September 14, 2013

Great ODI recoveries

A look at the top successful ODI chases achieved from positions of no hope
43

Andy Bichel and Michael Bevan starred in the famous World Cup win against England in 2003
Andy Bichel and Michael Bevan starred in the famous World Cup win against England in 2003 © Getty Images

ODI wins achieved from the brink of disaster

The year 2012: Still fresh in the memory of many a football follower. Arsenal, yes, we are talking about the "team", played Reading FC. Arsene Wenger showed his disdain for this lesser cup by selecting a team with many second XI players. Reading were all over Arsenal and were 4-0 up in 37 minutes. An Arsenal win was as unlikely as an Indian win over Brazil in football. Theo Walcott reduced the deficit by half time. Then the recovery started. Arsenal made it 4-2, but only a minute of play remained. Two goals were scored in the last minute to restore parity at 4-4. In the extra time, the floodgates opened. Arsenal scored three goals and Reading once, for the match to finish 7-5. Wenger could smile once again.

Move back 55 years to 1957. Charlton FC is 1-5 down with 26 minutes remaining. Half the stadium has left the ground. Then Johnny Summers scored four goals in 17 minutes and John Ryan followed up with two more goals to more than compensate the additional Huddersfield goal. The match ended in a 7-6 win for Charlton, the greatest comeback in recorded football history and the narrowest win in English football history. In a frenetic second half, ten goals were scored.

A sunny Sunday during July 1999. Paul Lawrie started the last round of the British Open, ten strokes behind the leader, Jean Van de Velde. How does one explain this to the non-golf persona? Let me say it is the equivalent of a four-goal deficit at half time or a 300-run deficit at the end of the first innings. Lawrie had a terrific round of 67, Van de Velde had a monstrous meltdown and the regulation 18 holes ended in a tie between Lawrie, Van de Velde and Justin Leonard. It was poetic justice that Lawrie won the claret jug, winning the play-off. He completed the biggest comeback in major championship and PGA Tour history by coming back from ten strokes behind in the final round.

Finally let us use the time machine to roll a few years forward. Sunny Melbourne, January 2002. Australian Open Final between Jennifer Capriati, seeded No. 1 and Martina Hingis, seeded at No. 3. This had been a tournament of upsets, the top four men's seeds not lasting beyond second round. Twenty-four hours later the unfancied 16th seed, Thomas Johansson would win the men's singles. Hingis won the first set comfortably and broke Capriati twice to lead 4-0. Capriati fought back and took the final to a tie-breaker. Hingis led 6-2 and now had four championship points. Capriati saved them all, and went on to win the second set. She was broken by Hingis but then won five straight games to win an extraordinary match. This is the only time a Grand Slam title had been won after saving four match points.

This preamble would have given readers an idea about the theme of this article. I will be looking at ODI matches in which teams were so far behind that some of the betting companies stopped taking bets and half the stadia emptied. Then they slowly clawed their way back and won the match. I look at the top ten recoveries in depth in this article and provide a potted summary of a few other recoveries.

Although this is seemingly an anecdotal article, I will not do my usual selection of matches based on my memory, knowledge aided by visual inspection of the scorecards and some analysis. This is going to be 100% objective and analytical and based on new measures developed for this type of analysis. So readers can rest be assured that no match will be left out.

For reasons briefly explained in this paragraph, I will only look at the second innings and chasing wins. The first innings is difficult to define in view of the absence of a clear target. The target in front of the first batting team is a notional one which varies from period to period and place to place. A 250, which was an excellent score during 1980 at Headingley, would seem wholly insufficient during 2010 in Faisalabad. Not just the arbitrary nature of the target, but it would be difficult to get a correct handle on the situation in the first innings.

Let us say a team is 50 for 4, batting first. Way below the requirement to reach 250, the notional target. Let us say this team reaches 150 and then dismisses the opponents for 120 winning the match. This clearly indicates an awful pitch. So the 50 for 4 is far better than it looks. Taking the other side, let us say a team is 100 for 1. No doubt an excellent situation; they reach 300 and the other team chases down this score in 40 overs for the loss of two only wickets. The 100 for 1, which was seemingly a very good position, was rather inadequate. Hence the first innings is excluded in all matches. At a later date I might find a way to solve all these complex situations.

So I will consider only successful chases. Even there, we have a huge population of 1691 innings. As I have already mentioned, I am going 100% analytical and objective. How do I do this?

At any point in the game, in the second innings, there is a clear target in front of the chasing team. The innings has 11 clear markers in the form of innings beginning and loss of each successive wicket. It is clear that the situation is at a defining point at the fall of a wicket and keeps on improving until the next wicket falls. There are two types of resources available. Wickets and balls. The wicket-related data is available for all the 3400+ matches. However, balls available for the team is available only for the recent 1600 matches. However, I am able to bring that factor into the equation because the limiting resource is the one which is lower.

The wickets resource is the more important one since my calculations prove that at no time in the later part of the innings does the balls remaining become the limiting resource. I am going to compare the target in front of the team and the wicket resource available using two metrics - Target Resources Factor-Wickets (TRF-W) and Target Resources Factor-Balls (TRF-B) - to get a handle of the situation. I compare this with the ratio between the target in front and balls available just to complete the analysis. Just to give an example of the importance of TRF-W. For the TRF-B value to be above 3.00, a scenario like the following needs to be there. The batting team needs to score 36 runs in 12 balls and achieve that win. To locate such instances we need ball-by-ball data, which is some time away.

Wicket resource! Easier said than done. Each wicket carries a different value. The first wicket carries a high value and the last wicket, a fairly low value. To associate correct values for each wicket, I have determined the resource utilised percentage value at the fall of each wicket in every match played, compiled and averaged the same and got a set of values across 3410 matches. The final values seem very sound since there are over 6800 innings to work with. The values are given below. These values are somewhat similar to the D/L table values. However, I suggest that readers should not compare these with the D/L values. The learned academic duo has their methods and I have mine. They are fixing target scores. I am analysing target scores. This table is going to be a very useful one for many future analyses.

Wkt  Res-Utl%  Res-Avl%

1. 17.53% 82.47% 2. 30.87% 69.13% 3. 45.36% 54.64% 4. 57.97% 42.03% 5. 68.84% 31.16% 6. 77.39% 22.61% 7. 84.37% 15.63% 8. 90.16% 9.84% 9. 95.33% 4.67% 10. 100.00% 0.00%

The target in front is determined through a linear function, as also the balls available resource. Since there have been over a hundred rule changes in the 40-plus years of ODI existence, there is no point trying to incorporate these rule changes into either target or balls working. The linear mappings are simple and effective. These three values are determined at the fall of each wicket and a ratio arrived at by dividing the task in front by the resource available. This ratio is 1.00 at the beginning of the innings and becomes 0.00 at the fall of the last wicket. None of the matches that we have considered will reach this stage since we are only looking at chasing wins. A ratio above 1.00 indicates a tough task and a ratio below 1.00 is an easier task. Let me explain this through a few examples.

Target Score-at-FoW Target-% BallsRes% TRF-B  WktRes-%  TRF-W

200 20/2( 60) 90.0 80.0 1.12 69.13 1.30* 200 50/0( 60) 75.0 80.0 0.94* 100.00 0.75 200 100/5(150) 50.0 50.0 1.00 31.16 1.60* 300 40/1( 90) 86.7 70.0 1.23* 82.47 1.05 300 200/6(210) 33.3 30.0 1.11 22.61 1.47* 300 210/8(240) 30.0 20.0 1.50 9.84 3.04*
It is clear that situation 2 is favourable to the batting team and others not so good. The last one is quite a desperate one. If a team won from this situation, it would be quite creditable. In the second and fourth situations, the restricting factor is the balls resource. However it is very likely that the balls resource is likely to be a factor only in low wicket situations. It is almost certain that when more wickets are lost, the wickets lost would prove to be a much more critical factor than balls remaining. The factor values appended with * are the limiting values.

Out of the 3408 matches, there were 1663 wins by wickets and these teams went through well over 10000 wicket-fall situations. Out of these only ten situations were so desperate that the TRF was over 3.0. These are the matches featured here. There were 18 other matches which had situations between 2.5 and 3.0 and their potted scores are also provided. Please note that there might be more than one very tough situations (even TRF values exceeding 3.0) for one match. But the worst one in each match has been selected.

It should be noted that D/L matches are also covered here. In fact, there are two D/L matches in this lot of 29. Of course, there might have been multiple target changes within an innings, but that is not documented anywhere.

Finally let me emphasise that this is a pure scorecard-based analysis. No information beyond the scorecard is needed. Other contextual factors such as relative team strengths, batting and bowling strengths, quality of batsmen at the crease, importance of the match etc are not relevant to this analysis.

Javed Miandad, Lance Klusener, Brendan Taylor, Shivnarine Chanderpaul, Ed Rainsford and Ryan McLaren have won ODI matches with sixes off the last ball. It is possible that the TRF-B in these matches at the beginning of the last ball was as high as 6. But the purpose of this analysis is much more than identifying such instances.

First let us look at a table containing the details of ten matches.

List of matches in which the teams recovered from TRF situations of 2.5 and above
MatchIdFoW ScrBallsDLTgtMaxFB ScrSB ScrBlsFTRF-BTgtFWktFTRF-W
Featured
3065107/8152 240300239/8243/90.4931.1230.5540.0985.632
26203/9 0 267360266/7267/90.0000.0000.2400.0475.133
2499231/9262 285300284/4286/90.1271.4960.1890.0474.057
1028 74/7 0 173258172/9173/90.0000.0000.5720.1563.661
1976135/8226 205300204/8208/80.2471.3840.3410.0983.470
2922187/9269 222300221/9222/90.1031.5260.1580.0473.376
3323140/8193 209300208 ao209/90.3570.9260.3300.0983.355
2182147/8202 218300217 ao218/80.3270.9970.3260.0983.310
241 92/7 0 178300177/8180/90.0000.0000.4830.1563.091
2794 6/5 48 153300152 ao153/80.8401.1440.9610.3123.083
Included
2617114/7143 213300212 ao213/70.5230.8880.4650.1562.974
2632 64/6113 194300193 ao195/80.6231.0750.6700.2262.966
1799 82/6129 246300245/8248/80.5701.1700.6670.2262.951
2455178/9230DL206252223/8205/90.0871.5570.1360.0472.911
2797 44/6118 125300124 ao127/80.6071.0680.6480.2262.869
1537 71/6130 196294195 ao196/80.5581.1430.6380.2262.823
2403134/8228 185300184 ao185/80.2401.1490.2760.0982.802
83 61/6 0 164300163 ao164/80.0000.0000.6280.2262.780
679158/8 0 216300215 ao217/90.0000.0000.2690.0982.729
676152/9 0 174330173/8175/90.0000.0000.1260.0472.707
2375 89/7175DL154264157/9153/70.3371.2520.4220.1562.700
3358133/7222 230300229/9230/80.2601.6220.4220.1562.698
88 80/6 0 204300203/7207/90.0000.0000.6080.2262.691
3161 92/6134 226300225/8228/70.5531.0720.5930.2262.625
31 39/6 0 94360 93 ao 94/60.0000.0000.5850.2262.590
3127169/8238 226300225 ao227/80.2071.2200.2520.0982.563
2634 49/5 65 231300230 ao233/60.7831.0060.7880.3122.528
2265 95/6156 221300220/8221/60.4801.1880.5700.2262.524

There are ten matches which had TRF values exceeding 3.0 and these are featured here. A brief perusal of the scorecards will clearly reveal how tough the task for the chasing team was. Most of these situations have occurred at 7/8/9 wickets down and only one has been at the fall of the fifth wicket. The TRF-B values are also calculated and shown in this table. The table is self-explanatory. The potted scores for the other 18 matches are available in the downloadable document which is in text format and can be viewed by Notepad or similar editor.

The potted scores of the top ten matches, along with brief commentaries, are detailed below.

1. ODI # 3065. Australia vs Sri Lanka.
Played on 3 November 2010 at Melbourne Cricket Ground.
Sri Lanka won by 1 wicket. Mom: Matthews A.D.
Australia: 239 for 8 wkt(s) in 50.0 overs
   MEK Hussey          71*( 91)
   NLTC Perera         8.0  0  46  5
Sri Lanka: 243 for 9 wkt(s) in 44.2 overs
   AD Matthews          77*( 84)
   SL Malinga          56 ( 48)
   XJ Doherty         10.0  1  46  4
The match at the top of the list is of recent vintage and still fresh in everyone's memory. At MCG, Australia posted a very competitive total of 239 and reduced Sri Lanka to 107 for 8. Only 9.84% of resources are available and 55.4% of target runs are yet to be scored. This leads to a TRF of 5.632, the highest in ODI history, for winning teams. Angelo Mathews and Lasith Malinga got together and set up arguably the most exciting stand in all ODI cricket and added 132 runs. Then, with four runs to go, Malinga is run out. Muttiah Muralitharan came and swatted the second ball for four, taking Sri Lanka to the most unlikely win amongst 3400-plus ODI matches.
2. ODI # 26. Pakistan vs West Indies.
Played on 11 June 1975 at Edgbaston, Birmingham.
West Indies won by 1 wicket. Mom: Sarfraz Nawaz.
Pakistan: 266 for 7 wkt(s) in 60.0 overs
   Majid Khan          60 ( 90)
   Mushtaq Mohammad    55 ( 82)
   Wasim Raja          58 ( 87)
West Indies: 267 for 9 wkt(s) in 59.4 overs
   CH Lloyd            53 ( 76)
   DL Murray           61*( 95)
   Sarfraz Nawaz      12.0  1  44  4
The first World Cup in 1975. An important league match. Pakistan post an imposing total of 266. West Indies are devastated by Sarfraz Nawaz and slump to 89 for 5, 166 for 8 and finally, 203 for 9. The key factors are 24.8% of target runs to be scored and 4.67% of resources available, a very high TRF of 5.133. Also, nine wickets down, so no second chance like the first match. Deryck Murray and Andy Roberts put together, arguably, the best last-wicket partnership ever, of 64 runs, and West Indies script an unlikely win. They went on to win that World Cup, the next, and only their disdain of India prevented them from completing a hat-trick of World Cup wins.

This was one of the matches in which even the early situations such as 151 for 7 and 166 for 8 produced TRF values exceeding 2.75. But the worst situation was at the fall of the ninth wicket.

3. ODI # 2499. Kenya vs Ireland.
Played on 2 February 2007 at Ruaraka Sports Club Ground, Nairobi.
Kenya won by 1 wicket (with 6 balls remaining). Mom: Odoyo T.M.
Ireland: 284 for 4 wkt(s) in 50.0 overs
   WTS Porterfield    104*(131)
   KJ O'Brien         142 (123)
Kenya: 286 for 9 wkt(s) in 49.0 overs
   N Odhiambo          66 ( 82)
   TM Odoyo            61*( 36)
   AC Botha            9.0  0  42  4
   WK McCallan        10.0  2  36  4
Minnows they might be but Kenya and Ireland produced a cracker of match. Ireland, helped by two hundreds by William Porterfield and Kevin O'Brien, set Kenya a huge task of 285. Kenya lost wickets steadily and were at 231 for 9. The key numbers were 18.9% and 4.67%. The TRF was 4.057. Then Thomas Odoyo played a magnificent innings of 61 in 36 balls, supported by Hiren Varaiya, who faced ten balls. Kenya won by a wicket with an over to spare. The TRF based on balls was 1.496.
4. ODI # 1028. Australia vs West Indies.
Played on 1 January 1996 at Sydney Cricket Ground.
Australia won by 1 wicket. Mom: Reiffel P.R.
West Indies: 172 for 9 wkt(s) in 43.0 overs
   CL Hooper           93*( 96)
Australia: 173 for 9 wkt(s) in 43.0 overs
   MG Bevan            78*( 88)
   CEL Ambrose         9.0  3  20  3
New Year's day at the SCG. West Indies, tied down by Paul Reiffel and Shane Warne, managed to reach 172 for 9 in the allotted 43 overs. Curtly Ambrose ripped through the Australian top order and they were soon struggling at 74 for 7. The TRF worked out to 3.661, with the constituent values being 57.2% and 15.63%. Michael Bevan then played one of the best finishing innings ever. He added 83 with Reiffel and nursed Warne and Glenn McGrath to take Australia to a win with nothing but a wicket to spare. Australia were 169 for 9 with a ball to go and Bevan smashed a four through long-on. Truly, a memorable finish.
5. ODI # 1976. Australia vs England.
Played on 2 March 2003 at St George's Park, Port Elizabeth.
Australia won by 2 wickets. Mom: Bichel A.J.
England: 204 for 8 wkt(s) in 50.0 overs
   AJ Stewart          46 ( 92)
   AJ Bichel          10.0  0  20  7
Australia: 208 for 8 wkt(s) in 49.4 overs
   MG Bevan            74*(126)
   AR Caddick          9.0  2  35  4
World Cup 2003 played at South Africa. Andy Bichel, producing one of two greatest spells of fast bowling in a World Cup (who can forget Gary Gilmour's 6 for 14 in 1975), limited England to 204 for 8. The Australian batsmen failed to Andy Caddick and Ashley Giles and were floundering at 135 for 8. That man, Bevan, steady at one end, was joined by the bowler of the World Cup: Andy Bichel. The TRF was 3.470 (34.1% and 9.84%). Bevan and Bichel finished off the job on hand themselves, reaching 208 for 8, winning by two wickets. McGrath's debatable batting skills were not needed.
6. ODI # 2922. Bangladesh vs Zimbabwe.
Played on 5 November 2009 at Zohur Ahmed Chowdhury Stadium, Chittagong.
Bangladesh won by 1 wicket. Mom: Naeem Islam.
Zimbabwe: 221 for 9 wkt(s) in 50.0 overs
   BRM Taylor         118*(125)
Bangladesh: 222 for 9 wkt(s) in 49.0 overs
   Naeem Islam         73*( 90)
Again, two unfancied teams. Brendan Taylor, with an excellent hundred, helped Zimbabwe reach a modest 221 for 9. Bangladesh, with some atrocious running between the wickets, were looking down the barrel at 187 for 9. The TRF was 3.376 (15.8% & 4.67%). Naeem Islam was already batting well at 40. He farmed the strike beautifully, allowing Nazmul Hossain to face only four balls and took Bangladesh to a one-wicket win, with an over to spare.
7. ODI # 3323. South Africa vs New Zealand.
Played on 19 January 2013 at Boland Park, Paarl.
New Zealand won by 1 wicket. Mom: Franklin J.E.C.
South Africa: 208 all out in 46.2 overs
   F du Plessis        57 ( 72)
New Zealand: 209 for 9 wkt(s) in 45.4 overs
   JEC Franklin        47*( 61)
   R McLaren           8.4  0  46  4
We finally come to 2013. South Africa are dismissed for 208 through a wonderful bowling performance by Mitchell McClenaghan. But New Zealand could not face Lonwabo Tsotsobe and Ryan McLaren. They were at 140 for 8, which produced the highest TRF of 3.355 (33.0% and 9.84%). The situation at 105 for 7 was only slightly better. James Franklin and Kyle Mills added 47 and paved the way. Still, the ninth wicket fell at 187. Franklin scored all the 22 runs for the last wicket and garnered an excellent one-wicket win.
8. ODI # 2182. England vs West Indies.
Played on 25 September 2004 at The Brit Oval, London.
West Indies won by 2 wickets. Mom: Bradshaw I.D.R.
England: 217 all out in 49.4 overs
   ME Trescothick     104 (124)
West Indies: 218 for 8 wkt(s) in 48.5 overs
   S Chanderpaul       47 ( 66)
   A Flintoff         10.0  0  38  3
England reached an average total of 217, aided by a hundred from Marcus Trescothick. The English bowlers struck regularly and reached 147 for 8. This produced a TRF of 3.310 (32.6% & 9.84%). The unlikely pair of Courtney Browne and Ian Bradshaw got together and scripted an unlikely win for West Indies, adding 71 priceless runs.
9. ODI # 241. Pakistan vs West Indies.
Played on 28 January 1984 at Adelaide Oval.
West Indies won by 1 wicket. Mom: Marshall M.D.
Pakistan: 177 for 8 wkt(s) in 50.0 overs
   Wasim Raja          46 ( 40)
West Indies: 180 for 9 wkt(s) in 49.1 overs
   MD Marshall         56*( 84)
   Wasim Raja         10.0  1  33  3
   Abdul Qadir        10.0  1  34  3
Pakistan and West Indies produced another thriller. Pakistan reached a below-par total of 177. West Indies were up the creek without a paddle at 72 for 7. The TRF was 3.091 (48.3% and 15.63%). Malcolm Marshall donned the unlikely role of a batting saviour and added the remaining 88 runs in partnerships with Eldine Baptiste, Michael Holding and Wayne Daniel.
10. ODI # 2794. Bangladesh vs Sri Lanka.
Played on 16 January 2009 at Shere Bangla National Stadium, Mirpur.
Sri Lanka won by 2 wickets. Mom: Sangakkara K.C.
Bangladesh: 152 all out in 49.4 overs
   Raqibul Hasan       43*(107)
Sri Lanka: 153 for 8 wkt(s) in 48.1 overs
   KC Sangakkara       59 (133)
   M Muralitharan      33*( 16)
   Nazmul Hossain     10.0  3  30  3
The last in this lot of featured matches, producing TRF values above 3.0, was played recently in Bangladesh. Bangladesh reached a poor total of 152. It looked like a cakewalk for Sri Lanka. They had reckoned without Nazmul Hossain and Shakib Al Hasan. The first five batsmen scored 2, 0, 0, 1 and 1. The score was 6 for 5 leading to a TRF of 3.091 (96.1% and 31.16%). Then Kumar Sangakkara steadied the innings with Jehan Mubarak. Both of them got out and the score became 114 for 8. The TRF was quite high even now but not as bad as 6 for 5. Then Murali took over. With an unconventional innings of 33 in 16 balls, he saw Sri Lanka through. This is the only instance of a TRF value exceeding 3.0, with the loss of only five wickets.

I have created a document file with details of all matches in which the TRF values exceeded 2.5. This includes multiple occurrences within the same match. This document also includes the potted scores of the 18 matches which contained TRF values between 2.5 and 3.00. To download/view the document, please CLICK HERE.

While I was working on this article I realised that this analysis has a lot of implications for evaluation of live matches. I hope no broadcaster picks up the idea from here without giving me credit! Let me give a few examples from recent matches.

1. Ireland: 269. England: 48 for 4. The TRF-W was 1.95 (82.22/42.03). England won from this position. But they would not have made it to a list of top 50 matches.
2. Australia: 315. England: 103 for 5. The TRF-W was 2.14 (66.77/31.16). But they did not win.
3. Australia: 315. England: 169 for 8. The TRF-W was 4.72 (46.52/9.84). But they did not win.

So this is only a theoretical exercise at the end of the match if the chasing team did not win. But it indicates the difficulty of the task ahead of the teams and has a lot of relevance during a match. It also puts in perspective what the teams featured here achieved.

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

  • TheNewStatsman on October 30, 2013, 2:12 GMT

    Ananth, I just love this piece of work. Rigorous use of existing data to provide a perfectly defensive analysis.

    Re your remark: "While I was working on this article I realised that this analysis has a lot of implications for evaluation of live matches. I hope no broadcaster picks up the idea from here without giving me credit!"

    I've had the same idea re using Duckworth-Lewis to provide an ongoing score prediction (1st innings) or win-loss prediction (2nd innings), which I've written about at the New Statsman, on Wordpress (nerdy but non-boring stats for sports facns). This work far exceeds mine or any that I've seen on this issue.

    Is there any chance that cricinfo will ever make its data publicly available, so others can use it for research?
    [[
    Cricinfo is very protective of their database. Anything in the public domain is available fee. Any derivations from this data is not proprietary. So my feeling is that as long as the methodology is original there should be no problems.
    D/L is another kettle of tea (okay, fish). They may not allow their tables to be used for any analysis unless royalty is paid.
    Ananth
    : ]]

  • Anshu.N.Jain on September 20, 2013, 4:58 GMT

    I am reminded of a match between India and Sri Lanka, way back in the late 90s. India were chasing a stiff 302 (for those times!), and were 60-odd for the loss of 4 wickets. A TRF-W of just under 2 at that stage. Azhar and Jadeja both hit 100s, but India fell short by 2 runs! I also remember that night that there was a power cut in Vadodara, where I was then, and we missed out on a definite classic as it must have panned out in the later overs.
    [[
    Anshu, excellent selection. ODI # 1223 in 1997. At 64 for 4, the TRF-W was 1.836 (78.87%/42.96%). The later value is the second innings-only wicket resource available. So the situation was tough but nowehere near the selection.
    Now we come to the other side, the losses from absolutely winning positions which is the theme of my next article. Azhar and Jadeja took them to 287 for 4 but could not win the match since I do not know when the fifth wicket fell. Since Mongia, Chauhan and Kumble together played only 5 balls, one could surmise but unfortunately I could not include this type of match because of the uncertainties and the complexity involved. The Balls-resource which is a very important component of my next article cannot be used if it is not a correct and actual figure. This is a rare match which went from a certain loss, to a certain win and finished at a loss.
    Ananth
    : ]]

  • Anshu.N.Jain on September 18, 2013, 13:37 GMT

    Thanks Anantha for entertaining my suggestion :-) I have a question on the calculation of the wicket resource percentages:

    1. Did you calculate % of total score in an innings contributed by each wicket, for each innings, and then average each wicket % contribution across all innings of interest?

    2. Or did you add all runs scored for the 1st wicket across all innings of interest, and divide by total runs scored in all such innings of interest, to derive the 1st wicket resource percentage, and similarly for other wickets through to the 10th?

    The first approach normalises % contribution by each wicket across all innings, while the second gives as much weightage as the volume of total runs scored by that wicket across all innings. I kind of prefer the second approach over the first. In a sense, it removes the distortion due to normalisation at the innings level and then averaging, rather than an "in the aggregate" calculation. Which do you prefer and why?
    [[
    The first one is correct since out-of-the-normal partnerships are given their due importance. 50 for 1 in an innings of 150 correctly has 33.33% as the contribution in that innings. These sort of variations will tend to get lost when we add all runs together and then do the division. Hebce the first method is the adopted one.
    Ananth
    : ]]

  • on September 18, 2013, 12:55 GMT

    I havent done all the math myself, but dont you the think that the World Cup 2011 match between England and Ireland would have made it to the atleast, atleast as a notable mention. Ireland were 111 for 5 chasing 327. And (with due respect to Andrew Strauss and Mahela Jayawardene) Kevin O Brien Played the innings of the tournament.
    [[
    Thanks, Prabhakar, for pointing to an excellent match. ODI # 3114 in 2011. 111 for 5 chasing 328 leads to a TRF-W vale of 2.12 (66.15%/31.17%). Extremely difficult situation but does not come to the cut-off value of 2.50. Possibly if other factors such as the size of target et al are introduced, this might very well qualify.
    Ananth
    : ]]

  • Ukri82 on September 17, 2013, 16:02 GMT

    This is a brilliant idea. Especially the indices TRF-B and WktRes are beautiful indicators because they are purely derived from past data. Nothing arbitrary, great. And the final index starting with a 1.00 in the beginning of the chase changing gradually as the match progresses is also very aesthetic.

    [[
    Yes, I love the absolutely certain and finite start at 1.00 and the way it moves during an innings. It is too much work for me to do a graph for every match. But it is worthwhile doing this for selected match or two.
    Ananth
    : ]]

    Couple of comments. 1. I still didn't understand your comments to ras's idea to use the first innings score itself as the target in case the team batting first wins. When he is talking about using 266 as the target score of India and apply the same calculation, why do you mention about the number 250, which he never indicated in both of his comments? I believe his idea is a good one.
    [[
    Unni, I understood Ras's suggestion. However it was my reluctance to use a score which might very well be a losing score which prompted to write that way. Currently the target score is a winning score. This is because it is the second innings. For the first innings some more work has to be done. Your later suggestion and Ras's can be the starting point.
    I will next do the other half of the clear wins, the "by run" wins. Then I will move on to the first innings.
    Ananth
    : ]]

    2. My idea in to handle this case would be to look into the bowling of the second team. So, in which situations did the team bowling second pulled off an unlikely win by getting the other team out? The indices could be designed analogus to the batting case, I think (using inverses of the same indices).

  • AnanthNarayanan on September 17, 2013, 3:42 GMT


    [[
    Anshu,
    I have completed the Wicket resource usage matrix by innings and posted the results in the form of a down-loadable document. There are enough significant varistions to conclude that the analysis should use the concerned innings' figures. Especially for first and later wickets.
    I have given the link below for downloading the document since I cannot edit my article. You can cut & paste this into your browser URL line.
    http://dl.dropbox.com/u/39210851/odisum4.txt
    Incidentally the first wicket calculations in the current analysis had a minor error since the 60 second innings which did not get under way were noit done correctly. It has since been corrected and the first wicket value, across all 6750 innings is 16.72%. All other wicket values remain the same.
    However please note that the current analysis will work more or less similarly since only the later wickets are the key ones.
    Ananth
    ]]

  • ras on September 16, 2013, 4:18 GMT

    I think I have not explained myself properly regarding 1st innings analysis.

    For eg., In India v Zim 1983 WC, consider India's notional target in 1st innings as 266 (that is the final score achieved) and then work out a ratio at 17/5 w.r.t how difficult it is to reach 266 in the end with available resources at 17/5. The analysis can be restricted to only the matches won by team batting 1st.

    Of course, it can be done for completed matches only and can't be applied to live games.

    I hope I make myself clear.
    [[
    Ras, I have explained in details this problem in the article.
    I wanted this analysis to be free of assumptions and what Milind calls, golden numbers. 250 is one suchnumber. I also wanted this analysis to be 100% reflective of batting success, not diluted with bowling efforts.
    9 for 4 and 17 for 5 are terrible situations. For what. In this case, 250 is perfect. India went past 250 by 16 runs and Zimbabwe fell short by 15 runs. Perfectly symmetrical. Enough for you to think that 250 is perfect for all situations.
    What if India reached only 175 and Zimbabwe scored 150. Won't that be Kapil's win as well as the bowlers' win.
    What if Zimbabwe had chased down the total of 266. Where would Kapil Dev's effort be placed.
    In Tests, that is the difference between Laxman's 281 and 73. His 281 was not enough to win the Test. It still required the bowlers to deliver. On the other hand the 73 was a pure batting win.
    There may be a way out to analyze these situations by not having golden numbers but context-based numbers. I have not found a suitable method which would cover all situations. Eventually I will.
    Ananth
    : ]]

  • on September 15, 2013, 18:52 GMT

    I remember there was a match where Kumble & Srinath batted out of their skins to take India to a near impossible win. Though I don't remember full details, I expected more details about that game somewhere.
    [[
    Thanks, Chetan. Match no 1129 during 1996 at KSCA, Bangalore. Australia scored 215 for 7. India were at 164 for 8 and Kumble and Srinath took them home. The TRF-W was 2.44 (24.07%/9.84%) and the match just missed the cut-off. Possibly India's best recovery.
    Ananth
    : ]]

  • AnanthNarayanan on September 15, 2013, 8:11 GMT

    <br><b>[[<br> At 8 for 3 in the Cardiff match played yesterday, the TRF-W was 1.74 (95.24/54.64). Quite desparate but nowhere near the situations at fall of the 8th/9th wickets. I cannot but feel that the TRF is an invaluable live measure of the team situation at the fall of a wicket. <br> Ananth<br>: ]]</b><br>

  • sanjay_srivastava on September 15, 2013, 6:38 GMT

    @Ananta Great collapses (or recoveries due to bowling efforts) are the flip side of the same coin - losses from very low index values. It will be nice to see this list as well.
    [[
    I had thought of it and looked at losses from low values, in other words impregnable situations. But 150 for 1 while chasing 250, cannot be analyzed by itself. The 150 for 1 could have been at the end of 25th over or 40th over. So the Balls available resource plays a major part in that. Without that it does not make sense. It virtually rules out the first 1400 matches.
    Having said that I could re-look at it from a different point of view and will come out with such instances in a later article. I could always assign notional scoring rates and solve the problem.
    Ananth
    : ]]

  • TheNewStatsman on October 30, 2013, 2:12 GMT

    Ananth, I just love this piece of work. Rigorous use of existing data to provide a perfectly defensive analysis.

    Re your remark: "While I was working on this article I realised that this analysis has a lot of implications for evaluation of live matches. I hope no broadcaster picks up the idea from here without giving me credit!"

    I've had the same idea re using Duckworth-Lewis to provide an ongoing score prediction (1st innings) or win-loss prediction (2nd innings), which I've written about at the New Statsman, on Wordpress (nerdy but non-boring stats for sports facns). This work far exceeds mine or any that I've seen on this issue.

    Is there any chance that cricinfo will ever make its data publicly available, so others can use it for research?
    [[
    Cricinfo is very protective of their database. Anything in the public domain is available fee. Any derivations from this data is not proprietary. So my feeling is that as long as the methodology is original there should be no problems.
    D/L is another kettle of tea (okay, fish). They may not allow their tables to be used for any analysis unless royalty is paid.
    Ananth
    : ]]

  • Anshu.N.Jain on September 20, 2013, 4:58 GMT

    I am reminded of a match between India and Sri Lanka, way back in the late 90s. India were chasing a stiff 302 (for those times!), and were 60-odd for the loss of 4 wickets. A TRF-W of just under 2 at that stage. Azhar and Jadeja both hit 100s, but India fell short by 2 runs! I also remember that night that there was a power cut in Vadodara, where I was then, and we missed out on a definite classic as it must have panned out in the later overs.
    [[
    Anshu, excellent selection. ODI # 1223 in 1997. At 64 for 4, the TRF-W was 1.836 (78.87%/42.96%). The later value is the second innings-only wicket resource available. So the situation was tough but nowehere near the selection.
    Now we come to the other side, the losses from absolutely winning positions which is the theme of my next article. Azhar and Jadeja took them to 287 for 4 but could not win the match since I do not know when the fifth wicket fell. Since Mongia, Chauhan and Kumble together played only 5 balls, one could surmise but unfortunately I could not include this type of match because of the uncertainties and the complexity involved. The Balls-resource which is a very important component of my next article cannot be used if it is not a correct and actual figure. This is a rare match which went from a certain loss, to a certain win and finished at a loss.
    Ananth
    : ]]

  • Anshu.N.Jain on September 18, 2013, 13:37 GMT

    Thanks Anantha for entertaining my suggestion :-) I have a question on the calculation of the wicket resource percentages:

    1. Did you calculate % of total score in an innings contributed by each wicket, for each innings, and then average each wicket % contribution across all innings of interest?

    2. Or did you add all runs scored for the 1st wicket across all innings of interest, and divide by total runs scored in all such innings of interest, to derive the 1st wicket resource percentage, and similarly for other wickets through to the 10th?

    The first approach normalises % contribution by each wicket across all innings, while the second gives as much weightage as the volume of total runs scored by that wicket across all innings. I kind of prefer the second approach over the first. In a sense, it removes the distortion due to normalisation at the innings level and then averaging, rather than an "in the aggregate" calculation. Which do you prefer and why?
    [[
    The first one is correct since out-of-the-normal partnerships are given their due importance. 50 for 1 in an innings of 150 correctly has 33.33% as the contribution in that innings. These sort of variations will tend to get lost when we add all runs together and then do the division. Hebce the first method is the adopted one.
    Ananth
    : ]]

  • on September 18, 2013, 12:55 GMT

    I havent done all the math myself, but dont you the think that the World Cup 2011 match between England and Ireland would have made it to the atleast, atleast as a notable mention. Ireland were 111 for 5 chasing 327. And (with due respect to Andrew Strauss and Mahela Jayawardene) Kevin O Brien Played the innings of the tournament.
    [[
    Thanks, Prabhakar, for pointing to an excellent match. ODI # 3114 in 2011. 111 for 5 chasing 328 leads to a TRF-W vale of 2.12 (66.15%/31.17%). Extremely difficult situation but does not come to the cut-off value of 2.50. Possibly if other factors such as the size of target et al are introduced, this might very well qualify.
    Ananth
    : ]]

  • Ukri82 on September 17, 2013, 16:02 GMT

    This is a brilliant idea. Especially the indices TRF-B and WktRes are beautiful indicators because they are purely derived from past data. Nothing arbitrary, great. And the final index starting with a 1.00 in the beginning of the chase changing gradually as the match progresses is also very aesthetic.

    [[
    Yes, I love the absolutely certain and finite start at 1.00 and the way it moves during an innings. It is too much work for me to do a graph for every match. But it is worthwhile doing this for selected match or two.
    Ananth
    : ]]

    Couple of comments. 1. I still didn't understand your comments to ras's idea to use the first innings score itself as the target in case the team batting first wins. When he is talking about using 266 as the target score of India and apply the same calculation, why do you mention about the number 250, which he never indicated in both of his comments? I believe his idea is a good one.
    [[
    Unni, I understood Ras's suggestion. However it was my reluctance to use a score which might very well be a losing score which prompted to write that way. Currently the target score is a winning score. This is because it is the second innings. For the first innings some more work has to be done. Your later suggestion and Ras's can be the starting point.
    I will next do the other half of the clear wins, the "by run" wins. Then I will move on to the first innings.
    Ananth
    : ]]

    2. My idea in to handle this case would be to look into the bowling of the second team. So, in which situations did the team bowling second pulled off an unlikely win by getting the other team out? The indices could be designed analogus to the batting case, I think (using inverses of the same indices).

  • AnanthNarayanan on September 17, 2013, 3:42 GMT


    [[
    Anshu,
    I have completed the Wicket resource usage matrix by innings and posted the results in the form of a down-loadable document. There are enough significant varistions to conclude that the analysis should use the concerned innings' figures. Especially for first and later wickets.
    I have given the link below for downloading the document since I cannot edit my article. You can cut & paste this into your browser URL line.
    http://dl.dropbox.com/u/39210851/odisum4.txt
    Incidentally the first wicket calculations in the current analysis had a minor error since the 60 second innings which did not get under way were noit done correctly. It has since been corrected and the first wicket value, across all 6750 innings is 16.72%. All other wicket values remain the same.
    However please note that the current analysis will work more or less similarly since only the later wickets are the key ones.
    Ananth
    ]]

  • ras on September 16, 2013, 4:18 GMT

    I think I have not explained myself properly regarding 1st innings analysis.

    For eg., In India v Zim 1983 WC, consider India's notional target in 1st innings as 266 (that is the final score achieved) and then work out a ratio at 17/5 w.r.t how difficult it is to reach 266 in the end with available resources at 17/5. The analysis can be restricted to only the matches won by team batting 1st.

    Of course, it can be done for completed matches only and can't be applied to live games.

    I hope I make myself clear.
    [[
    Ras, I have explained in details this problem in the article.
    I wanted this analysis to be free of assumptions and what Milind calls, golden numbers. 250 is one suchnumber. I also wanted this analysis to be 100% reflective of batting success, not diluted with bowling efforts.
    9 for 4 and 17 for 5 are terrible situations. For what. In this case, 250 is perfect. India went past 250 by 16 runs and Zimbabwe fell short by 15 runs. Perfectly symmetrical. Enough for you to think that 250 is perfect for all situations.
    What if India reached only 175 and Zimbabwe scored 150. Won't that be Kapil's win as well as the bowlers' win.
    What if Zimbabwe had chased down the total of 266. Where would Kapil Dev's effort be placed.
    In Tests, that is the difference between Laxman's 281 and 73. His 281 was not enough to win the Test. It still required the bowlers to deliver. On the other hand the 73 was a pure batting win.
    There may be a way out to analyze these situations by not having golden numbers but context-based numbers. I have not found a suitable method which would cover all situations. Eventually I will.
    Ananth
    : ]]

  • on September 15, 2013, 18:52 GMT

    I remember there was a match where Kumble & Srinath batted out of their skins to take India to a near impossible win. Though I don't remember full details, I expected more details about that game somewhere.
    [[
    Thanks, Chetan. Match no 1129 during 1996 at KSCA, Bangalore. Australia scored 215 for 7. India were at 164 for 8 and Kumble and Srinath took them home. The TRF-W was 2.44 (24.07%/9.84%) and the match just missed the cut-off. Possibly India's best recovery.
    Ananth
    : ]]

  • AnanthNarayanan on September 15, 2013, 8:11 GMT

    <br><b>[[<br> At 8 for 3 in the Cardiff match played yesterday, the TRF-W was 1.74 (95.24/54.64). Quite desparate but nowhere near the situations at fall of the 8th/9th wickets. I cannot but feel that the TRF is an invaluable live measure of the team situation at the fall of a wicket. <br> Ananth<br>: ]]</b><br>

  • sanjay_srivastava on September 15, 2013, 6:38 GMT

    @Ananta Great collapses (or recoveries due to bowling efforts) are the flip side of the same coin - losses from very low index values. It will be nice to see this list as well.
    [[
    I had thought of it and looked at losses from low values, in other words impregnable situations. But 150 for 1 while chasing 250, cannot be analyzed by itself. The 150 for 1 could have been at the end of 25th over or 40th over. So the Balls available resource plays a major part in that. Without that it does not make sense. It virtually rules out the first 1400 matches.
    Having said that I could re-look at it from a different point of view and will come out with such instances in a later article. I could always assign notional scoring rates and solve the problem.
    Ananth
    : ]]

  • on September 15, 2013, 6:22 GMT

    There are three more instances of 9tth/10th wickets adding >30 runs in 2nd innings in tied matches:
    1. In tied ODI, there has been only 1 partnership of over 50 for the 9th/10th wickets (ODI 1235, Zim vs NZ,1 October 1997, chasing 233 NZ were 177/8 before 9th wicket of Harris and Larsen added 55, Larsen run out last ball for a tie)
    2ODI No 692: India vs WestIndies, Perth, 6th Dec 1991 Chasing India's 126, WI were 76-8, Ambrose and Cummins added 37 for 9th wicket
    3)ODI No 1821, South Africa vs Australia, Pochefstroom, 27 March 2002, Chasing South Africa's 259., Aus were 223 /9 in 45.2 overs, Maher and Hauritz added 36 for 10th wicket to ensure a tie
    4)ODI No 2258, England vs Australia, Lords, 2 July 2005, Chasing Australia's 196, England were 162/8 in 45.1 overs, 9th wicket of Giles and Gough added 32, before Gough fell at 194 on the second last ball. In my opinion, these 4 instances should deserve a mention along with this article.
    [[
    Let me say that I strictly went by wins. Ties are in a way wins for both teams. As such these deserve an honorary mention in this article. Since I do not have access to the article for editing, I will take youir presentation as the mention.
    The TRF-W for the first match was 2.38 (23.5%/9.84%). Let us not forget that only the 8th wicket had fallen.
    Ananth
    : ]]

    Also,I suggest that as a sequel to this article, can an article be done which analyses losses from strong positions of teams batting seconds. It could be under the heading great batting collpses, or bowling recoveries depends on how one perceives the match situation
    [[
    Pl see my response to Sanjay.
    Ananth
    : ]]

  • on September 15, 2013, 6:17 GMT

    How much is the TMF for the match with the record last-innings partnership between Mohammed Amir and Saeed Ajmal in 2009 (I know they lost the match, but I still want to know)? Here is the scorecard- http://www.espncricinfo.com/ci/engine/match/426722.html Thanks E.Shai
    [[
    Absolutely outstanding choice. Because my program only threw up wins this match was not seen. And it was difficult for me to see such near-misses.
    Dear ES, the TRF-W at 101 for 9, chasing 212, was a mammoth 11.21 (52.36%/4.67%). This would have been the leader, by the proverbial 100 miles, if Aamer and Ajmal had scored another 8 runs. Many thanks for bringing this wonderful match to light.
    Ananth
    : ]]

  • on September 14, 2013, 19:22 GMT

    where is 2002 natwest final.. india were 146 for 5 chasing 326 which itself would have been a world record chase then?
    [[
    The TRF-W at that point was 1.76. This is determined by dividing the target % ahead of India, 55.5% (181/325) by 31.16%, the resources available at the fall of the fifth wicket. This was a tough target no doubt but nowhere near these 8/9 wickets-loss situations. Just to get a perspective, compare the only 5-wkt loss situation featured here. Sri Lanka were 6 for 5.
    I agree that the number of runs to be scored were 181, quite high. But to offset that, this was a batting paradise, 470 for 10, at that point. Still I agree that the situation at 146 for 5 was quite tough, as evidenced by the TRF value of 1.76.
    Ananth
    : ]]

  • on September 14, 2013, 12:45 GMT

    Pakistan VS southafrica may also deserve a mention where Razzaq scored a centruy in UAE. It was played in 2011 i think.
    [[
    An excellent suggestion. Chasing 286/8, Pakistan were 136 for 5. The TRF-W at this point works to 1.68 - (52.6% (151/287) divided by 31.16%). But the situation was still worse later. At the fall of the ninth wicket the factor was 2,23 (10.45%/4.67%). In fact I think this match was much tougher than the 2002 Natwest Final. But not enough to make the cut.(
    Ananth
    : ]]

  • TheNewStatsman on October 30, 2013, 10:31 GMT

    Thanks for the response, Ananth.

    I also derive all my research from public-domain data, but I'm a one-person operation and can only give it 10 hours a week - I imagine your project has taken many more hours than this. The best I can do is to look at common statistical misconceptions or shallow use of statistics in sports media (often a misunderstanding of dependent variables or ignorance of confidence intervals or significance tests) & demonstrate how a genuine analysis *could* be undertaken - I just don't have the resources to actually do the research.

    Duckworth-Lewis may well ask for a fee, but I'm sure TV broadcasters can afford it. It would certainly add value well beyond the usually meaningless graphs and charts that our Australian coverage provides.

    Meanwhile, any Google-competent blogger can download a DL App for free, and use it for analysis, as I have.

    I look forward to reading more of your posts here, and directing my (very few!) readers to you.

    New Statsman / Wordpress
    [[
    To take one example, let us say the team score is 150 for 2 in 30 overs. Today the projections are straight from 5th standard arithmetic. 150+20*4, 150+20*5, 120+20*6 and 150+20*8 and the like. Any broadcaster with any self-respect could look at the batsmen left, bowler overs available, the pattern of scoring, the location and get a simple simulation going. The result could be anything between 265 and 335. I once told a broadcaster that I could mail them a simulated score at the end of each over. The answer was "It is too complex and no one will understand.". The real need was to keep it simple so that all advertisers would understand.
    Anyhow will mail you directly.
    Ananth
    : ]]

  • Anshu.N.Jain on September 30, 2013, 8:41 GMT

    Hi Anantha, On calculation of TRF-W again, in your method:

    (i) if a 2nd innings only had 4 wickets fall, do you assign 0% contribution to all other wickets for that particular innings, or do you discard this inning in the count when you average the % contribution across all innings for all other wickets (in this case, wickets 5 to 10)?

    (ii) a single inning where the % contribution from any wicket is disproportionately high will skew the average massively in any sample. This is especially true for late order wickets (7 to 10), where the frequency of disproportionately high % contribution is low, therefore the skew is more pronounced.

    An aggregate method removes both the above discrepancies.
    [[
    Anshu, will look into this.
    Ananth
    : ]]

  • on September 23, 2013, 11:26 GMT

    I would like to mention ODI no 682 between Pakistan and WI at Sharjah. Chasing 235 WI were 57/5 before Dujon and Richardson added 154, though Pakistan eventually won by 1 run. I doubt the TRF-W values would be high as comparable to theinstances mentioned in the articles, yet it should be considered as a great comeback considering the strength of Pakistan bowling and the venue.
    [[
    I cannot really alter the rukles to a near-victory. It does not qualify for a great win. Unfortunately it does not qualify for the second article also since there is lack of clarity at 211 for 5. How many overs were left? If 4 overs were left for scoring 26, it was relatively easy, if 3 overs, somewhat difficult and if 2 overs, quite difficult.
    The situation at 57 for 5 was quite difficult. TRF-W was 2.28 (76.27%/33.34%).
    Ananth
    : ]]

    Also a couple of requests: 1) What was the TRF-W value before the last over of the famous 1999 Aus-SA WC Semifinal 2) What was the maximum TRF-W achieved during SA chase of 434?

  • Anshu.N.Jain on September 20, 2013, 11:53 GMT

    Anantha If you prefer the average of inning wise percentages to an aggregate calculation, I suggest you also evaluate the option of using the wicket resource table for "chasing and winning" innings only. If the idea is to identify wins after good recoveries, the template is surely to be found in the sample of successful chases and not all chases. Further, this approach will eliminate those chases which resulted in losses because the middle and lower orders either couldn't build on the good work of the top order, or weren't called into service because the team ran out of overs.
    [[
    Too many subsets of data. I think we have to trust the population size taking care of minor variations. To me 305 & 302 for 8 is almost similar to 305 & 306 for 8. I do not want to lose the numbers available from either innings.
    Ananth
    : ]]

  • harshthakor on September 19, 2013, 11:38 GMT

    Anantha,I would still add India's chase of 326 runs after being placed at 146-5 considering the likes of Tendulkar,Dravid,Ganguly etc were all dismissed,the run rate more than 8 runs per over and the target above 325.

    I would also add the Aussies win in 2001-2002 against the Kiwis in the final league game of the Carlton and United series when Bevan steered his team home by 2 wickets.

    To remind you a game that could have been a classic win was New Zealand's chase of South Africa's target of 302 runs in the 1997-8 Carton and United series in Australia from positions of 129-6 and 220-8 to lose by 2 runs.The hitting by Adam Parore and Chris Cairns was extraordinary ,aided by Dion Nash.Inspite of run -rate climbing to over 10 runs the kiwis almost crossed over the line.

    Another class win would have been Sri Lanka's chase against West Indies in 1995 chasing above 300 an from about 130-6 coming withinn 4 runs of the target.

  • harshthakor on September 16, 2013, 12:19 GMT

    Ananth,I may also add the match between India and Pakistan at Eden Gardens in 1987 when Pakistan needed 65 of 6 overs with just 4 wickets in hand and Salim Malik's unbeaten 72 carried them over the line by a margin of 2 wickets .I can also remember Pakistan chasing Australia's target of 260 by a margin of 2 wickets in 1986-87 in the 4 nation tournament after being precariously placed at 120-6 and 190-8.Qasim Omar made a major contribution in the latter game aided by the likes of Wasim Akram.The Ist match of the 1987 Reliance cup also had a dramatic chase by England who won needing 34 of the last 3 overs chasing a target of 220 runs.Alan Lamb took them home after England were on the verge of losing at 111-6.The 1987 last match of the one day series at Edgbaston between England and Pakistan also had a great twist with the hitting of Defreitas stealing a win for England by a margin of 1 wicket. Anway,good work Ananth,Appreciate your choices.
    [[
    No choices. The algorithm chose the matches for me.
    1. Ind-Pak 1987 (Odi 474). Surprisingly the situation was not as difficult as it seems. The worst situation was when Pak were 174 for 6, chasing 239. The TRF-W was only 1.02 (23.01%/22.61%). But I know the scoring rate was really the problem. However no data for this is available.
    2. Pak-Aus 1987 (414). The TRF_W at 129 for 6 was 2.34. At 224 for 8 it was 1.85. So it is clear that it was very tough all through.
    3. Eng-Win 1987. (452). 162 for 7 and 209 for 8, chasing 244 are not really that difficult situations.
    4. Eng-Pak 1987 (450). Similar to 1 and 3 above. 167 for 8 chasing 214 is not a desperate situation.
    Overall very nice selections, all within a 12 month period.
    Ananth
    : ]]

  • cric_options on September 16, 2013, 9:17 GMT

    Ananth - I might have missed this part somewhere, if you have already explained it in the article or the comments, but as some readers have pointed out, doing this exercise for 1st inning batting team wins - I propose, can't we just consider the first teams total in such calculation as the eventual total of the second team (at which they collapsed or ran out of balls). That would take care of eliminating the contribution made by the bowlers of the winning team in the match. Thanks !
    [[
    These are some of the options I have to consider when I look at this topic again. The fact is that the clearest and cleanest of situations exist in wicket wins. So I went for it first.
    Ananth
    : ]]

  • calcu on September 16, 2013, 8:57 GMT

    Could you make a list of those matches in which The TRF-W was least, yet the team lost it?
    [[
    Pl look at the responses given to Sanjay and Pawan.
    Ananth
    : ]]

  • ras on September 16, 2013, 2:55 GMT

    I think the first innings notional target can be taken as the actual no. of runs scored by the team in the end. And then ratios can be calculated at fall of each wicket to gauge how difficult it was to reach this target from a given position. Only matches won by team batting 1st may be considered. Of course, the method seems so simple that I think you would have considered and discarded it because of some inherent statistical flaw. Please enlighten me.
    [[
    I think the enlightement has to come from your end. What exactly are you referring to. The first innings situations. If so, please read the article carefully. Notional targets are what say they are, notional. The notional target at Faisalabad during 2010 could be 325 and at Hamilton in 2002 could be 175.
    Ananth
    : ]]

  • on September 15, 2013, 22:45 GMT

    Hi Anantha:

    Wonderful work with these numbers. I differ with you on certain finer points, but the entire work you do is absolutely phenomenal. Terrific stuff!

    Thanks for all the wonderful efforts. I know you are a Lara fan, slightly ahead of SRT, I am a SRT fan, slightly ahead of BCL! [sounds like a t-20 league, but those are Lara's initials.... :)
    [[
    Thanks, Roger. It is very easy to idologize Federer and Lara. What is important is to have immense respect for and appreciate Tendulkar, Kallis, Djokovic and Nadal.
    Ananth
    : ]]

  • on September 15, 2013, 20:47 GMT

    This doesn't take into account which wickets were remaining, the quality of the bowling attack, the quality of the batsman at the crease. Those factors may alter the equation a bit. 107/8 with Kumble and Srinath at the crease is different from 107/8 with Ganguly and Kumble at the crease.
    [[
    You should never complicate a simple analysis. As it is, it is easy to understand. I could have brought in all sorts of other variables. Why, I will modify your own example. If Ganguly was coming off a horrible stretch of 100 runs in the last 15 innings and Srinath was coming off a purple patch of 100 runs in the last 5 innings. And what about Wasim and Waqar each having 5 overs to bowl as compared to one of them having bowled out. And Powerplay yet to come. And so on.
    Ananth
    : ]]

  • cric_options on September 15, 2013, 19:05 GMT

    Awesome piece of analysis Ananth. Quite interesting to know that India does not feature in the top 10, either as the winning or the losing team. Wonder what it tells about the Indian team.

    An offshoot of these calculations would be to normalize the career contributions of ODI batsmen based on their batting positions and see where they would stand in this new measure and how often they contributed greater or lesser than expected.

    To those matches where the ball data is available, maybe incorporate that too and see how this analysis would look like. - Thanks Som.
    [[
    Yes, I agree it opens new areas of analysis. Especially opening and coming in 4/5 down. Many great batsmen have come at no.6/7 and they have certainly faced more difficult situations. Maybe less hostile bowling but situations were certainly tough.
    Ananth
    : ]]

  • Anshu.N.Jain on September 15, 2013, 16:50 GMT

    contd.. (2/2) 2. In general, the resource utilised percentage table could be used to gauge the relative ease or difficulty of wins or losses of a particular team while chasing. One of my hypotheses is that in the last 10 years, while chasing in ODIs, India has, in general, won narrowly and lost heavily. This hypothesis could be tested if the TRF-W values across all chases, or across all chase wins or all chase losses, could be used to derive a composite figure which will help understand how easy or difficult a team has made it for itself while chasing and winning. As suggested by several other readers, losing from easy positions could also be similarly analysed. This analysis for to-date history, or particular slices in time, or for/against a particular team (say India vs Australia in the Tendulkar era) could be very interesting. Great attempt with this innovative idea.
    [[
    I will certainly look at expanding the scope of wicket resource utilization analysis.
    Ananth
    : ]]

  • Anshu.N.Jain on September 15, 2013, 16:48 GMT

    More thoughts (1/2) 1. It would be interesting to see the resource utilised percentage values for various slices: a. By 1st and 2nd innings b. By decades c. By teams Combination of two or more of these factors will be very interesting. It would convey what strategies teams use to chase successfully. India in the tendulkar-ganguly and tendulkar-sehwag era should have a higher 1st wicket percentage value than the average for all teams, implying that Indian batting in this era was served well by its openers in successful chases. NZ on the other hand is likely to depend heavily on its middle and lower order, given the plethora of their bits and pieces players. This should be reflected in the NZ specific resource utilised percentage table.
    [[
    A lot of work, Anshu. This is possible only if a separate article is dedicated to this topic. There is nothing wrong in doing that, sometime in the future.
    Ananth
    : ]]

  • Anshu.N.Jain on September 15, 2013, 16:46 GMT

    contd.. (2/2) 4. Looking at the Wicket resources table again, the 1st wicket is worth the most in % terms at 17.53%, and quite obviously so. Each wicket is progressively worth lesser percentage resource as we move from wicket 1 to wicket 10. The trend is however broken when wicket 2 (30.87-17.53=13.34%) is worth less than wicket 3 (45.36-30.87=14.49%). Assuming my understanding of its calculation in point 1 above is correct, does this imply that on average, in ODI history, the 2nd wicket has contributed lesser to the total target chased than the 3rd wicket?
    [[
    It is quite possible that the game dynamics are such that the 1st wicket falls within the first 10 overs and exposes the second wicket partnership slightly more than the third wicket partnership. It is also possible that the no.4 batsman has been very good. But these are actual numbers over 6428 innings and the message sent by these numbers is sacrosanct.
    Ananth
    : ]]

  • Anshu.N.Jain on September 15, 2013, 16:44 GMT

    (1/2) This is a nice idea. Quick thoughts: 1. I could not gather the working used to calculate the resource utilised percentage value at the fall of each wicket. Is this the average of % target score scored by each wicket across all innings?
    [[
    Certainly not target score. That has no meaning for this computation. Take a match scoreline of 300 & 200. What is important are the numbers, 300 and 200, which are actual scores. 250 (the notional target score) and 301 (the second innings target score) really have no meaning. So the denominator is always the actual score at the end of the innings.
    Ananth
    : ]]

    2. I'd think that the resource utilised percentage value table at the fall of each wicket would be different for first innings and for second innings. It would be interesting to look at these two tables for each innings (setting and chasing) separately. 3. I suggest that we only use values in the resource utilised percentage value at fall of wickets derived from all 3400+ chase innings only, and not all 6800+ innings. As you yourself say, the target is clearly defined only in the second innings.
    [[
    I have taken all 3413 matches. And both innings. It is quite possible that the values would be slightly different if I do it by innings. I will do it and post the information to get a good idea of how the distribution changes. If the diffrence is significant it is worth considering the values separately for different innings.

    Ananth
    : ]]

    contd.

  • ras on September 15, 2013, 15:04 GMT

    I distinctly remember the Ajmal-Aamer game in UAE. We had a power cut just after the fall of 9th wicket & I went to sleep. The power came back within 15 mins., but I didn't bother stirring up thinking that all would have been over by then. (NB:I am of the kind who usually doesn't like to miss even a ball). Next day after seeing the scorecard I realized that I had missed a gem.

    Nice article btw, but I think some way can be found to quantify 1st innings recoveries, because they are as important as 2nd innings.

    Thanks

  • on September 15, 2013, 12:14 GMT

    In number 1 take nothing away from sri lanka but clarke's captaincy we deplorable. malinga and matthews hitting everything down the ground yet no fielders put out there
    [[
    Clarke was probably hoping that there would have been near-catches. Possibly the fact that this was only the second match as ODI Captain for Clarke might have caused this.
    Ananth
    : ]]

  • RaviMarathe on September 15, 2013, 10:27 GMT

    Ananth, good article. I need to re-read it before I give more comments. Till then a couple of questions reg. match saving efforts batting first and setting competitive targets - Kapil Dev's performance against Zimbabwe in 1983. and Aus vs WI at Mohali in the 1996 world cup. Have you considered them?
    [[
    Kapil's was in the first innings. What was the notional target in front of the Indians. As I have explained in the article 10 for 4 leading to 150 all out could still win the match. Same with Mohali. 207 was a winning total. So what was the notional target at 15 for 4.
    Ananth
    : ]]

  • Vyasa_Shastry on September 15, 2013, 9:08 GMT

    2 of 2

    2) You rightly pointed out that wickets are more valuable that balls remaining in most scenarios, but there must be a way to combine the 2 difficulties (a simple sum or otherwise). Scoring 36 in 12 balls is more difficult lower down the order than say... at 2 down knowing that you have more batters to come after you. I liked the approach where you looked at difficulty of batting conditions in various test matches based on RPW and RPO and then tabulating the runs scored. Don't you feel it would be more appropriate to use the same philosophy here (how do you deal with those zeros?)?
    [[
    Where data is available I have considered the Balls-resource-factor and compared it with Wickets-resource-factor. Pl see the few examples given.
    Ananth
    : ]]

  • Vyasa_Shastry on September 15, 2013, 9:03 GMT

    Hello Anantha,

    Very nice step in this direction. I did not understand the column titled Balls in your table. How come some of them are 0? What does this zero represent? The others are balls remaining at the FOW scenario. I think I have missed something in this article. Besides this, I have a few comments to add:
    [[
    That means no balls data was available. If you re-rtead the article carefully I have explained this clearly.
    Ananth
    : ]]

    1) A point just for academic interest, but wouldn't it be more relevant to change the weightage for each wicket based on the last 5 years data rather than ODIs to date? It might be more accurate since batting strength might have increased in the lower order, no (considering that many of these matches have 6 down scenarios). What would be even better is to give the CTD batting average (batting average is also not bad) of the 2 batters at the crease to give a better perspective of how unlikely the scenario was in those specific 30 cases!
    [[
    Too much nit-picking and complicating will spoil the essential simplicity of the analysis. The numbers look very good and are applicable for all the 40 years. If the 9th wicket resource for the past years is 0.9487 instead of 0.9533, what does it mean.
    Ananth
    : ]]

    1 of 2

  • Bonehead_maz on September 15, 2013, 6:26 GMT

    lol - this is a bad day for you to bring this up Ananth ;) Yes - I actually watched it .... desperation slowly leading to exasperation ? Will comment on what I'm sure is another excellent article when I've read it. Meantime, the specials article on Thommo 1973 is a corker, and worth a read. I class Johnny Pym as a friend and never met a cricketer so enlivened with theories of how to better perform . LMAO he once batted against me with a bat full of holes so the air wouldn't resist it's downswing! Sandy Morgan a likewise "legend" to me ! Barry Knight - the supplier of my training facility and a great bloke, Greg Bush a great bloke and wonderful team mate. I played U 21 reps a year earlier than that story as a 14 yo opening batsman - needless to say Thommo destroyed any rofl "Sachin" in me ! Took me another 3 years to take the easy single to fine leg - needed another 10 years having no thigh pad to fully overcome. ( if I don't score off that I deserve to be hurt etc etc lol)

  • on September 15, 2013, 6:10 GMT

    A great analysis. I fully agree with your selection of matches won batting second in situations where it was thought that the loss for that team was inevitable at some stage of the match. This analysis also prompted me to use statsguru to find such matches where partnerships of 9th-10th wicket (>=30 runs as criteria) almost carried a team to victory. i confine my analysis to ties only in this comment.

  • on September 14, 2013, 22:00 GMT

    awesome analysis dear... thumbs up...

  • mad_ush on September 14, 2013, 20:51 GMT

    Nice article. Stats bewildered me, but the summery took is more than enough to spark the intensity of those matches and to relieve those moments.

  • on September 14, 2013, 19:43 GMT

    How can we forget 6/23 by Ashish Nehra vs England in 2003 worldcup ? Was not that great spell? :(
    [[
    This article is not about great bowling spells.
    Ananth
    : ]]

  • on September 14, 2013, 15:23 GMT

    @Anantha Are you planning to do a similar one for Test in fourth innings?
    [[
    I have been thinking about that also but Test scene is a lot more complex because of the non-limiting nature. Also a draw is a valid and (often successful) rtesult. But I will certainly do it some time in the future.
    Ananth
    : ]]

  • Charith99 on September 14, 2013, 8:54 GMT

    Had malinga taken the fifth wicket against SA during 2007 World Cup match it would have even topped his batting performace in the above match.

  • on September 14, 2013, 6:50 GMT

    Not much of a stats guy. But just goes on to show that even when measured objectively cricket really can still be a game of uncertainty. Proud of Naeem Islam. Slept over the game, and heard about the facts from a friend.

  • on September 14, 2013, 6:50 GMT

    Not much of a stats guy. But just goes on to show that even when measured objectively cricket really can still be a game of uncertainty. Proud of Naeem Islam. Slept over the game, and heard about the facts from a friend.

  • Charith99 on September 14, 2013, 8:54 GMT

    Had malinga taken the fifth wicket against SA during 2007 World Cup match it would have even topped his batting performace in the above match.

  • on September 14, 2013, 15:23 GMT

    @Anantha Are you planning to do a similar one for Test in fourth innings?
    [[
    I have been thinking about that also but Test scene is a lot more complex because of the non-limiting nature. Also a draw is a valid and (often successful) rtesult. But I will certainly do it some time in the future.
    Ananth
    : ]]

  • on September 14, 2013, 19:43 GMT

    How can we forget 6/23 by Ashish Nehra vs England in 2003 worldcup ? Was not that great spell? :(
    [[
    This article is not about great bowling spells.
    Ananth
    : ]]

  • mad_ush on September 14, 2013, 20:51 GMT

    Nice article. Stats bewildered me, but the summery took is more than enough to spark the intensity of those matches and to relieve those moments.

  • on September 14, 2013, 22:00 GMT

    awesome analysis dear... thumbs up...

  • on September 15, 2013, 6:10 GMT

    A great analysis. I fully agree with your selection of matches won batting second in situations where it was thought that the loss for that team was inevitable at some stage of the match. This analysis also prompted me to use statsguru to find such matches where partnerships of 9th-10th wicket (>=30 runs as criteria) almost carried a team to victory. i confine my analysis to ties only in this comment.

  • Bonehead_maz on September 15, 2013, 6:26 GMT

    lol - this is a bad day for you to bring this up Ananth ;) Yes - I actually watched it .... desperation slowly leading to exasperation ? Will comment on what I'm sure is another excellent article when I've read it. Meantime, the specials article on Thommo 1973 is a corker, and worth a read. I class Johnny Pym as a friend and never met a cricketer so enlivened with theories of how to better perform . LMAO he once batted against me with a bat full of holes so the air wouldn't resist it's downswing! Sandy Morgan a likewise "legend" to me ! Barry Knight - the supplier of my training facility and a great bloke, Greg Bush a great bloke and wonderful team mate. I played U 21 reps a year earlier than that story as a 14 yo opening batsman - needless to say Thommo destroyed any rofl "Sachin" in me ! Took me another 3 years to take the easy single to fine leg - needed another 10 years having no thigh pad to fully overcome. ( if I don't score off that I deserve to be hurt etc etc lol)

  • Vyasa_Shastry on September 15, 2013, 9:03 GMT

    Hello Anantha,

    Very nice step in this direction. I did not understand the column titled Balls in your table. How come some of them are 0? What does this zero represent? The others are balls remaining at the FOW scenario. I think I have missed something in this article. Besides this, I have a few comments to add:
    [[
    That means no balls data was available. If you re-rtead the article carefully I have explained this clearly.
    Ananth
    : ]]

    1) A point just for academic interest, but wouldn't it be more relevant to change the weightage for each wicket based on the last 5 years data rather than ODIs to date? It might be more accurate since batting strength might have increased in the lower order, no (considering that many of these matches have 6 down scenarios). What would be even better is to give the CTD batting average (batting average is also not bad) of the 2 batters at the crease to give a better perspective of how unlikely the scenario was in those specific 30 cases!
    [[
    Too much nit-picking and complicating will spoil the essential simplicity of the analysis. The numbers look very good and are applicable for all the 40 years. If the 9th wicket resource for the past years is 0.9487 instead of 0.9533, what does it mean.
    Ananth
    : ]]

    1 of 2

  • Vyasa_Shastry on September 15, 2013, 9:08 GMT

    2 of 2

    2) You rightly pointed out that wickets are more valuable that balls remaining in most scenarios, but there must be a way to combine the 2 difficulties (a simple sum or otherwise). Scoring 36 in 12 balls is more difficult lower down the order than say... at 2 down knowing that you have more batters to come after you. I liked the approach where you looked at difficulty of batting conditions in various test matches based on RPW and RPO and then tabulating the runs scored. Don't you feel it would be more appropriate to use the same philosophy here (how do you deal with those zeros?)?
    [[
    Where data is available I have considered the Balls-resource-factor and compared it with Wickets-resource-factor. Pl see the few examples given.
    Ananth
    : ]]