Twenty20 May 17, 2009

T20 bowling - a lateral look

It is my pet theory that it does not matter which bowler the captain bowls in T20
32

A hectic and harrowing few days have left me longing for an article which will not have over 100 responses daily, quite a few of them calling for my blood. Hence before I do the follow-up article to the Test batsmen one, I decided to do a simple one which will elicit few, but informed, responses and will be of interest to the discerning readers.

It is my pet theory that it does not matter which bowler the captain bowls in T20.

If the batsman sets his mind, any bowler will go for plenty. This has been substantiated by the happenings in the IPL. The bowlers who have gone for plenty (around 20 runs) are the experienced bowlers such as Flintoff (twice), Agarkar, Vettori (19), Sreesanth (thrice), Ishanth Sharma, Vaas (the 23-run last over) et al. The lesser bowlers also did go but not like this. Hence I decided to put this, possibly unsound, theory to test.

This analysis pertains to T20 International matches. Reluctantly I have come to the conclusion that the IPL is not going away and am building my data base. So the IPL analysis will follow later. I may not be a great fan of IPL, but that entity is a treasure-trove for analysis.

Again the selection criteria plays an important part. I have set the criteria as players who have bowled in a minimum of 3 matches. Since the maximum number of matches for a bowler is 17, this represents a sub-20% lower limit. Initially I set 5 matches as the criteria but enough bowlers did not come through. These and the "all bowlers" data is also summarised later. This is the only criteria I have set. The measure used for classification is Balls per (Bowler) match.

The BpM measure ranges from 24.0 for a few bowlers to around 10.0 for some others. 141 bowlers satisfy this criteria (bowled in 3 or more matches) and are available for analysis. Out of these, 65 bowlers have bowled 21.0 balls or more indicating that these are very regular bowlers. 14 bowlers have bowled 15 balls or less. The middle group, between 15 and 21 balls per match can be termed as somewhat regular bowlers. These limits are used only for some form of classification and are not set in stone. Now let us look at how these bowlers, as a group have performed.

Only bowlers who have bowled in minimum of 3 matches

BpM  Bow  Balls   Runs Wkts  St/Rt   RpO

Very regular bowlers >=21 65 9825 11245 535 18.36 6.87 Regular bowlers 15-21 62 6298 7752 331 19.03 7.39 Irregular bowlers <15 14 825 1181 41 20.12 7.14

Total 141 16948 20178 907 18.69 7.14

The averaging has been done within a group. The first group denotes bowlers who have almost always completed their spells barring very few matches. The third group denotes bowlers who have rarely completed their spells. The middle group comprises of bowlers who bowled regularly in a support manner. Quite a few of them might have completed their spells also.

What is amazing is that the Strike Rate and RpO among the three groups does not vary that much. There is a differential of less than 10% between the values. If anything, the third group has a better RpO than the middle group.

Compare these with the total figures. Again less than 10% variation.

It seems to give substance to my own conclusion that the lesser bowlers do as well or as badly as the better bowlers particularly during the later half of the innings. No captain would like to take the opening overs lightly.

Now we can see two other summaries. The first is the one where the bar is set higher, to a minimum of 5 matches in which the bowler should have bowled. The second is the one including all bowlers who ever delivered a ball. Hence there is an additional classification in the last table.

Only bowlers who have bowled in minimum of 5 matches

BpM  Bow  Balls   Runs Wkts  St/Rt   RpO

Very regular bowlers >=21 40 7904 9105 427 18.51 6.91 Regular bowlers 18-21 20 2988 3676 167 17.89 7.38 Irregular bowlers <18 14 1754 2394 86 20.40 8.19 Total 74 12646 15175 680 18.60 7.20

All matches - No limit

BpM  Bow  Balls   Runs Wkts  St/Rt   RpO

Very regular bowlers >=21 105 11146 12960 602 18.51 6.98 Regular bowlers 18-21 62 4910 5960 269 18.25 7.28 Irregular bowlers 12-18 52 2932 3934 143 20.50 8.05 Casual bowlers <12 32 421 623 22 19.14 8.88

Total 251 19409 23477 1036 18.73 7.26

The results are self-explanatory. There is very little difference between the groups, even in the last one.

Does it mean the bowlers are irrelevant in T20 matches and that the team can field 11 batsmen, a few of them bowling their part-time spin/pace. No way. The team might score 250 quite frequently but almost always the opposing team would knock off these runs in 18 overs.

It is also true that in places like South Africa, where the current IPL is a revelation, quite a few totals of around 150 have been defended successfully. In these places I would advocate a 4 full time bowlers + 7 batsmen (incl keeper). Flintoff/IrfanPathan//Kallis would be considered full-time bowlers. The fifth bowler quota could be met from amongst the batsmen who could bowl. One cannot take any chances in these types of pitches where low totals might have to be defended more often than not.

However the situation is different on the very good batting pitches like the subcontinent where the regular bowler is as likely to go for as many runs as the part-timer. Here it makes sense to play only 3 full-time bowlers and meet the fourth and fifth bowler quotas from amongst the 3/4 batsmen/all-rounders. This would leave enough batting fire-power to chase any total. The adage of "6 batsmen should do the job" might hold good in Tests and, to a lesser extent, ODIs but not in T20 because of the high risks needed to be undertaken by the batsmen. However one should not compromise on the three bowlers. That should comprise of a very good opening pair and a top-flight spinner.

Let me sum this in a different way. I think the captain should be able to say, in a subcontinent match, that his strengthened batting lineup would chase whatever total the other team puts up and of course, up his own team's target score, batting first, by about 20%. On the other hand, the captain in a match in South Africa, should be able to say that his better set of bowlers would defend even a lower total and of course, be able to lower his own team's target score, batting first, by about 20%. These ideas go against the normal thinking that in subcontinent one should strengthen the bowling and on bowler-friendly pitches strengthen the batting.

Let me illustrate my idea with two team selections.

India

In the sub-continent In South Africa et al

Gambhir Gambhir Sehwag Sehwag Raina Raina Yuvraj Singh Yuvraj Singh R.Sharma R.Sharma Dhoni Dhoni Y Pathan Y Pathan A Nayar / Badrinath P Ojha Harbhajan Harbhajan Zaheer Khan Zaheer Khan R.P.Singh R.P.Singh

The only difference is at no.8 where in the sub-continent a batsman Abhishek Nayar/Badrinath plays while at South Africa, the bowler Ojha plays. These may not be the exact players to consider, but the reader should get the drift. The underlying assumption is that Ojha might go for as many runs as a combination of Nayar/Sharma but Nayar or Badrinath would give the batting edge. A similar exercise for Australia.
Australia

In the sub-continent In South Africa et al

Haddin Haddin Clarke Clarke Ponting Ponting M Hussey M Hussey Symonds Symonds Watson Watson Hopes Hopes D Hussey / Warner Hilfenhaus Johnson Johnson Lee Lee Bracken Bracken

Here the choice could be between D.Hussey/Warner who are primarily batsmen and Hilfenhaus, a bowler. D Hussey could as well bowl a few overs in India as Hilfenhaus.

The choice for Sri Lanka could be between Kandamby/Silva (batsman) and Thushara (bowler). For England it could be between Mascarenhas and Anderson. For Pakistan the choice could be between Fawad Alam and Shoaib Akhtar. And so on.

Umar Gul, Vettori, Afridi, Stuart Clark, Abdur Razzak, RP Singh, Shoaib Malik et al are representative of very good T20 bowlers and bowlers like these and others cannot easily be replaced by the casual bowlers.

To all the readers:

I have mentioned in my replies to your comments that all my analysis is based on the scorecard data since I do not have access to the proprietary ball-by-ball data. However I have seen that there is quite a lot of over-based data available in public domain.

I also realize that to do a proper analysis of the bowlers and their deployment, it is necessary for me to use what is available in the public domain and develop my own over-based data.

I have started work on this and once that work is over, expected to take about a week, I should be able to do much more in-depth analysis of T20 and IPL matches and then we can look at this particular analysis segment, amongst others, with lot more clarity. The middle overs, for that matter, the early and end overs, will cease to be a mystery.

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

Comments have now been closed for this article

  • ganesh on May 28, 2009, 7:22 GMT

    Its simple, at death overs mostly captains have given ball to exp bowlers like R.P,Munaf,Vaas etc...So there names are in top of list of most expensive bowlers. And anyway no body is perfect.....even if ur very good bowler in most of the matches there will some odd day when u will be surely very expensive.

  • Vijay on May 23, 2009, 23:32 GMT

    Ananth, Can you repeat the same analysis with a different classification of bowlers and see how it looks? Here's the classification I suggest: a) Strike bowlers - at least 3 overs i.e. >18 balls b) Support bowlers - at least 1 over but not more than 3 i.e. 6

  • Vishal on May 23, 2009, 19:06 GMT

    "The results are self-explanatory. There is very little difference between the groups, even in the last one. " Common! there is a difference of >1 run per over between very regular bowlers and irregular bowlers (last table) and almost 2 runs per over between very regular and casual bowler. That means if in a 20 overs match a team of irregular bowlers is conceding 20 more runs/innings (now if you look at the margin of victory and defeat isn't 20 significant??) The case becomes totally obvious when you compare very regular bowlers to team of causal bowlers the run differential is (20 x 1.9)= 38 runs!!!

  • Nick on May 21, 2009, 18:38 GMT

    Its not about what the score is after the 9th over, its about what the bowler concedes in the 9th over. Either way a 23 run over reflects badly and a 3 run over looks good. in any data set you are going to have outliers and strays, (maidens, 17 run overs) but i'm sure we can agree that on average more runs will be scored in the 18th over than the average score for the 7th over. [[ Nick The more I think about it, your point makes more sense. Looking at each over in isolation gives us 20 sets of data to work with. The problem only is the shortage of data. We have not even crossed 100 T20 matches yet. But these analyses would become relevant as we get more data. Shahid Africi has bowled 61 overs in T20 (the most). Mostly he would have bowled the middle overs. So there is a strong chance that he bowled, say, the 9th/10th/11th overs around 7/8 times. Against a total population of around 100 overs this type of analysis offers good possibilites. I will look at it after I build up the data. Thanks This is genuinely a lateral approach Ananth: ]]

  • Youvi on May 21, 2009, 4:09 GMT

    I believe T20 is going thru a phase where it is still to be figured out what really works so one could say with some basis a particular team has the best chance of being the champion. It doesn't matter whether it is IPL or T20 world cup. It seems therefore at this juncture in T20's evolution luck (or the unknown factor) plays a significant role. One doesn't seem to know what one is going to get on a given day. It is almost as if statistics at least the way these are applied, are not truly reflective of the T20 game itself. [[ Youvi, You are correct. Traditional playing methods do not apply to T20 nor traditional methods of analysis. I am hoping next week to come out with something different for both IPL and T20. Ananth: ]]

  • Nicholas on May 20, 2009, 20:03 GMT

    Here is an interesting idea (i'd do it myslef, but have neither the software, data nor time).

    Take each over, ie: 9th over of an innings. for each bowler who has ever bowled a 9th over, calculate the average 9th over excluding that bowler, then calculate the average deviation of that bowler from the average over. sum them up per bowler. You now have an average deviation for each bowler. Now plot these figures against an experience measure, perhaps total overs bowled or wickets taken (try either?) then see if these fit some sort of Regression or trend.

    i believe this method will account for discrepincies in which over is being bowled, as well as sort out the need for grouping bowlers by some arbitrary allocation. I also believe that it may indeed produce results that show that experience will indeed prosper. [[ Nicholas The problem, in addition to getting the data for each over (which I am working on), is that the score at, say, end of 9th over could vary from 32 for 5 to 97 for 0 (real IPL situations). Ananth: ]]

  • Pat on May 20, 2009, 8:18 GMT

    A good interesting topic and I appreciate the small sample size and the limited access, so I'll nit-pick:

    Australia

    In the sub-continent In South Africa et al

    Haddin Haddin Clarke Clarke Ponting Ponting M Hussey M Hussey Symonds Symonds Watson Watson Hopes Hopes D Hussey / Warner Hilfenhaus Johnson Johnson Lee Lee Bracken Bracken

    The subcontinent team has 5 players selected on bowling (Lee, Bracken, Johnson, Hopes and Watson), whereas the SA team has 6 (+ Hilfenhaus). It seems au-contraire to your theory with these teams.

  • Adam on May 19, 2009, 15:13 GMT

    Firstly - I don't agree at all with the extremely simplistic idea that an increase in RPO automatically means that T20 is "a batsman's game"; all it means is that we need different standards of excellence. A dominating bowling performance can win a match as easily as a dominating batting performance. The original analysis merely shows that those bowlers bowling in the middle innings (where on average less runs are scored per over), may only bowl 2 or 3 overs rather than finishing their spell (which we already knew). To attempt to read more into the data than that is dishonest. Anecdotal evidence suggests that the bowlers we might have expected to succeed don't necessarily do so - that says more about our faulty logic than a problem intrinsic to the format itself. T20 is a game similar in style to baseball - where both teams have a roughly even chance of winning any one game - which is why baseball is played over long series of 5 or 7 games, so the better team has time to emerge.

  • Kamlesh on May 19, 2009, 14:31 GMT

    There is hardly any difference in averages for the regular and not so regular bowlers. But if we see the differences between the economy rates, regular bowlers are quite miser compared to non-regular bowlers. Difference is 1 per over between very-regular and irregular bowlers and almost 2 per over between very-regular and casual bowlers. I don't agree with the theory of fielding a side with 3 full-time bowlers in subcontinent. As it would cost around 8*2=16 extra runs and hardly a no.8 player get a chance to bat. Your South African idea might work as it would cost just 8 run, but crucily a extra batsman would be handy in bowler friendly condition.

  • BASS 1298 on May 19, 2009, 12:56 GMT

    The analysis was fantabulous as always.as everyone say T20 is an absolute batsmen's game and this form does not judge a bowler's ability.R.P singh was the one who got 4 scalps in his first match against KKR and the same person was hit around the park in the second match against the same team.so it is really hard to speculate the ability of a bowler.It absolutetly depends on the batsman he faces and the mood of the batsman.Lee was exceptional in few matches but could not face gilly and symo though his team won.And when the best of the lots are hit muscled away to boundaries it is absolutely difficult to say who is good and who is bad.

  • ganesh on May 28, 2009, 7:22 GMT

    Its simple, at death overs mostly captains have given ball to exp bowlers like R.P,Munaf,Vaas etc...So there names are in top of list of most expensive bowlers. And anyway no body is perfect.....even if ur very good bowler in most of the matches there will some odd day when u will be surely very expensive.

  • Vijay on May 23, 2009, 23:32 GMT

    Ananth, Can you repeat the same analysis with a different classification of bowlers and see how it looks? Here's the classification I suggest: a) Strike bowlers - at least 3 overs i.e. >18 balls b) Support bowlers - at least 1 over but not more than 3 i.e. 6

  • Vishal on May 23, 2009, 19:06 GMT

    "The results are self-explanatory. There is very little difference between the groups, even in the last one. " Common! there is a difference of >1 run per over between very regular bowlers and irregular bowlers (last table) and almost 2 runs per over between very regular and casual bowler. That means if in a 20 overs match a team of irregular bowlers is conceding 20 more runs/innings (now if you look at the margin of victory and defeat isn't 20 significant??) The case becomes totally obvious when you compare very regular bowlers to team of causal bowlers the run differential is (20 x 1.9)= 38 runs!!!

  • Nick on May 21, 2009, 18:38 GMT

    Its not about what the score is after the 9th over, its about what the bowler concedes in the 9th over. Either way a 23 run over reflects badly and a 3 run over looks good. in any data set you are going to have outliers and strays, (maidens, 17 run overs) but i'm sure we can agree that on average more runs will be scored in the 18th over than the average score for the 7th over. [[ Nick The more I think about it, your point makes more sense. Looking at each over in isolation gives us 20 sets of data to work with. The problem only is the shortage of data. We have not even crossed 100 T20 matches yet. But these analyses would become relevant as we get more data. Shahid Africi has bowled 61 overs in T20 (the most). Mostly he would have bowled the middle overs. So there is a strong chance that he bowled, say, the 9th/10th/11th overs around 7/8 times. Against a total population of around 100 overs this type of analysis offers good possibilites. I will look at it after I build up the data. Thanks This is genuinely a lateral approach Ananth: ]]

  • Youvi on May 21, 2009, 4:09 GMT

    I believe T20 is going thru a phase where it is still to be figured out what really works so one could say with some basis a particular team has the best chance of being the champion. It doesn't matter whether it is IPL or T20 world cup. It seems therefore at this juncture in T20's evolution luck (or the unknown factor) plays a significant role. One doesn't seem to know what one is going to get on a given day. It is almost as if statistics at least the way these are applied, are not truly reflective of the T20 game itself. [[ Youvi, You are correct. Traditional playing methods do not apply to T20 nor traditional methods of analysis. I am hoping next week to come out with something different for both IPL and T20. Ananth: ]]

  • Nicholas on May 20, 2009, 20:03 GMT

    Here is an interesting idea (i'd do it myslef, but have neither the software, data nor time).

    Take each over, ie: 9th over of an innings. for each bowler who has ever bowled a 9th over, calculate the average 9th over excluding that bowler, then calculate the average deviation of that bowler from the average over. sum them up per bowler. You now have an average deviation for each bowler. Now plot these figures against an experience measure, perhaps total overs bowled or wickets taken (try either?) then see if these fit some sort of Regression or trend.

    i believe this method will account for discrepincies in which over is being bowled, as well as sort out the need for grouping bowlers by some arbitrary allocation. I also believe that it may indeed produce results that show that experience will indeed prosper. [[ Nicholas The problem, in addition to getting the data for each over (which I am working on), is that the score at, say, end of 9th over could vary from 32 for 5 to 97 for 0 (real IPL situations). Ananth: ]]

  • Pat on May 20, 2009, 8:18 GMT

    A good interesting topic and I appreciate the small sample size and the limited access, so I'll nit-pick:

    Australia

    In the sub-continent In South Africa et al

    Haddin Haddin Clarke Clarke Ponting Ponting M Hussey M Hussey Symonds Symonds Watson Watson Hopes Hopes D Hussey / Warner Hilfenhaus Johnson Johnson Lee Lee Bracken Bracken

    The subcontinent team has 5 players selected on bowling (Lee, Bracken, Johnson, Hopes and Watson), whereas the SA team has 6 (+ Hilfenhaus). It seems au-contraire to your theory with these teams.

  • Adam on May 19, 2009, 15:13 GMT

    Firstly - I don't agree at all with the extremely simplistic idea that an increase in RPO automatically means that T20 is "a batsman's game"; all it means is that we need different standards of excellence. A dominating bowling performance can win a match as easily as a dominating batting performance. The original analysis merely shows that those bowlers bowling in the middle innings (where on average less runs are scored per over), may only bowl 2 or 3 overs rather than finishing their spell (which we already knew). To attempt to read more into the data than that is dishonest. Anecdotal evidence suggests that the bowlers we might have expected to succeed don't necessarily do so - that says more about our faulty logic than a problem intrinsic to the format itself. T20 is a game similar in style to baseball - where both teams have a roughly even chance of winning any one game - which is why baseball is played over long series of 5 or 7 games, so the better team has time to emerge.

  • Kamlesh on May 19, 2009, 14:31 GMT

    There is hardly any difference in averages for the regular and not so regular bowlers. But if we see the differences between the economy rates, regular bowlers are quite miser compared to non-regular bowlers. Difference is 1 per over between very-regular and irregular bowlers and almost 2 per over between very-regular and casual bowlers. I don't agree with the theory of fielding a side with 3 full-time bowlers in subcontinent. As it would cost around 8*2=16 extra runs and hardly a no.8 player get a chance to bat. Your South African idea might work as it would cost just 8 run, but crucily a extra batsman would be handy in bowler friendly condition.

  • BASS 1298 on May 19, 2009, 12:56 GMT

    The analysis was fantabulous as always.as everyone say T20 is an absolute batsmen's game and this form does not judge a bowler's ability.R.P singh was the one who got 4 scalps in his first match against KKR and the same person was hit around the park in the second match against the same team.so it is really hard to speculate the ability of a bowler.It absolutetly depends on the batsman he faces and the mood of the batsman.Lee was exceptional in few matches but could not face gilly and symo though his team won.And when the best of the lots are hit muscled away to boundaries it is absolutely difficult to say who is good and who is bad.

  • Yogesh on May 19, 2009, 4:24 GMT

    RPO difference of 0.5 is quite significant in T20. What i see is that there is more uncertainity with respect to bowlers. Even your form bowlers can go to pieces a' la Balaji against KKR 'cos of a few bad balls. Except for a few picky ones, most bowlers have had bad overs once in a while. Hence you see captains wanting to have 4/5 bowlers plus 2-3 part timers. Because they feel that if one or two bowlers are hit(which is a given), they need back-up. Also in powerplays, they want to have four different bowlers. I saw many teams under-using one of their bowlers because the others managed to do the job on that day. So, what are we seeing is a theory contrary to yours ? Pick 4-5 bowlers (where two/three of them can bat) and 2-3 part-timers. If you are defending smaller totals, you might really need strike bowlers and not part-timers. So the current trend seems quite risk-less concept than yours. I think your theory is a case of statistic misleading people.

  • Rupert on May 19, 2009, 0:38 GMT

    There seem to be many comments pointing out that regular bowlers have to bowl both at the beginning and the death, making conditions far more difficult for them than for the part timers. I have a question related to that. During the first few overs, the prime bowlers are facing prime bats. But at the death !?! If they're still facing prime batsmen, then the batting team hasn't lost too many wickets - which means that the batters should be well set in the middle overs, and subsequently, should be blasting the part time bowlers out of the water in the middle overs. On the other hand, if the batters at the death are tailenders and they are hitting the prime bowlers as well as the top order bats, then probably experience doesn't count ... for the batters as well as the bowlers.

  • Peter Parker on May 18, 2009, 18:23 GMT

    In line with some of the other comments here, I think you underestimate the variance that you see in the numbers. A 10% difference in RPO translates to 15-20 runs in a typical T20 - and that is a significant number in this format. As for team composition - it is my hypothesis that in the subcontinent a team should need LESS batsmen, not more. Teams are losing a lot more wickets per inning in IPL2 than they did in IPL1. As such, a batsman at no.8 would most likely be wasted in the subcontinent. A specialist bowler will give the team more options in that position. In SA, however, a batsman at no.8 will probably be used at least in 50-75% of the matches, and in bowler-friendly conditions, you could trust your part-timers to do a decent enough job. [[ Peter I have deliberately projected a contrarian view to let people have a debate. I understand that the RpO differential is significant. However that difference is less pronounced when we fix a minimum limit for bowlers (even 3 matches). Ananth: ]]

  • Ananth on May 18, 2009, 11:09 GMT

    To all the responders: I have mentioned in my replies to your comments that all my analysis is based on the scorecard data since I do not have access to the proprietary ball-by-ball data. However I have seen that there is quite a lot of over-based data available in public domain. I also realize that to do a proper analysis of the bowlers and their deployment, it is necessary for me to use what is available in the public domain and develop my own over-based data. I have started work on this and once that work is over, expected to take about a week, I should be able to do much more in-depth analysis of T20 and IPL matches and then we can look at this particular analysis segment, amongst others, with lot more clarity. The middle overs, for that matter, the early and end overs, will cease to be a mystery. Ananth

  • adam on May 18, 2009, 10:19 GMT

    Excellent article, although I only half agree with your theory.

    The point was made about the specialisation of the bowling tasks throughout an innings. Attacks tend to be made up from 2 quick bowlers who bowl the first 5/6 and last 2/3 overs, and then a selection of 3 or 4 spinners and medium pacers who share the middle overs. Two decent quick bowlers who can bowl to a field at the start, and then bowl yorkers, bouncers, slower balls etc effectively at the death cannot be substituted for just any other bowler. In the middle overs, its more about the captain having options, and being able to chop and change the attack so that the batsmen can never get settled against one bowler. By and large, any half decent bowler can sneak in a couple of effective overs in this period, although having a dangerous spinner like Warne or Vettori as an option is clearly a bonus. I think the captain needs at least 4 options no matter what the pitch.

  • gautam on May 18, 2009, 8:02 GMT

    This is certainly very interesting analysis and a lot of valid comments/pushback from others. I do believe that the "skill" factor goes down considerably as the game reduces from 5 days to 1 day to 20 overs. Consistency is not rewarded and honestly i think thats one of the reasons we won the T20 world cup. Batsman like sehwag, yuvraj, yusuf etc can go without fear knowing that there are only 20 overs to play. That said i guess people who love the IPl would say its a different skill that's needed. i agree with your team selection - i blasted the mumbai team choice in an article on www.thecricketnation.com and would love your views. It is a batsman's game and it certainly is a captains game. Shane Warne has proven that time and again. btw , i think they'll get in by the skin of their teeth.

  • Mahek on May 18, 2009, 7:08 GMT

    In what world is 6.98 runs an over almost the same as 8.88? If you wanted to do a proper stats test you could have done an Analysis of Variance to see if there is any statistically significant difference in the numbers for the groups you mentioned.

    Coming to the cricketing side, the part-time bowlers usually bowl in the middle overs where batsmen aren't willing to take as many risks. Also, captains usually introduce their part-timers when there's a new batsman at the crease as he sees this as an opportunity to get through some overs. [[ Mahek I agree that in the group of "All bowlers" there is wide difference between the RpO figures of the top and fourth classifications. However this was only presented for information and should be ignored since this contains bowlers who have bowled a sum total of 5 balls. Even in this, note the RpW column. They are almost the same. For practical purposes we should only look at the 3 matches or 5 matches tables. Ananth: ]]

  • venkat on May 18, 2009, 3:36 GMT

    Ananth you are right your analysis is not skewed the data you have produced is very clear and it's a definite proof of that. I think it's a wrong choice of words on my part and probably preview screen on the comment section would have made me realize that before i posted the comment. But still the result or the conclusion of the article looks contrived to me. [[ Venkat The article might look contrived but the purpose was to steer the readers into a meaningful discussion on T20 team composition. Yesterday after seeing Munaf Patel's disastrous last over I have realized that once a quality batsman sets his mind any bowler, barring couple of bolwers like Malinga or Gul or Vettori would go for plenty. But that is mainly during the last 5 overs. Ananth: ]]

  • Observer on May 18, 2009, 2:49 GMT

    The conclusion seems to be one that would have to wait. For one what has been overlooked is that it is the "Very Reg" bowlers who bear the brunt of power hitting in 1st 6 and last 4 overs. That they still have a better economy should surely account for something. More importantly - overlooking even small differences in averages doesn't make sense. Not in IPL.. Diff between RPO (all matches - no limit) of "very reg" and "irregular" bowlers amounts to close to 11 runs. It will be interesting to note that how many matches have been decided with less than half of that separating the winner and loser. And with that information, can captains still afford to risk not having 5 at least "regular/very regular bowlers".. By your data itself the scenario is even more stark for casual bowlers. Don't have an analysis but a casual observation from last season when some captains did try the teams like you have suggested for sub-continent is that it was really difficult to hide any casual bowlers. [[ O, My feeling is that in the sub-continental wickets, as long as you have three very good bowlers one could have 8 overs (mind you, not 20 which would be required in ODIs) with bowlers such as Yuvraj, Sehwag, Dilshan, Sharma, Jayasuriya, Symonds, Clarke, Duminy, Gayle, Oram, Fawad Alam et al. However in South Africa this would be suicidal. What you have mentioned was tried mostly in ODIs which is certainly not advisable. Anyhow very few T20 matches (mind you, not IPL) have been played in the sub-continent. Ananth: ]]

  • Satyajit Roy on May 18, 2009, 0:53 GMT

    As a statistician you should realize that your sample size is very small; and in such situations any hypothesis made tend to be more circumstantial. Just as the 50-overs game gives more impetus for batsmen than bowlers, in the even shorter version, this is more so. BTW I never really understood why a limit exists for bowlers, when there is no equivalent limit on batsmen. That is where I feel the quality bowlers are at a disadvantage and you get these anomalies in figures coming in... Also the T20 format is still in infancy and evolving in terms of strategy. One interesting feeling I am getting is that the Captain's tactical role is very crucial in this shortest format. Decisions on bowlers and even batting order have to be constantly in focus throughout the 20 overs and the "thinking" captains are getting better results. The case in point is Sachin, who is not very cerebral, thus even with a glowing lineup the team is doing badly. Zaheer is on the bench and Dhawal playing. Warne/Dhoni [[ Satyajit I can only do with what sample I have. For that reason can I mix IPL and T20 Internationals. Ananth: ]]

  • Nicholas on May 17, 2009, 23:49 GMT

    surely the flaw here is not that the overs bowled at different time has effect on bowlers performance, but surely its a rather obvious fact that should a bowler bowl two shocker overs, his cptn is not going to bring him back on, irrespective of whether he's a specialist bowler or not. hence the quality of your bowling effects the number of balls you bowl, and thus the breakdown of regular etc bowlers is fundamentally flawed. The RPO of the"regular bowlers" was consistently the lowest for this very reason.

  • Aneesh on May 17, 2009, 22:03 GMT

    Interesting pet theory, and you're right, it seems to be supported by the data. If you have easy access to the raw data, it might be interesting to compute the standard deviation of the economy rates & strike rates. That would help put the small gap between the 3 groups in perspective.

  • James on May 17, 2009, 21:40 GMT

    Hi again Ananth, your theory is pretty good. But I think it is slightly biased by the usage of part timers. You don't see JP Duminy of David Hussey bowling the final overs, they bowl in the middle when the captain knows he can get away with it. So part timers always get the best conditions/situation to bowl in because the captain knows they are a big risk.

    I also disagree with your selection theory. I think the key in T20 is bowling OPTIONS. Sure in theory you could select just 3 or 4 specialists, but what if Hayden starts lashing into Sreesanth, or Rohit Sharma starts blasting Mortaza? You as captain don't want to bowl them for another over as it may cost you another 20+ runs, and a part timer is in even more danger. I personally believe you need 4 specialists and 2 ARs, one of which is a pretty good bowler eg. Flintoff, Afridi, I.Pathan. Because I see more advantage of a bowling option at #7 or #8 than a depth batsman. You shouldn't have to rely on a #7/8 batsman to win. [[ James Thanks. This is the sort of the comment I expected when I put up an almost provocative theory. I also believe that specialists should do their jobs. However this is applicable much more in Test matches and then ODI than T20. Ananth: ]]

  • arijit on May 17, 2009, 18:25 GMT

    I'm sceptical. Why then would Malinga be a better T20 bowler than, say, Dinda? Would Munaf, who choked Mumbai in the last over, have conceded 27 against Deccan as Mortaza did? Well, even if your theory is true, don't tell Lalit Modi -- it makes nonsense of his idea of the auction with high and low bids [[ Arijit, Amazing. As you were probably writing this mail, Munaf Patel bowled a 25-run over and probably cost the Rajasthan team a place in the Semi Finals. Ananth: ]]

  • Raghav on May 17, 2009, 16:54 GMT

    An example of what has been highlighted by some:

    Regular Bowlers (70%) Overs 1-5 & 16-20 @ 10 runs per over (30%) Overs 6-15 @ 6 runs per over Overall they get Economy of 10X0.7 + 6X0.3 = 8.8

    Irregular Bowlers (30%) Overs 1-5 & 16-20 @ 12 runs per over (70%) Overs 6-15 @ 8 runs per over Overall they get Economy of 12X0.3 + 8X0.7 = 9.2

    Though the overall economy if only 5% apart actualdiffernce is @25%. Can this theory be confirmed or negated.

    Add to this the fact that Irregular bowlers only bowl overs 3&4 if they are on a roll. Can you do an analysis of spells where entire quota of 4 overs is bowled. Irregular bowlers may be on top of regulars in this analysis because they bowl 4 overs only if it is their day.

  • venkat on May 17, 2009, 16:01 GMT

    Continuing on my previous post it would be shocking and probably erroneous to conclude from any analysis that it doesn't matter who is the bowler or batsmen in the middle in a shorter format. I think orange and purple cap standings proves without any doubt at the end of the day class stands apart. With data it is always possible to skew the analysis and contrive the results to the original intent when it comes to aggregation and averages. But at the end of the day a game is about winning and losing and here specialist talent and skill will succeed more times than it fails [[ Venkat Pl note that even though I started with a proposition and myself termed it as probably unsound, I have finally concluded with only a suggestion for playing one player or another. Your Orange Cap (or is it Purple Cap) bowler is in my team as well as other top bowlers. And kindly do not term my analysis results as skewed. there is nothing wrong in those, no assumptions nor any subjective considerations. Ananth: ]]

  • venkat on May 17, 2009, 9:36 GMT

    The major flaw in this approach is treating all 20 overs from 1-20 on the same scale and with same criteria. For overs 1-6 criteria and should be genuine opening bowlers against irregular bowlers. For 9-16 the criteria should remain the same are irregular vs regular bowlers. The big change in this exercise will come in the death overs where the criteria of putting regular bowlers against irregular bowlers is highly generic and looks the situation from a very broad perspective which may not give the desired benefit. Better approach would be put the contribution of specialist death bowlers against all other bowlers(regular+irregular) To name a few specialist death bowlers(agit agarkar,Pathan, Kallis and flintoff will not make this list) but whereas Malinga, Zaheer, vettori,RP singh( may be a stretch), Lee(may be a stretch), Ojha,Balaji,warne etc. If this analysis is made it could skew the result to confirm my hypothesis that it's about horses for courses not about regular or irregular [[ Absolutely right. I have not done any over-by-over analysis since I do not have the data for that. Ananth: ]]

  • Marcus on May 17, 2009, 9:20 GMT

    I don't really agree with your pet theory. There are some bowlers who've done very well in T20 (such as Umar Gul, Dan Vettori and so on) who can't really be replaced by part-timers. Also, part-timers, being part-timers, will receive a huge boost to their records if they do have occasional successes, whereas specialists who have a couple of bad games will experience quite a drain on theirs. For example, if David Hussey bowls just 2 overs and takes, say, 2-14, then his figures will look eminently respecatable, whereas if Vettori bowls his four overs for 0-42 then the impact on his record will probably be quite exaggerated.

    I also agree with Giri regarding batsmen relaxing against part-timers, but the same thing happens in every form. Look how many times Virender Sehwag, Jesse Ryder and Wavell Hinds have virtually secured all-rounder status at one time or another with a couple of important breakthroughs! [[ Marcus If you read the last paragraph of my article, I have mentioned exactly what you have commented. In fact you have referred to a subset of my bowler group mentioned therein. Ananth: ]]

  • Arvind on May 17, 2009, 8:23 GMT

    Its a follow up for my previous post as I ran out of the character limit

    The averages are very deceiving. Perhaps the differences in the quality of bowlers would be revealed in higher order moments. A good bowlers is more likely to perform better in any give over (at least thats what we would expect). So the over-wise performance distribution of a good bowler in twenty20 matched would be heavy tailed so better described by 'median' while for non-regular bowlers this distribution would be more or less symmetric and can be captured by 'averages'.

    I would be interested in seeing (as a figure ;) ) the average runs_per_over distribution of regular bowlers and non-regular ones. [[ Arvind Unfortunately I do not have over-by-over data. Otherwise I could easily do whatever you are suggesting. My data is the public domain one and as of now I don't have access to the proprietary data. Ananth: ]]

  • Arvind on May 17, 2009, 8:22 GMT

    This analysis provides support for my hypothesis that shorter the format of a cricket game, more role random performances would play and differences between the quality of players would not be revealed.

    As you analysis showed (on average) for bowlers it seems that the randomness become strong already at 20 overs format (where a bowlers get max. 4 overs). I guess for the batsmen this limit will be 10 overs i.e. if you play 10 overs games it wouldnt matter who is batting.

    In fact the later can be tested. If you study the 50 over matches by dividing them in 10 over segments, you may find that all batsmen perform nearly same in 10 over segments. The quality of the batsmen in revealed in how long they can bat. Similarly for bowlers the quality of the bowler is in maintaining the performance over 10-20 overs not for 2-4 overs.

    By the way regular bowlers have lower runs_per_over is likely to be due to the fact that they were allowed to bowl full spell because they were economical earlier

  • Ashwin on May 17, 2009, 7:43 GMT

    It would be interesting to apply this to IPL teams, and see what is the best combination each team can play, within the '4 internationals' limit. [[ Ashwin Amazingly one of the articles in the pipeline is to analyze the theoretical strength of IPL teams. I am in the process of develoing the club-based IPL Database and then I have to link the same with T20(Intl) and ODI and then only can I do this type of analysis. Ananth: ]]

  • giri on May 17, 2009, 6:21 GMT

    great analysis as always. i think a qualitative analysis can also be presented. usually the 'regular' bowlers bowl within the first 6 overs and at the death, in both situations, the batsmen look for quick runs. the part-timers operate in the middle overs, batsmen look for ones and twos with the odd boundary. also the batsmen sometimes relax aginst the part-timers, leading to dismissals. this would account for the almost similar strike-rates. [[ Giri What you say makes sense. Possibly the overs 1-6 and 15-20 are the overs where anything goes. Also during the middle overs, once the batsmen get into a safe 5/6-rpo mode, the part-timers should suffice, again I qualify, probably less so in SAF than Ind/Pak/Slk. Ananth: ]]

  • No featured comments at the moment.

  • giri on May 17, 2009, 6:21 GMT

    great analysis as always. i think a qualitative analysis can also be presented. usually the 'regular' bowlers bowl within the first 6 overs and at the death, in both situations, the batsmen look for quick runs. the part-timers operate in the middle overs, batsmen look for ones and twos with the odd boundary. also the batsmen sometimes relax aginst the part-timers, leading to dismissals. this would account for the almost similar strike-rates. [[ Giri What you say makes sense. Possibly the overs 1-6 and 15-20 are the overs where anything goes. Also during the middle overs, once the batsmen get into a safe 5/6-rpo mode, the part-timers should suffice, again I qualify, probably less so in SAF than Ind/Pak/Slk. Ananth: ]]

  • Ashwin on May 17, 2009, 7:43 GMT

    It would be interesting to apply this to IPL teams, and see what is the best combination each team can play, within the '4 internationals' limit. [[ Ashwin Amazingly one of the articles in the pipeline is to analyze the theoretical strength of IPL teams. I am in the process of develoing the club-based IPL Database and then I have to link the same with T20(Intl) and ODI and then only can I do this type of analysis. Ananth: ]]

  • Arvind on May 17, 2009, 8:22 GMT

    This analysis provides support for my hypothesis that shorter the format of a cricket game, more role random performances would play and differences between the quality of players would not be revealed.

    As you analysis showed (on average) for bowlers it seems that the randomness become strong already at 20 overs format (where a bowlers get max. 4 overs). I guess for the batsmen this limit will be 10 overs i.e. if you play 10 overs games it wouldnt matter who is batting.

    In fact the later can be tested. If you study the 50 over matches by dividing them in 10 over segments, you may find that all batsmen perform nearly same in 10 over segments. The quality of the batsmen in revealed in how long they can bat. Similarly for bowlers the quality of the bowler is in maintaining the performance over 10-20 overs not for 2-4 overs.

    By the way regular bowlers have lower runs_per_over is likely to be due to the fact that they were allowed to bowl full spell because they were economical earlier

  • Arvind on May 17, 2009, 8:23 GMT

    Its a follow up for my previous post as I ran out of the character limit

    The averages are very deceiving. Perhaps the differences in the quality of bowlers would be revealed in higher order moments. A good bowlers is more likely to perform better in any give over (at least thats what we would expect). So the over-wise performance distribution of a good bowler in twenty20 matched would be heavy tailed so better described by 'median' while for non-regular bowlers this distribution would be more or less symmetric and can be captured by 'averages'.

    I would be interested in seeing (as a figure ;) ) the average runs_per_over distribution of regular bowlers and non-regular ones. [[ Arvind Unfortunately I do not have over-by-over data. Otherwise I could easily do whatever you are suggesting. My data is the public domain one and as of now I don't have access to the proprietary data. Ananth: ]]

  • Marcus on May 17, 2009, 9:20 GMT

    I don't really agree with your pet theory. There are some bowlers who've done very well in T20 (such as Umar Gul, Dan Vettori and so on) who can't really be replaced by part-timers. Also, part-timers, being part-timers, will receive a huge boost to their records if they do have occasional successes, whereas specialists who have a couple of bad games will experience quite a drain on theirs. For example, if David Hussey bowls just 2 overs and takes, say, 2-14, then his figures will look eminently respecatable, whereas if Vettori bowls his four overs for 0-42 then the impact on his record will probably be quite exaggerated.

    I also agree with Giri regarding batsmen relaxing against part-timers, but the same thing happens in every form. Look how many times Virender Sehwag, Jesse Ryder and Wavell Hinds have virtually secured all-rounder status at one time or another with a couple of important breakthroughs! [[ Marcus If you read the last paragraph of my article, I have mentioned exactly what you have commented. In fact you have referred to a subset of my bowler group mentioned therein. Ananth: ]]

  • venkat on May 17, 2009, 9:36 GMT

    The major flaw in this approach is treating all 20 overs from 1-20 on the same scale and with same criteria. For overs 1-6 criteria and should be genuine opening bowlers against irregular bowlers. For 9-16 the criteria should remain the same are irregular vs regular bowlers. The big change in this exercise will come in the death overs where the criteria of putting regular bowlers against irregular bowlers is highly generic and looks the situation from a very broad perspective which may not give the desired benefit. Better approach would be put the contribution of specialist death bowlers against all other bowlers(regular+irregular) To name a few specialist death bowlers(agit agarkar,Pathan, Kallis and flintoff will not make this list) but whereas Malinga, Zaheer, vettori,RP singh( may be a stretch), Lee(may be a stretch), Ojha,Balaji,warne etc. If this analysis is made it could skew the result to confirm my hypothesis that it's about horses for courses not about regular or irregular [[ Absolutely right. I have not done any over-by-over analysis since I do not have the data for that. Ananth: ]]

  • venkat on May 17, 2009, 16:01 GMT

    Continuing on my previous post it would be shocking and probably erroneous to conclude from any analysis that it doesn't matter who is the bowler or batsmen in the middle in a shorter format. I think orange and purple cap standings proves without any doubt at the end of the day class stands apart. With data it is always possible to skew the analysis and contrive the results to the original intent when it comes to aggregation and averages. But at the end of the day a game is about winning and losing and here specialist talent and skill will succeed more times than it fails [[ Venkat Pl note that even though I started with a proposition and myself termed it as probably unsound, I have finally concluded with only a suggestion for playing one player or another. Your Orange Cap (or is it Purple Cap) bowler is in my team as well as other top bowlers. And kindly do not term my analysis results as skewed. there is nothing wrong in those, no assumptions nor any subjective considerations. Ananth: ]]

  • Raghav on May 17, 2009, 16:54 GMT

    An example of what has been highlighted by some:

    Regular Bowlers (70%) Overs 1-5 & 16-20 @ 10 runs per over (30%) Overs 6-15 @ 6 runs per over Overall they get Economy of 10X0.7 + 6X0.3 = 8.8

    Irregular Bowlers (30%) Overs 1-5 & 16-20 @ 12 runs per over (70%) Overs 6-15 @ 8 runs per over Overall they get Economy of 12X0.3 + 8X0.7 = 9.2

    Though the overall economy if only 5% apart actualdiffernce is @25%. Can this theory be confirmed or negated.

    Add to this the fact that Irregular bowlers only bowl overs 3&4 if they are on a roll. Can you do an analysis of spells where entire quota of 4 overs is bowled. Irregular bowlers may be on top of regulars in this analysis because they bowl 4 overs only if it is their day.

  • arijit on May 17, 2009, 18:25 GMT

    I'm sceptical. Why then would Malinga be a better T20 bowler than, say, Dinda? Would Munaf, who choked Mumbai in the last over, have conceded 27 against Deccan as Mortaza did? Well, even if your theory is true, don't tell Lalit Modi -- it makes nonsense of his idea of the auction with high and low bids [[ Arijit, Amazing. As you were probably writing this mail, Munaf Patel bowled a 25-run over and probably cost the Rajasthan team a place in the Semi Finals. Ananth: ]]

  • James on May 17, 2009, 21:40 GMT

    Hi again Ananth, your theory is pretty good. But I think it is slightly biased by the usage of part timers. You don't see JP Duminy of David Hussey bowling the final overs, they bowl in the middle when the captain knows he can get away with it. So part timers always get the best conditions/situation to bowl in because the captain knows they are a big risk.

    I also disagree with your selection theory. I think the key in T20 is bowling OPTIONS. Sure in theory you could select just 3 or 4 specialists, but what if Hayden starts lashing into Sreesanth, or Rohit Sharma starts blasting Mortaza? You as captain don't want to bowl them for another over as it may cost you another 20+ runs, and a part timer is in even more danger. I personally believe you need 4 specialists and 2 ARs, one of which is a pretty good bowler eg. Flintoff, Afridi, I.Pathan. Because I see more advantage of a bowling option at #7 or #8 than a depth batsman. You shouldn't have to rely on a #7/8 batsman to win. [[ James Thanks. This is the sort of the comment I expected when I put up an almost provocative theory. I also believe that specialists should do their jobs. However this is applicable much more in Test matches and then ODI than T20. Ananth: ]]