets have bowling average of, say, 20 cannot be undertaken objec-
been lost. tively. Further, these traditional measures offer no way of
Again writing i ¼ 6u, the consumption of resource, pi, for aggregating a player’s performance in batting and bowling
the ith ball remaining is represented by either either for an individual match, over a series of matches, or
pi ¼ Pði; wÞ  Pði  1; wÞ ð5aÞ even over a player’s career.

or
Net contributions
pi ¼ Pði; wÞ  Pði  1; w þ 1Þ ð5bÞ The net batting and bowling contributions, proposed in this
depending, respectively, on whether or not the batsman paper, have a common ratio scale on which better
survives that ball. performances of both are represented by larger values. This
Comments that were made previously in relation to the enables batting and bowling performances to be compared
effects of extras on net contributions are relevant, in a directly and objectively. Further, they provide a simple and
similar way, in the discussion of resource percentages. objective way of combining a player’s performances in
Resources are not consumed by the batsman nor contributed batting and bowling to produce a measure of all-round
by the bowler for a no ball or a wide but these forms of performance. A simple addition is all that is necessary.
extras are debited against the bowler. Runs made by the
batsman off the no ball are credited to the batsman and Resource averages and runs per match
debited against the bowler.
The author’s two resource averages and the corresponding
RM measures of Beaudoin and Swartz do not assist in being
Measure of resource utilization
able to compare batsmen with bowlers objectively since good
For any ball i, the batsman scores si runs and consumes pi performances occur at opposite ends of the scale. Beaudoin
resources. His average run contribution, per unit of resource and Swartz6 suggest, however, that their two RM measures
consumed, to the team’s total (shortened later to [batting] could be used together to evaluate ‘all round’ performance.
808 Journal of the Operational Research Society Vol. 56, No. 7

In their concluding comments, they suggest that subtracting who did not lose his wicket, scored more runs than Hayden
the bowling RM from the batting RM provides a measure of did for his dismissal. Hogg’s strike rate (92.20) is somewhat
all-round performance. The greater the difference, they higher than Hayden’s (75.82). On these measures, Hogg is
suggest, the better the player as an all-rounder. This ranked as having performed better than Hayden. When the
difference is interpreted as the average magnitude of victory, stage of the innings is taken into account, via the D/L
in runs, for the player’s team had he been the only batsman methodology, Hayden is ranked ahead of Hogg. Hayden’s
and the only bowler. net contribution is þ 21.39 runs compared with Hogg’s
It is questionable whether the scale so produced is a valid þ 16.90. Using the batting resource average Hayden has
interval scale and the measure depends heavily on whether contributed 3.34 runs per unit of resource consumed but
or not the player is deemed, or purports to be, an all- Hogg contributed 3.03. (The corresponding RM measures
rounder. A batsman who does no bowling would almost would be 334 and 303, respectively.)
certainly have a measure exceeding an all-rounder as there It will be noted that Bevan made a modest contribution of
would be nothing to subtract. This is seen as a serious only 10 runs. But his batting resource average is almost as
problem with this suggested measure of a player’s all-round high as Hayden’s. Bevan was not dismissed, having retired
ability in batting and bowling. hurt in this match, thereby not consuming resources for the
loss of his wicket. However, his net batting contribution of
þ 2.78 places his small contribution into the right perspec-
Illustrations of the performance measures tive and places him lower down the rankings of player
batting performance. This example suggests that the net
Illustrations of how the proposed measures of performance
batting contribution is less susceptible to distortion through
operate will take two forms. One is an illustration of how
small volumes of input than is the case for the batting
they operate in a single match; the VB Series final of 25th
resource average and corresponding RM measure.
January 2003. This will also highlight issues of the judgement
Table 2 illustrates some of the issues concerning bowling
of players’ performances in awards made by officials for
performance measures. If a batsman is run out then he is
performances in this match.
debited with the loss of the wicket and the corresponding
The second set of illustrations will be an overview of these
resource consumption. But the bowler is not credited with
performance measures over the whole of the VB Series of
the wicket nor its resource contribution. Table 2 includes
matches for 2002/03. (http://uk.cricinfo.com/link_to_data-
these items for the purposes of accounting for the number of
base/ARCHIVE/2002-03/OD_TOURNEYS/VBS/SCORE
wickets lost and associated resources.
CARDS accessed 20 July 2004.) in order to assess the
It will be noted that, in Table 1, Australia’s total of 229
viability of the measures in the more medium term.
was six runs below the average of Z(50,0) ¼ 235. The total
net contributions of the batsmen correspond to this slight
underachievement to the average, by six runs.
VB series final, 25 January 2003, Sydney
Similarly, in Table 2, England’s total of 224 was 11 runs
Tables 1 and 2 summarize several measures of performance below the 235 ODI average, consequently the combined net
of the Australian team from this match. The standard contributions of Australia’s bowlers was a positive 11.
measures are provided on the left with the D/L measures The various bowling performance measures show con-
proposed in this paper on the right. sistency in this match and there is unlikely to be any dispute
Since only one innings is considered the usual batting in their message; Lee’s performance is assessed as being the
average (total runs divided by total number of dismissals) best by all the measures. Below this, however, it will be noted
cannot be captured; only the total runs. In Table 1, Hogg, that Warne had a better average and strike rate than Hogg.

Table 2 England’s innings—bowling measures


Balls Strike Net bowling Bowling Bowling resource
Bowler Runs Wkts Average bowled rate contribution resources average

Lee 30 5 6.00 57 11.4 þ 35.28 28.6 1.04


Warne 58 2 29.00 60 30 8.87 21.5 2.69
Hogg 41 1 41.00 60 60 3.35 16.5 2.48
Williams 46 1 46.00 60 60 þ 0.64 20.5 2.24
Bichel 42 0 — 60 — 16.16 11.3 3.71
run out 1 þ 3.50 1.5
Byes & leg byes 7

Totals 224 10 11.0 99.9


AJ Lewis—Player performance in one-day cricket 809

But according to the D/L measures, Hogg’s performance is contribution is not only the best of the column but also is
rated better than Warne’s, reflecting the notion that bowling better than any net batting contribution. This quantitative
economy, and not just wickets, is an important factor in evidence supports the decision of the adjudicators in this
bowling performance. match to award the ‘man of the match’ to Lee. Although the
Williams has an inferior bowling average to that of Hogg match scorecard shows that four of his five wickets were the
but, by the D/L measures, Williams’ performance has been ‘cheaper’ lower order batsmen, which slightly devalues his
rated as the superior one reflecting the stage of the innings bowling average, he was instrumental in turning around a
when Williams conceded his runs and took his wicket match that England looked very likely to win. No doubt the
compared with Hogg. adjudicator was influenced not only by his superior bowling
Bichel would not normally be ranked as he took no average, as in Table 2, but also, subjectively, by Lee’s
wickets but by his net contribution and resource average, his significant effect on the outcome of the match.
performance can be placed in context with his fellow It will also be noted, from Table 3, that in addition to his
bowlers, albeit as the lowest-ranking performance. bowling Lee made a small positive contribution when
batting that added to his overall positive contribution to
the match. The evidence from the analysis, summarized in
Combined analysis Table 3, supports the judgement that awarded Lee the man-
As indicated previously, the traditional measures of perfor- of-the-match accolade. Table 3 also illustrates, however, that
mance cannot be used to compare or combine the separate Caddick’s bowling performance (despite a slight negative
performances for batting and bowling. The net contribu- contribution for his batting) resulted in a strong total
tions, however, possess these qualities so that overall contribution in the match for England. In quantitative
performance in a particular match can be assessed objec- terms, the award to Lee of the man-of-the-match was not
tively. This is an important characteristic as nearly every quite so clear-cut as it might have seemed.
match has a ‘man-of-the-match’ award. In major tourna- The measure of total net contribution as an indicator of
ments usually a media personality makes the judgement for all-round performance would be a potentially valuable aid to
this award. The net contribution would be a valuable the adjudicator in determining the man-of-the-match award.
measure to assist in this decision. Table 3 provides the
summary of net batting, bowling and match contributions of
the players from Australia (Aus) and England (Eng) that Assessment of the D/L performance measures in the
took part. medium term
Looking first, in Table 3, at just the net batting and In order to observe how the measures of performance
bowling contributions it can be seen that Lee’s bowling proposed in this paper would apply over more than one
match, the complete VB series of 2002–03, which contained
Table 3 Net player contributions, Aus v Eng, 14 matches, will be used for the purposes of evaluation. This
25 January 2003
series was a ‘round robin’ competition between Australia,
Net batting Net bowling Match England and Sri Lanka (SriL) that resulted in a best-of-three
Name Country contribution contribution contribution set of final matches between England and Australia. After 14
Lee Aus þ 2.70 þ 35.28 þ 37.98 matches, several of the traditional performance measures
Caddick Eng 2.00 þ 31.20 þ 29.20 can be compared more meaningfully with the measures
Hayden Aus þ 21.39 0 þ 21.39 proposed in this paper and by Beaudoin and Swartz. In
Stewart Eng þ 19.91 0 þ 19.91 addition, the combined contribution measure can be used to
Vaughan Eng þ 24.35 8.49 þ 15.86 evaluate the appropriateness of the ‘player-of-the-tourna-
Hogg Aus þ 16.90 3.35 þ 13.55
Collingwood Eng þ 4.27 0 þ 4.27 ment’ award. Squads for each team contained up to 18
Irani Eng 12.54 þ 16.33 þ 3.79 players so that some 54 players took part in at least one
Bevan Aus þ 2.78 0 þ 2.78 match. For brevity, only the top 20–22 players for each
Gilchrist Aus þ 0.99 0 þ 0.99 measure will be presented in order to discuss the relative
Williams Aus 0 þ 0.64 þ 0.64
merits of the measures.
Hussain Eng 1.06 0 1.06
Blackwell Eng 3.50 0.98 4.48
Flintoff Eng +1.87 8.23 6.36
Martyn Aus 10.45 0 10.45 Batting performance measures
Bichel Aus +0.81 16.16 15.35
Warne Aus 7.38 8.87 16.25 Table 4 provides the traditional batting averages for the
Ponting Aus 16.78 0 16.78 series. It will be noted that, based on total runs to times out,
Symonds Aus 16.95 0 16.95 Clarke has been placed at the head of the table, as his
Trescothick Eng 17.00 0 17.00 average was ‘infinite’. He was not dismissed in his one
Knight Eng 20.41 0 20.41
innings of 36 runs. Traditionally he, and several other
810 Journal of the Operational Research Society Vol. 56, No. 7

Table 4 VB series 2002–03 batting averages Table 5 VB series 2002–03 net batting contributions
Player Country Runs Innings Times out Average Rank Net Batting
Net contribution average
Clarke Aus 36 1 0 — — Player Country contribution rank rank
Nissanka SriL 12 1 0 — —
Lehmann Aus 202 4 2 101.00 1 Hayden Aus 152.76 1 3
Hogg Aus 89 3 1 89.00 2 Collingwood Eng 121.05 2 8
Hayden Aus 458 9 7 65.42 3 Jayasuriya SriL 116.65 3 6
Bevan Aus 221 7 4 55.25 4 Gilchrist Aus 105.25 4 11
Samaraweera SriL 55 2 1 55.00 5 White Eng 62.29 5 14
Jayasuriya SriL 359 7 7 51.28 6 Lehmann Aus 46.28 6 1
Knight Eng 461 10 9 51.22 7 Ponting Aus 43.14 7 12
Collingwood Eng 289 8 6 48.16 8 Trescothick Eng 37.58 8 20
Martyn Aus 276 8 6 46.00 9 Knight Eng 36.18 9 7
Gunawardene SriL 45 1 1 45.00 10 Bichel Aus 23.22 10 15
Gilchrist Aus 310 8 7 44.28 11 Sangakkara SriL 20.49 11 17
Ponting Aus 306 8 7 43.71 12 Jayawardene SriL 20.19 12 16
Watson Aus 86 5 2 43.00 13 Martyn Aus 18.04 13 9
White Eng 121 4 3 40.33 14 Watson Aus 17.22 14 13
Bichel Aus 39 2 1 39.00 15 Atapattu SriL 16.35 15 18
Jayawardene SriL 261 8 7 37.28 16 Vaughan Eng 14.24 16 19
Sangakkara SriL 234 8 7 33.42 17 Nissanka SriL 8.75 17 —*
Atapattu SriL 265 8 8 33.12 18 Clarke Aus 7.97 18 —*
Vaughan Eng 130 4 4 32.50 19 Samaraweera SriL 7.72 19 5
Trescothick Eng 316 10 10 31.60 20 Hogg Aus 7.46 20 2
Maher Aus 94 4 3 31.33 21
*These players were not dismissed and so do not qualify for a standard
Hussain Eng 278 10 10 27.80 22
batting average.

players in a similar situation such as Nissanka, would be next three positions. Similarly other more recognized bats-
excluded from the head of the table and placed below all men, such as Jayasuriya and Ponting have their perfor-
players who have a finite average. Since these players had mances more highly rated than through the simple batting
not been dismissed, they did not qualify for inclusion in the average.
rankings. This is clearly a deficiency of the standard batting This measure is accounting more fairly for all of a
average. And even when a player has qualified through at batsman’s innings and putting much less weight on whether
least one dismissal, the batting average exhibits weakness as or not the batsman loses his wicket, which is the outcome
a valid measure of performance. In Table 4, Lehmann and from only one ball. It is also making allowance for the stage
Hogg head the list for the standard batting average largely of the innings that batsmen score their runs. Since the D/L
because in two of their innings they were not out; their ‘true’ model (1) expects more runs per ball towards the end of the
average performance has clearly been distorted. Several of overs available when there are wickets in hand then net
the other players’ rankings are also distorted by the low contributions from the final overs, colloquially called the
frequency of dismissals for the number of times they batted. ‘slog’ overs, are lower. For example Bevan, who was fourth
Bichel, for example, averages 39 from two innings but only in the batting averages of this tournament and, in January
one dismissal. 2004, was ranked fifth in the PriceWaterhouseCoopers
ratings (http://uk.cricinfo.com/db/STATS/RATINGS/PW
CODIINDIV.html accessed 20 July 2004), scores many of
Net batting contribution
his runs towards the end of the innings and is frequently not
Table 5 provides rankings based on net batting contribution. out. Under the net contribution measure, the value of a
This measure enables the performances of Clarke, Nissanka wicket, in terms of loss of run-scoring potential, becomes less
and others, legitimately and objectively, to be included for and less important as the team’s innings progresses, and
ranking purposes. With this measure, Lehman falls down the declines to zero right at the end. These aspects of the net
rankings as the effect of being ‘not out’ is reduced. Similarly contribution measure are reflected in Bevan slipping from
for Hogg; who descends to 20th place after the reduction of fourth place to be outside the top twenty batsmen of the
the effect of only having been dismissed once in his three tournament.
innings.
Conversely, several recognised batsmen who were quite
Batting resource average
low in the rankings in Table 4 have been promoted using the
net batting contribution measure. Hayden becomes top and Table 6 summarizes the top 20 players of the VB 2002-03
Collingwood, Jayasuriya, and Gilchrist are promoted to the series using the batting resource average as defined
AJ Lewis—Player performance in one-day cricket 811

Table 6 VB series 2002–03 batting resource averages


Batting Bat. res Res Batting average Net contribution
Player Country Runs resource avge avge rank rank rank

Nissanka SriL 12 1.5 8.10 1 — 17


Bichel Aus 39 7.0 5.56 2 15 10
White Eng 121 26.8 4.52 3 14 5
Collingwood Eng 289 77.2 3.75 4 8 2
Gilchrist Aus 310 91.6 3.39 5 11 4
Hayden Aus 458 137.4 3.33 6 3 1
Jayasuriya SriL 359 115.5 3.11 7 6 3
Clarke Aus 39 13.7 2.85 8 — 18
Lehmann Aus 202 71.0 2.85 9 1 6
Watson Aus 86 31.2 2.75 10 13 14
Samaraweera SriL 55 21.3 2.58 11 5 19
Ponting Aus 306 118.9 2.57 12 12 7
Trescothick Eng 316 127.2 2.49 13 20 8
Vaughan Eng 130 52.7 2.47 14 19 16
Flintoff Eng 16 6.5 2.46 15 30 23
Hogg Aus 89 36.4 2.45 16 2 20
Gunawardene SriL 45 18.7 2.41 17 10 21
Knight Eng 461 192.8 2.39 18 7 9
Sangakkara SriL 234 99.3 2.36 19 17 11
Martyn Aus 276 118.5 2.33 20 9 13

previously. The corresponding RM measures as defined by Bowling performance measures


Beaudoin and Swartz5,6 would involve the multiplication of
The bowlers’ performances will be ranked under the two
the resource averages values in Table 6 by 100. For
measures suggested by the D/L methodology and compared
comparative purposes, the same players’ rankings by the
with the standard bowling averages.
other batting measures are included.
Table 6 clearly shows a strength and a weakness of the
batting resource average (and batting RM measure). A
strength, as with the net contribution measure, is that it
Standard bowling averages
enables players who were not dismissed to be ranked, such as Table 7 provides the standard bowling averages of the VB
Nissanka and Clarke. A further strength is that it also places Series of 2002-03. Bracken heads the list with an average of
less weight on not being dismissed as reflected in the slides 13.50 runs conceded per wicket taken. Behind him are
down the rankings of Lehman and Hogg and the promo- several of the more regular bowlers in Muralitheran, Lee and
tions of the more recognized batsmen such as Collingwood Caddick. Bracken’s low average comes partly from his
and Gilchrist. playing in only two games. In one of these, he took three
A weakness in these measures can be identified through wickets for 21 runs in a low-scoring innings. His top-place
Nissanka’s topping of the rankings and the surprise ranking has therefore come about from one very good
elevation of Bichel to second place. One aspect of this performance in his small total input to the tournament. This
ranking is that these two players’ resource averages are illustrates, again, how measures using averages are prone to
based on few innings and so, in a similar way to the standard distortions when based on small volumes of data. The
batting average, cannot be relied upon to be representative. regular bowlers Lee and Caddick played in nine matches
Further, and importantly, these two players batted late in each. Consequently, the effects of outstanding, or poor,
their respective innings and in matches where there was no single performances are smoothed over the larger number of
hope of winning the matches. Runs earned have thus come games. This means that bowling averages provide a mean-
from the consumption of very little resource. When most of ingful measure of performance only for bowlers whose input
the wickets have been lost in an innings and plenty of overs has been substantial. Bowlers with low volume input cannot
are in hand the D/L curves (1) and (4) have virtually zero fairly be compared and ranked through the standard
gradient so that little is contributed to the denominator of bowling average.
the resource percentage average for the balls received. Table 8 summarizes all the performance measures of the
This criticism is also relevant to the measure of net top 20 bowlers, as ranked by the net bowling contribution
contribution but its effect is less pronounced as the measure measure proposed in this paper. The final three columns
is additive and not a ratio. show the rankings of these players under the three
812 Journal of the Operational Research Society Vol. 56, No. 7

mechanisms. It will be seen that there is some agreement but oppositions’ innings and so consisted of many ‘tail-end’
also significant variations that deserve comment. batsmen not possessing the batting expertise of the higher-
order batsmen. Further, his performance in the final

The Australian bowler Lee is not ranked in the top three match of the series, as outlined previously, gave him a
by any of the three measures—yet he received the ‘Player- high profile towards the end of the tournament that may
of-the-Series’ award. Analysis of the score sheets, on the have coloured the objective judgement of the award’s
pages of the Internet previously cited, shows that many of adjudicators.
his 18 wickets were taken towards the end of the
Bracken’s first ranking by the standard bowling average is
mirrored by being placed in the same position, and by a
clear margin, using the bowling resource average (and the
Table 7 VB 2002–03 series, standard aggregate bowling corresponding RM measure). As both of these measures
averages are based on averages, the small volume of input with one
Player Country Runs/wkts Avge Rank outstanding performance in only two games is arguably
having a distorting effect on the ratios, whereas the
Bracken Aus 54/4 13.50 1 averages of most other bowler are based on several more
Muralitheran SriL 107/7 15.28 2
Lehmann Aus 83/5 16.60 3 matches. The net bowling contribution measure, with its
Lee Aus 338/18 18.77 4 additive structure places less weight on that one out-
Clarke Aus 24/1 24.00 5 standing performance and ranks him second, by a clear
Caddick Eng 324/13 24.92 6 margin, behind Caddick.
Williams Aus 178/7 25.42 7
The England bowler Caddick was ranked sixth under the
Watson Aus 262/10 26.20 8
FernandoCRD SriL 355/13 27.30 9 standard bowling average but ranked first using net
Blackwell Eng 303/11 27.54 10 bowling contribution. As the opening bowler, his wickets
Warne Aus 167/6 27.83 11 include many of the higher order batsmen who have
Hogg Aus 341/12 28.41 12 greater run scoring potential and whose dismissal attracts
Collingwood Eng 57/2 28.50 13 greater merit than the dismissal of the less skillful lower-
Anderson Eng 373/13 28.69 14
Lakshitha SriL 65/2 32.50 15 order batsmen. This contributes to his top ranking.
Bichel Aus 230/7 32.85 16
Under the standard bowling average Mubarak and
Vaughan Eng 67/2 33.50 17 Jayawardene do not qualify for a finite average, as they
Nissanka SriL 174/5 34.80 18 did not take any wickets, yet their performances can be
Jayasuriya SriL 300/8 37.50 19 evaluated, and ranked, by using the two D/L-based
de Silva SriL 197/5 39.40 20
measures.

Table 8 VB series 2002–03 summary of bowling performance measures and rankings


Std bowl Net bowl Bowl Bowl res Std avge Res avge Net cont
Player Country Run/wkts avge cont. res avge rank rank rank

Caddick Eng 324/13 24.92 54.27 165.6 1.95 6 4 1


Bracken Aus 54/4 13.50 48.29 44.1 1.22 1 1 2
Muralitheran SriL 107/7 15.28 47.56 69.1 1.54 2 2 3
Lee Aus 338/18 18.77 41.94 166.4 2.03 4 6 4
Blackwell Eng 303/11 27.54 37.25 149 2.03 10 7 5
Williams Aus 178/7 25.42 29.77 90.9 1.95 7 5 6
Lehmann Aus 83/5 16.60 29.34 48.8 1.70 3 3 7
Bichel Aus 230/7 32.85 22.68 110.8 2.07 16 8 8
Watson Aus 262/10 26.20 6.65 117.7 2.22 8 9 9
Warne Aus 167/6 27.83 3.78 74.5 2.24 11 10 10
Jayawardene SriL 8/0 —* 0.40 3.2 2.47 34 15 11
Vaughan Eng 67/2 33.50 0.89 28.7 2.33 17 11 12
Clarke Aus 24/1 24.00 2.17 9.4 2.56 5 18 13
Mubarak SriL 29/0 —* 2.97 11.1 2.61 34 19 14
Martyn Aus 96/1 96.00 6.88 39.0 2.46 30 12 15
Flintoff Eng 56/1 56.00 8.23 20.6 2.72 26 21 16
Symonds Aus 82/1 82.00 8.32 32.6 2.51 28 17 17
Collingwood Eng 57/2 28.50 10.11 20.7 2.75 13 22 18
White Eng 162/3 54.00 11.26 65.8 2.46 25 13 19
Lakshitha SriL 65/2 32.50 13.11 23.0 2.83 15 24 20
*See footnote of Table 5.
AJ Lewis—Player performance in one-day cricket 813

Combined performance measure The resource average measures and the RM measures of
Beaudoin and Swartz5,6 do not have the ability to compare
As outlined in relation to the VB Series final, net
batsmen and bowlers on the same scale of measurement.
contribution can compare performances between batting
Beaudoin and Swartz indicate that the difference of the RM
and bowling and evaluate performances in the combination
measures of an individual who has batted and bowled in the
of these two facets of the game. Table 9 provides the batting
match would provide an indicator of the best ‘all round’
and bowling net contributions, and their totals, for the top
performer in the match. This difference is interpreted as the
20 players over the whole of the 14 matches of the VB Series
runs the player scored or conceded if they undertook all of
2002-03.
Table 9 provides information to assist in making the batting and bowling, respectively, for the team; the
judgements concerning players’ performances over the whole higher the figure the better the all-round performance.
of the VB Series. It must be stressed that just the batting and Table 10 summarizes the RM measures of appropriate
bowling performances are included in this analysis. Other players that utilize information in Tables 6 and 8.
aspects of players’ games such as wicket keeping, fielding Several comparisons of interest can be made between
and catching are not accounted for and are more difficult to Tables 9 and 10. First is the exclusion from Table 10 of any
quantify. As a consequence, in assessing overall player player, such as Hayden, Gilchrist and Bracken who have not
performances an adjudicator needs to make subjective contributed in both batting and bowling; the latter through
allowance for these further aspects of the game, such as not having batted in any of his matches for his country. If
the number of catches taken, runs saved or run-outs effected. Table 10 were used to assist adjudicators of Player-of-the-
Nevertheless, it would seem that the Australian bowler Lee’s Series awards then there is an automatic bias against
nomination as the ‘Player of the Series’ ignored the strong batsmen who do not bowl and towards bowlers, most of
contributions made by many other players in the series. whom will have batted in the series of necessity.
From Tables 5 and 8, straight comparisons of the batting Table 10 shows Nissanka and Bichel heading the list. Both
and bowling contributions places Lee well behind others. of these players took only a small part in the tournament so
Looking at Table 9, which summarizes candidate players’ that their RM measures are distorted for this and for other
batting and bowling contributions and their total, Lee is reasons outlined previously. Consequently, their top rank-
ranked 14th. His total net contribution of þ 33.99 is well ings need to be viewed with considerable caution. Through
below those of countrymen Hayden and Gilchrist, below the total net contribution of Table 9, however, these two
Englishman Collingwood and below Sri Lankan Jayasuriya. players have their more limited contributions placed into
The Australian player Lehman was the highest ranked context in the whole VB series of 14 matches; Nissanka is
player who made positive net contributions in both batting not ranked in the top 20 and Bichel is placed 10th overall.
and bowling. It could be argued, therefore, that he could be A further issue in relation to the RM difference is whether
regarded as the best all round player in the tournament. it is a valid interval scale for the measure of all-round

Table 9 Net contributions for batting, bowling and their total


Player Country Batting contribution Bowling contribution Total net contribution Rank

Hayden Aus 152.76 0.00 152.76 1


Collingwood Eng 121.05 10.11 110.94 2
Gilchrist Aus 105.25 0.00 105.25 3
Jayasuriya SriL 116.65 19.47 97.18 4
Lehmann Aus 46.28 29.34 75.62 5
Caddick Eng 1.73 54.27 56.00 6
White Eng 62.29 11.26 51.03 7
Bracken Aus 0.00 48.29 48.29 8
Muralitheran SriL 0.31 47.56 47.87 9
Bichel Aus 23.22 22.68 45.90 10
Ponting Aus 43.14 0.00 43.14 11
Trescothick Eng 37.58 0.00 37.58 12
Knight Eng 36.18 0.00 36.18 13
Lee Aus 7.95 41.94 33.99 14
Williams Aus 0.00 29.77 29.77 15
Watson Aus 17.22 6.65 23.87 16
Sangakkara SriL 20.49 0.00 20.49 17
Jayawardene SriL 20.19 0.40 19.79 18
Atapattu SriL 16.35 0.00 16.35 19
Vaughan Eng 14.24 0.89 13.35 20
814 Journal of the Operational Research Society Vol. 56, No. 7

Table 10 Beaudoin and Swartz6 all-rounder rankings for VB 2002–03 series


Batting Batting Bowling Bowling RM RM all-rounder
Player Country RM RM rank RM RM rank difference rank

Nissanka SriL 809.71 1 292.96 25 516.75 1


Bichel Aus 555.63 2 207.60 8 348.03 2
White Eng 451.62 3 246.20 13 205.42 3
Lehmann Aus 284.70 9 170.02 3 114.68 4
Collingwood Eng 374.53 4 275.44 22 99.09 5
Muralitheran SriL 229.18 22 154.94 2 74.24 6
Watson Aus 275.27 10 222.55 9 52.72 7
Jayasuriya SriL 310.73 7 279.14 23 31.59 8
Caddick Eng 226.31 23 195.70 4 30.61 9
Clarke Aus 285.42 8 256.16 18 29.26 10
Vaughan Eng 246.65 14 233.18 11 13.47 11
Hogg Aus 244.80 16 246.40 14 1.60 12
Martyn Aus 232.83 20 246.07 12 13.24 13
Jayawardene SriL 230.98 21 247.52 15 16.54 14
Lee Aus 183.62 33 203.16 6 19.54 15
Flintoff Eng 245.85 15 272.02 21 26.17 16
Blackwell Eng 155.39 37 203.34 7 47.95 17
Warne Aus 176.00 36 224.12 10 48.12 18
Harmison Eng 210.18 29 293.22 27 83.04 19
de Silva SriL 176.93 35 297.99 28 121.06 20

performance. Consider Caddick and Clarke in Table 10. insufficient in that they do not take into account the context
Their RM differences are approximately equal at around 30. in which runs are made or conceded and wickets are taken or
Caddick’s figure is the difference of a slightly below average given away. Compared with the longer version of the game,
batting RM of 226.31 and a better than average bowling there are more occasions in which the performance of the
RM of 195.70. Both these are relative to the 235 average individual needs to take second place to the strategy for the
runs scored in ODIs. Clark has a better batting performance team in its objective in trying to win the match. Further,
than Caddick, but a much worse bowling performance yet these batting and bowling averages are incompatible for the
they end up with approximately the same RM difference and purposes of comparing performances in the two disciplines
hence adjacent rankings. and combining them in order to assess all-round perfor-
In Table 9, Caddick is ranked very highly with a total net mance.
contribution of þ 56.00. Clarke, on the other hand, features The methodology of measures proposed in this paper
outside the top 20 in 22nd place with a net contribution of is based on the Duckworth/Lewis model and the D/L
þ 5.80. It would be for Beaudoin and Swartz to argue resource percentages that are used for resetting targets in
whether a 30-point difference in RM measures has the same one-day cricket. It requires data on the outcome of every ball
meaning for whatever two RM batting and bowling of a match. Although this is available in verbal form on
measures gave that difference. This paper argues that the Internet (http://www.cricket.org/link_to_database/
combined net contribution retains the additive ratio-scale ARCHIVE/ accessed 29 January 2004), the data’s avail-
nature of the net batting and bowling contributions and ability commercially in electronic form has enabled sub-
hence produces a more meaningful and reliable measure of stantial analysis to be undertaken. Using several matches
all-round performance. It also avoids the arbitrary decision over a series, players’ performances can be more reliably
of whether or not the player regards himself as an all- evaluated taking into account the context in which runs and
rounder in batting and bowling. wickets are earned or conceded. The measures of net batting
and bowling contribution have the considerable virtue of
being in the same units and hence they provide a common
and additive scale of measurement for the comparison and
Summary
aggregation of batting and bowling performances.
The standard measures of batting and bowling average are A further measure of performance evaluation using the D/
simple to calculate and have been used to assess player L methodology is the resource average. This is the average of
performances over many years. In the one-day form of runs earned or conceded per unit of resource consumed by
cricket, they provide some indication of a player’s abilities in batsmen or contributed by bowlers. This measure, equiva-
batting and bowling, for a particular match, for a series, and lent to the runs per match measure of Beaudoin and
even throughout their careers. These measures, however, are Swartz,5,6 may well have considerable virtue in the longer
AJ Lewis—Player performance in one-day cricket 815

term where the ratios are not based on small divisors. But, as regarded as essential so that the summary statistics and
has been shown through the examples, they are not as player rankings can be produced in a timely manner
reliable measures of comparing performances in batting and following the conclusion of the many one-day matches that
bowling in individual matches or in a short series as the net are now played regularly around the world.
contribution measures. Further, the resource averages (and
the corresponding RM measures) do not enable objective
Acknowledgements—The author would like to acknowledge, with
comparison to be made of batting and bowling perfor- thanks, the assistance received in preparing this paper. In particular
mances, and the use of the difference of a player’s batting thanks are due to his colleague Frank Duckworth for valuable
and bowling RMs as a measure of ‘all-round’ performance is comments on the draft manuscript. Further thanks are expressed also
for the comments of the anonymous referees that have led to
questionable. improvements in the presentation of the paper. Finally, additional
thanks are due to Champion Data (Australia) and the Press
Association (Sports) for the provision, in electronic form, of the
Further developments ball-by-ball details of matches that have been absolutely essential for
undertaking the analysis summarized in this paper.
Net contribution shows potential for the more reliable
assessment of player performances in batting and bowling,
and in comparing performances between these two dis- References
ciplines. In addition, it provides a valuable measure in 1 Engels M (ed) (2004). Wisden Cricketers’ Almanack. John Wisden
assessing players’ all-round ability in batting and bowling. & Co. Ltd.: Hampshire.
This potential has been shown not just for an individual 2 Duckworth FC and Lewis AJ (2004). A successful operational
match but also for a series of matches. Further evaluation of research intervention in one-day cricket. J Opl Res Soc 55:
749–759.
this potential is necessary on how the measure would
3 Duckworth FC and Lewis AJ (1998). A fair method for resetting
perform in the longer term if the methodology were to be the target in interrupted one-day cricket matches. J Opl Res Soc
incorporated into or to supplant the existing standard and 29: 220–227.
commercial measures of performance and rankings. For this 4 Lewis T (2003). Towards fairer cricketing averages. In:
to be achieved, several more series of matches need to be Bulletin of the International Statistical Institute, Contributed
Papers, Volume LX, Book 1. International Statistical Unit:
analysed and across more countries in order for there to be a
Voorburg, The Netherlands, pp 722–723.
greater pool of data to assess how the measures will reflect 5 Swartz T and Beaudoin D (2003). The best batsmen and bowlers
and rank these attributes of cricketers. in one-day cricket. In: Bulletin of the International Statistical
The resource averages have been shown to have more Institute, Contributed Papers, Volume LX, Book 2. International
limited reliability over a few matches. But, over more Statistical Unit: Voorburg, The Netherlands, pp 499–500.
6 Beaudoin D and Swartz TB (2003). The best batsmen and
matches, reliability of resource averages (and their equiva-
bowlers in one-day cricket. South African Statist J 37:
lent in RMs) in assessing just batting and bowling 203–222.
performance is likely to increase. Again analysis of more 7 Johnston MI, Clarke SR and Noble D (1993). Assessing player
matches will confirm this expectation or otherwise. performance in one-day cricket using dynamic programming.
The availability of ball-by-ball data is an essential element Asia Pac J Op Res 10: 45–55.
8 Clarke SR (1988). Dynamic programming in one-day cricket:
in undertaking the extensive analysis of this methodology.
optimal scoring rates. J Opl Res Soc 39: 331–337.
For the longer term, this would need to be organized
through suitable sports data-collection organizations. A
further necessity is the development of software that will Received January 2004;
process this ball-by-ball data accurately and quickly. This is accepted July 2004 after one revision

You might also like