November 2, 2012

Influence of wides and no balls on Test bowler averages

A stats analysis to determine the effect of wides and no-balls on the bowling averages in Tests

Waqar Younis has the lowest number of wides and no balls per Test amongst the top pace bowlers Jack Atley / © Getty Images

The previous article was a huge data-rich one involving bowler-pairs. Even understanding the significance of data presented there would have taken some time. This is also a bowler-centric article but much simpler to read and understand, if not in analysis. I must thank Arijit Dasgupta for this suggestion.

A few days after India won the World Cup during 1983, it was decided that the wides and no balls would be debited to the bowler analysis. It was a simple idea but had significant problems in execution and it took more than 2 years for the first scorecards to be rolled out with the wides/no balls information shown against the bowler's analysis. It is possible that the bowlers and wides were debited to the bowlers from Test # 961 onwards (India-Pakistan match played at Bangalore during September 1983). However the debited numbers were not shown against the bowler analysis until Test # 1029 (Australia-New Zealand Test match at Brisbane starting on 8 November 1985).

So this is our real cut-off Test. So one can say that the cut-off is data-driven than law-implementation-driven especially as the 1983-87 period is a grey area. Milind's query to MCC on this has resulted in a vague unclear answer. The only assumption, and a valid one, is that where wides/no balls have been shown against a bowler analysis, these are included in the bowling analysis runs.

Arijit's idea was that this single law change had a profound influence on the Bowling averages of bowlers over the past 25 plus years and that I should study this in depth. This seemed an excellent idea and I have been working on it for the past month. It turned out to be a complex exercise, because of the reasons outlined below.

This analysis concentrates almost totally on the post-1985 period, or more correctly the post-T1029 period. For the pre-T1029 period, I will present a single summary of the wides/no balls which would have been debited to the bowlers. This will give an idea of the possible increase in bowling averages.

For the first 1028 Tests the total number of wides/no balls stands at 15353 and the number of bowler wickets at 31727. This leads to a notional increase of 0.48 runs per wicket. For the next 1027 Tests the total number of wides/no balls stands at 29118 and the number of bowler wickets at 31449. This has led to an increase of 0.93 runs per wicket. This clearly shows a doubling of the number of wides/no balls post-T1029. This may be due to the stricter interpretation of extras-related rules, stricter handling of short balls, significant law changes and the increase in declaring a ball wide owing to the influence of ODIs.

Before anything, let me be clear that the overriding consideration in all these adjustments is simply that Team wides/no balls are king. These would remain sacrosanct. All adjustments will be done only with the bowler level wides/no balls. With this guiding principle I have ensured that the Team scores remain untouched. The salient points of this analysis are outlined below.

1. All scorecards, including the Cricinfo ones, do not include full information about the number of deliveries that were declared wide/no ball and how many runs were scored off these deliveries. When 5 wides are conceded, 5 runs are added properly to the team score, extras total, no balls tally in the team extras line and bowler runs conceded. However, somewhat inconsistently, this is shown as only 1 wide delivery against the bowler. Same is true for the rare case where 5 runs are scored off a no ball. As an example, please peruse the very last innings scorecard. Test # 2055, India-New Zealand Test at Bangalore. In the Indian second innings, Southee bowled a 5-ball wide in the team's second over. The extras shows 5 wides, team total and bowler runs are increased by 5 runs but only 1 wide is shown against Southee in the bowling analysis.

2. Even after Test # 1029, there are matches in which no wides and no balls are shown against the bowler analysis even though there are team wides and no balls. Nothing can be done about these and such Test matches are kept out of these calculations.

3. There are matches in which the team wides/no balls are less than the sum of the bowler wides/no balls. The adjustment here is simple. Reduce the bowler wides/no balls proportionately to match the team wides/no balls. Very logical and correct. If there is something wrongly done, I would rather assume that all mistakes are restricted to the bowler figures rather than the team numbers.

4. Now comes the difficult part. The team wides/no balls are higher than the sum of the bowler wides/no balls. This is a highly nuanced situation. There are many types of differences so I had to use a different method for each adjustment.

  • Situations such as team line showing 5/6/7/8... wides/no balls and only one bowler 's analysis showing 1/2/3/4... wide/no ball. No problems since these are similar to the Southee situation. Just make the specific bowler's wides/no balls as 5 or 6 or whatever. Could as well be team showing 10 wides and bowler showing 2 wides. The key is an exact difference of multiple of 4 wides/no balls.
  • Situations where the difference is x wides but more than one bowler has bowled the wides. I have taken a logical, if not the most accurate approach that the difference will be attributed to the bowler who has bowled most wides. For instance, if the team has 5 wides and two bowlers have bowled 1 and 2 wides, the additional 2 wides are allotted to the bowler who has captured 2 wides. May not be perfect but there is a good chance that it is correct.
  • Sometimes there is no pattern. Team has been debited 8 wides and the bowler total is 3 wides. Nothing can be done about it and it is left as it is.
  • For recent matches there was the option of perusing the commentary. However it is a thankless option and there is no way I could have done it because of the difficulty and the time required. 130 overs are shown in three screens and perusal of each delivery was needed. Browser search of wide is useless because of the "wide of point" type of commentary.
  • In a few cases I have perused the scorecards and taken an intuitive and informed decision. By now I had got a clear idea of the bowler characteristics.
5. At the end of about a week's tough work, I had improved the bowler level byes/no balls availability from about 21000 to just over 22800. This works to about 78% of the total tally of wides/no balls for this period of 29000+ and represents a major successful move forward. The overall impact might only be a reasonably small figure but the impact on individual bowlers sometimes goes as high as 7%. The analysis from this point onwards is a simple one and my tables indicate the current average, adjusted average and the % difference. Multiple tables are shown so that we can get different insights.

In view of my own physical condition, I am not going to offer detailed comments on the tables. I will present all the tables and at the end, summarise what I perceive are the key points. Readers can always contribute their bit and make this a discussion forum. Let me also say that I will publish many more comments without any response from me. I have to necessarily cut the amount of work I do. More often than not, I have to let my silence speak.

Since it is not possible to identify the wides/no balls conceded by the pre-T1029 bowlers, it is not possible to increase their averages. So the only fair method of comparison will be a notional reduction of the post-T1029 bowlers' averages.

75+ wickets bowlers ordered by reduction % of bowling average

BowlerTypeCtryStartEndMatsWktsAvgeWDsNBsTotalAdjAvge% chg
             
Malinga L.SRFMSlk200420103010133.163522926430.547.88%
Reiffel P.RRFMAus199219983510426.96519820325.017.24%
Collins P.TLFMWin199920063210634.621223524732.296.73%
Edwards F.HRFWin200320125415838.386933840735.806.71%
Lee BRFAus199920087631030.825358764028.756.70%
Pollock S.MRFMSaf1995200810842123.122860263021.626.47%
FernandoRFMSlk200020124010037.843421024435.406.45%
KasprowiczRFMAus199620063811332.88321922230.915.98%
Morkel MRFSaf200620124215030.375221526728.595.86%
Roach K.A.JRFWin20092012218227.702011313326.075.86%
Flintoff ARFMEng199820097922632.793536640131.015.41%
ShoaibAkhtarRFPak199720074617825.691722724424.325.34%
Mohd SamiRFPak20012012368552.743020923949.935.33%
AmbroseRFWin198820009840520.99642242819.935.04%
Abdur RazzaqRFMPak199920064610036.931017518535.085.01%
Srinath JRFMInd199120026723630.473130733829.044.70%

200+ wickets bowlers with over 2.0% reduction % of bowling average

BowlerTypeCtryStartEndMatsWktsAvgeWDsNBsTotalAdjAvge% chg
 
Lee BRFAus199920087631030.825358764028.756.70%
Pollock S.MRFMSaf1995200810842123.122860263021.626.47%
Flintoff ARFMEng199820097922632.793536640131.015.41%
AmbroseRFWin198820009840520.99642242819.935.04%
Srinath JRFMInd199120026723630.473130733829.044.70%
Vaas WPUJCLFMSlk1994200911135529.582042044028.344.19%
Wasim AkramLFMPak1985200210441423.62939140022.654.09%
Gough DRFEng199420035822928.401924426327.254.04%
Zaheer KhanLFMInd200020128529132.076730537230.793.99%
McDermottRFAus198419967129128.63430530927.573.71%
McGrath G.DRFMAus1993200712456321.642636439020.953.20%
Walsh C.ARFWin1984200113251924.441239240423.663.19%
Hoggard M.JRFMEng200020086724830.502521624129.533.19%
Caddick A.RRFMEng199320036223429.91621522128.973.16%
Cairns C.LRFMNzl198920046221829.402217519728.503.07%

Top bowlers in order of wickets

BowlerTypeCtryStartEndMatsWktsAvgeWDsNBsTotalAdjAvge% chg
             
MuralitharanrobSlk1992201013380022.73238839022.242.15%
Warne S.KrlbAus1992200714570825.422316318625.151.03%
Kumble ArlbInd1990200813261929.651027028029.201.53%
McGrath G.DRFMAus1993200712456321.642636439020.953.20%
Walsh C.ARFWin1984200113251924.441239240423.663.19%
Kapil Dev NRFMInd1978199413143429.653202329.590.18%
Hadlee R.JRFMNzl197319908643122.3002222.290.02%
Pollock S.MRFMSaf1995200810842123.122860263021.626.47%
Wasim AkramLFMPak1985200210441423.62939140022.654.09%
HarbhajanrobInd199820119840632.226152132.170.16%
AmbroseRFWin198820009840520.99642242819.935.04%
Ntini MRFSaf1998200910139028.834513017528.381.56%
Botham I.TRFMEng1977199210238328.4000028.400.00%
MarshallRFWin197819918137620.958536120.780.77%
Waqar YounisRFMPak198920038737323.567455223.420.59%
Imran KhanRFPak197119928836222.8100022.810.00%
Vettori D.LlspNzl1997201211236034.42412112534.081.01%
Lillee D.KRFAus197119847035523.9200023.920.00%
Vaas WPUJCLFMSlk1994200911135529.582042044028.344.19%
Donald A.ARFSaf199220027233022.25629115321.792.08%
WillisRFEng197119849032525.2000025.200.00%
Lee BRFAus199920087631030.825358764028.756.70%
Gibbs L.RrobWin195819767930929.0900029.090.00%
Trueman F.SRFEng195219656730721.5800021.580.00%

Bowlers who have conceded most wides/no balls per Test

BowlerTypeCtryStartEndMatsWktsAvgeWDsNBsTotalWdNb/Test
            
Malinga L.SRFMSlk200420103010133.16352292648.80
Lee BRFAus199920087631030.82535876408.42
Collins P.TLFMWin199920063210634.62122352477.72
Edwards F.HRFWin200320125415838.38693384077.54
Mohd SamiRFPak20012012368552.74302092396.64
Morkel MRFSaf200620124215030.37522152676.36
Roach K.A.JRFWin20092012218227.70201131336.33
FernandoRFMSlk200020124010037.84342102446.10
KasprowiczRFMAus199620063811332.8832192225.84
Pollock S.MRFMSaf1995200810842123.12286026305.83
Reiffel P.RRFMAus199219983510426.9651982035.80
ShoaibAkhtarRFPak199720074617825.69172272445.30
Flintoff ARFMEng199820097922632.79353664015.08
Srinath JRFMInd199120026723630.47313073385.04
.....
Swann G.SrobEng200820124619229.595050.11

General Comments

1. Probably the most significant bowling average change is one for Curtly Ambrose. The already very low average of this great bowler has been further reduced to below-20.0, 19.93 to be exact, making comparisons with Syd Barnes possible.

2. The next two significant reductions are for two very dissimilar bowlers, Brett Lee and Shaun Pollock. Their bowling averages have been reduced by over 6%. Pollock's average has been reduced to a wonderful figure of 21.62. They have bowled over 600 wides/no balls. What may be the reason for an accurate bowler like bowling so many no balls?

3. Similarly two totally dissimilar bowlers head the % reduction: Malinga and Reiffel. A real paradox. One a tear-away slinger and the other, a seemingly accurate medium-pacer.

4. Look at Swann's tally of wides/no balls. He has bowled a single wide in his 5 year career. That was in the 2011 Birmingham Test against India. That must have been a momentous occasion. And the icing on the cake: these were 5 wides.

5. If Swann is so accurate, how does one explain Muralitharan's 388 no balls? Warne, on the other hand, seems to have bowled only 163 no balls.

6. Not just Murali. Why do so many top bowlers bowl so many no balls. Surely some of these resulted in loss of a wicket.

7. Look at Waqar Younis' accuracy, only 52 wides/no balls and compare the same with Wasim Akram's 400 wides/no balls. The faster bowler has far fewer wides/no balls.In fact Waqar Younis has the lowest wides/no balls per Test amongst the top pace bowlers.

8. Saqlain Mushtaq and Muralitharan have the highest average of wides/no balls per Test, for spinners, either side of 3.0. If we conclude that unconventional off-spin is the reason, then Harbhajan seems to have excellent control with only 0.22 wides/no balls per Test.

9. Now for the maximum number of wides and no balls bowled by a bowler in an innings. These are either from the raw unadjusted scorecard data or where the changes are 100% correct (1 wide in analysis line for a single bowler and 5 wides in team extras line).

  • No Balls: Unusually high numbers are present. Wasim Akram, in Test # 1283 against South Africa (1995), conceded 21 No balls. Ambrose's analysis in Test # 1363 shows 21 No balls, but this is wrong since the bowler total is in excess of the Team totals. Hence only Wasim Akram's is the correct figure indisputably.
  • Wides: There are two candidates. In Test # 1809 Pak against Eng (2006), Umar Gul conceded either 10 or 14 wides but not very clear. In Test # 1836 against Eng (2007), Collymore conceded either 10 or 14 wides but not very clear. I could peruse the commentaries to sort this out but it would amount to too much work for very little gains.
  • Total WDs/NBs: There are two 22s and 23s each but in unclear situations. Hence the total high must be allotted to Wasim Akram, with 21 No balls and Vaas, in Test 1592 against Pak (2002) who conceded 1 wide and 20 no balls.

10. Cannot resist this comment. I read a report wherein Kohli has mentioned that England and India gave India flat tracks for practice matches and green tops for Tests. If this is true, then how did England score 474/8, 269/6, 221, 544, 710/7 and 591/6. Around 60 Runs per wicket. Green tops? And Australia scored 330, 240, 659/4, 369, 604/7 and 167/5. RpW is around 51.5. India has the right to select any team, no spinners or 11 batsmen or 11 bowlers or 11 Zonal players for the practice match. Why make the silliest of statements to support it?

One final summary. I have given below the top-10 of the current Bowling average tables, for bowlers who have captured more than 150 wickets.

Barnes, Davidson, Marshall, Garner, Ambrose, Laker, Trueman, McGrath, Donald, Hadlee.

Now the same table with the wides/no balls removed for post-T1029 bowlers.

Barnes, Ambrose, Davidson, Marshall, McGrath, Garner, Laker, Trueman, Pollock, Donald.

The most significant change is Shaun Pollock moving from no.15 to no.9. Ambrose moves from 5 to 2. McGrath moves from 8 to 5. Muralitharan remains at no.11 in both tables.

To download the complete bowler tables and match summaries, please CLICK HERE.

Finally my thanks once again to Arijit.

Anantha Narayanan has written for ESPNcricinfo and CastrolCricket and worked with a number of companies on their cricket performance ratings-related systems

Comments