An over in a Test match is normally a small incremental step towards building a long innings or planning a dismissal. Seemingly nothing might move for many overs but these are part of a long-term (relatively speaking) plan to achieve progress. Considering both batting and bowling aspects of the game an over is but 0.25% (1 in 400) of a Test match.

A T20 over is a totally different entity. It forms a huge 2.5% (1 in 40) component of a T20 match. An over lost is a big step backwards and couple of overs lost would invariably lead to defeat. I would venture to say that a ball in a T20 match would be approximately equivalent in importance to an over in Test cricket. The planning is almost down to each ball. A dot ball is a resounding success, especially during the late stages, and a four conceded might lead to wild celebrations where a six was needed.

Milind, may the force be with his tribe, has created the ball-by-ball data for almost all 400 T20 international matches. It is complete and is a jewel in my database. We have validated the data together and have in hand now a collection of gold dust. This is especially true for T20s since the T20 analysis is a lot more well-defined and the data lends itself to multiple shades of nuanced analysis. This is the first in a series of such analyses. The unit of analysis is the T20 Over.

Redefinition of the Dot Ball

I have made a very significant and common-sense-based re-definition of one of the pillars of bowling analysis. Henceforth I will treat a dot ball as one in which no run was added to the opposing team. Thus a maiden over comprises of six such tougher-defined dot balls. I am sure most readers will agree with me. The current definition of dot ball and maiden over, which dates back to 1877, is outdated and archaic for the fast-paced T20 game.

A bowler should earn his maiden today. Already we have amendments to the law that do not allow wides and no-balls to be exempt while looking at dot balls and maiden overs. I have simply extended this concept to byes and leg byes. When Dale Steyn bowled six balls to Virender Sehwag and Gautam Gambhir in the league game in the 2012 World T20 in Colombo and conceded five leg byes, he, wonderful bowler though he is, did not deserve a maiden. There is no denying that five runs were accrued to the Indian total, which is all that matters. Interesting sidebar is that India won by a single run. Similarly, Shaun Tait bowled a wonderful first over to Imrul Keyes in a league game in the 2010 World T20. It is classified as a maiden. But he conceded four byes and that was not a dot ball.

One last point to support my definition. A bowler agrees with the wicketkeeper that the third ball will be a googly. The keeper moves down leg side anticipating this. The bowler forgets and bowls a vicious legbreak. There's a good chance it will go for four byes. Whose fault is this? Just a simple example. The bottom line, as far as I am concerned, is that it is the runs conceded to the other team that matter, not the runs conceded to the batsman. This is to emphasise the team game concept.

I have talked about this in depth since I know that the point will be raised by readers. I will accept and post all such comments but will not change my interpretation. The bowling analysis has already been changed to reflect this. I have 151 maidens and the other scorecards, including ESPNcricinfo, have 192. Of these, 41 overs have had one to five byes/leg byes conceded and thus have been classified by me as non-maiden overs. Good change: makes the maiden that much more treasured.

A look at the most extraordinary overs: Batsman-dominant

Stuart Broad to Yuvraj Singh, 2007: Everyone, and their neighbour's dog, knows about these five minutes of madness. Broad bowled six balls, mostly of good length, and was sent over the fence six times by Yuvraj Singh, then at the peak of his form. Four of these were on the on side and two over the off-side ropes. I have always felt that Broad could have bowled a deliberate wide, if nothing else, at least to break the rhythm of Yuvraj. But then history would not have been made in a very symmetric fashion.

Wayne Parnell to Jos Buttler, 2012: The sequence of events, fairly self-explanatory, ran thus: 6, 6, 2, nb+0, nb+4, 4, 6 and 2. A disjointed sequence, but England moved from 74 to 106 in this 11-over match, and won comfortably.

Izatullah Dawlatzai vs England, 2012: The sequence was Buttler 4, Buttler lbw, Jonny Bairstow nb+6, nb+1, Luke Wright 6, 6, 6 and 1. England, not exactly moving the world at 155 for 3 in 18, suddenly propelled to 187 for 5 at the end of the 19th over.

Daryl Tuffey to Ricky Ponting, 2005: This was the very first T20I match played, and Ponting took Tuffey to the cleaners to the tune of 6, 2, 6, 6, 4 and 6. Tuffey never really recovered from this mauling.

Bilawal Bhatti vs Australia, 2014: This was in the recent World T20. Aaron Finch scored 4 and 1 and then Glenn Maxwell took over with 4, 6, 6, nb+4 and 4. The irony was that this was in the eighth over. Australia moved from 72 for 2 to 102 for 2 in eight overs. They needed 90 runs in 12 overs but messed up the chase and lost. Bhatti redeemed himself when he was entrusted with the last over. He conceded only six runs and captured two wickets.

A look at the most extraordinary overs: Bowler-dominant - through wickets

Mohammad Amir vs Australia, 2010: I consider this to be the most incredible over in T20I history. Do I hear Yuvraj's 36 against Broad? Excellent credentials indeed. But I feel a move from 191 for 5 to 191 all out during the course of a single over gets the biscuit. I will always take the side of the bowlers anyhow. And the opponents were Australia and established batsmen were at the crease. The sequence was Brad Haddin caught, Mitchell Johnson bowled, Michael Hussey run out, Steven Smith run out, Tait dot ball and Tait bowled. Five dismissals, three wickets to the bowler, six dot balls, this was the Twilight Zone.

RP Singh vs New Zealand, 2007: The sequence was Daniel Vettori bowled, Shane Bond 4, Bond run out, Craig McMillan 1, McMillan run out and Jeetan Patel run out. New Zealand moved from 185 for 6 to 190 all out. But still won the match.

Doug Bracewell vs Zimbabwe, 2011: Forster Mutizwa 1, Mutizwa run out, Ray Price run out, Kyle Jarvis caught and Chris Mpofu caught (5 balls). Bracewell hastened the end of Zimbabwe innings getting rid of four batsmen in five balls.

Haseeb Amjad vs Nepal, 2014: Malla 6, Malla 2, Malla c&b, Vesawkar run out, 1 bye and Budayair run out.

Al-Amin Hossain vs West Indies, 2014: Marlon Samuels caught, Andre Russell caught, 1 bye, Dwayne Bravo caught, 1 bye, Denesh Ramdin run out.

Now let us have a look at the table containing key indices for the 20 overs. This is a massive table containing around ten key indices for each over. Detailed interpretation of this table could run to pages. So a brief commentary is provided.

First, an explanation of the number of matches covered and the base numbers. Out of 400 matches played so far, two matches (26 and 68) were abandoned after the toss. Two matches (9 and 335) do not have ball-by-ball data available. In two matches (273 & 318), no second innings was played. So we have ball-by-ball data for 396 matches and 790 innings. Out of these 790, two innings (119 and 318) lasted four and two balls respectively. So the number of first overs is 789. The rest follow based on this.

T20 Over indices: By over
Over#Overs20+ Runs10+ Runs4 Runs or lessMaidensDot Balls/OverRpOBpWWktsMax Runs
1 789 0.25%16.85%37.50% 3.17% 3.43 6.0824.7 192 24
2 788 0.89%25.12%35.02% 2.54% 3.19 6.8322.6 209 25
3 787 1.02%30.63%25.80% 1.02% 2.92 7.6121.5 220 25
4 786 1.02%31.53%22.63% 1.78% 2.78 7.7720.1 235 23
5 785 1.91%31.59%26.24% 1.40% 2.72 7.8922.3 211 27
6 782 0.90%31.97%26.98% 2.43% 2.73 7.7020.4 230 22
7 779 0.39%19.52%37.38% 0.77% 2.33 6.4123.5 199 22
8 773 0.91%19.41%33.25% 0.78% 2.27 6.7123.4 198 30
9 769 0.26%18.86%31.09% 0.91% 2.23 6.6822.6 204 24
10 762 0.79%22.71%29.27% 0.13% 2.07 7.0123.1 198 32
11 758 0.92%22.55%29.41% 0.79% 2.11 6.9718.9 241 25
12 752 0.93%25.78%27.11% 0.53% 2.01 7.4023.6 191 28
13 747 1.61%28.93%26.65% 0.27% 1.98 7.5919.5 230 25
14 739 0.54%26.52%25.30% 0.68% 1.99 7.4818.7 237 21
15 726 1.65%28.91%26.43% 0.69% 2.05 7.7315.8 276 26
16 708 2.40%32.05%26.12% 0.56% 1.89 8.0214.9 286 28
17 686 2.48%39.23%18.67% 0.58% 1.82 8.7413.8 299 24
18 658 2.74%36.49%23.26% 0.30% 1.83 8.5111.7 337 27
19 605 2.98%44.63%21.32% 0.17% 1.79 9.37 9.9 365 36
20 519 3.27%43.70%17.52% 0.38% 1.63 9.92 6.8 456 29
All14699 1.32%28.29%27.70% 1.03% 2.32 7.5517.65014 36

The number of overs has already been explained. One additional point: 19.4 and 18.4 overs will add to 38.2, not 39. In other words the overs are converted to balls, added and then converted back. This is done to get an accurate measure when doing the percentage calculations and per over values.

Only two of the first overs went for above 20 runs while as many as 17 of the last overs went for above 20 runs. As expected, the 20th over leads the table in this regard. The total number of such overs is 194. In other words, an average of one 20-plus over every two matches. The pattern is a slow increase, with one exception. Look at the drop in the ninth over. From seven occurrences in the eighth over, the figure drops to two in the ninth over and then jumps to six in the 10th over. And look at the jump in the 13th over and abrupt drop in the 14th over.

The ten-plus runs numbers follow a similar increasing pattern to the 20-plus-run overs except that the kinks in the curve as seen in the ninth, 13th and 14th overs do not exist. The abrupt drop in the seventh over is clearly a result of the removal of field restrictions. Look at the last few overs. Nearly half the overs are ten-plus run overs. Some people might even argue that they expect more.

The four-and-below run overs represent the tight overs. As expected the first over leads the field. Three out of eight first overs have had only four or fewer runs taken off. Let me remind the readers that these refer to team runs: that means including all extras. That the seventh over comes close to the first one should not surprise anyone. But there is a surprise in the 17th over. Why is there a big drop from the 16th? Are the teams which have lost around five wickets or so trying to play safe to ensure that they can go on an all-out attack mode in the last three overs.

I will cover fascinating topic of maidens in depth in the next part of the article. Here I will refer to the overall percentage figures. As expected, maidens occur most frequently in the first over: just short of one every three innings. This drops very drastically to 2.5% in the second over and then to 1% in the third over. Goes upto 2.4% in the sixth over and then a huge drop to 0.77% in the seventh over. The figures keeps on dropping to the lowest at 0.17% in the 19th over and then to 0.38% in the last over. The last figure is misleading. It only indicates two maidens, as against one, the previous over. The overall percentage value is 1%, meaning one every five innings. But it is a rather steady decline, overall.

Now we come to Dot Balls per over. The overall figure is 2.32. But this starts with a fairly high 3.43 in the first over through to 1.73 in the last over. It is interesting to note that there is no great variation in this measure. It is possible that in the early stages the non-dot-balls are ones and twos while in the later stages these are boundaries.

We now come to the most important measure here: the average Runs per Over. The average across 400 matches is 7.55. The first over is a fairly low 6.1 and quickly reaches the average in the third over. Then this value drops off drastically in the seventh over and takes a further seven overs to pick up. The last two overs are above nine and the last over is fast approaching ten. However this figure is only over 500 overs as against the 6.1 over 789 overs. There are no major surprises here.

The number of wickets, which fell in the specific overs, is less relevant than the balls per wicket figure since the number of overs figure varies a lot. This figure, surprisingly, has a choppy ride. It starts off with an understandable 24-plus (once in four first overs) for first over to 20 for the sixth over. Then there is a sharp increase in the seventh over as caution takes over and the fielders are banished to the outfields. Even within the first group, see the sudden increase in over No. 5. Afterwards the value drops steadily from over No. 7 to 10. Then there is a completely inexplicable sudden drop in over No. 11 and a sudden increase in over No. 12. From over 15 onwards the number drops abruptly to a value of around 6.8 for over No. 20: a wicket in the last over of every match. Quite a performance indeed.

The final column lists maximum runs conceded. I have already written about the 36-run over and the 30-plus run overs. The 32-run over was the 11th of the innings and the 36-run over was the 19th. Most other overs have the 20s as their highest run count. For reasons not clear the 14th over has the lowest high score: a mere 21. Possibly the lull before the storm in the 16-20 overs.

The graph below is self-explanatory.

It is clear that the Dot balls per over graph follows a steady downward trend with a slight variation at the six-to-seven over mark. The range is in a narrow band of 3.4 to 1.6. The Runs per over graph is a more volatile one with clear pronounced drops at over no 6 and 18. But the trend is reasonably similar to the Dot balls per over graph.

However the Balls per wicket graph moves like a yo-yo for no ostensible reason, until over No. 12. Afterwards there is a continued drop until over No. 20. It is virtually impossible for me to describe the movements between overs three and 12. And let us not forget that nearly 400 matches have been played across venues and during ten years. So we have data for either side of 700 overs.

Finally a few interesting extracts from the huge T20-Over Matrix file I have uploaded are given below. You could download this file and extract more of these gems.

Innings in which there were three overs with 20+ runs.
Match 202: Sri Lanka vs Australia
Match 268: England vs Afghanistan
Match 328: Australia vs England
Match 377: Netherlands vs Ireland

Innings in which there were 14/15 overs with ten-plus runs.
Match 13: Australia vs England. 15 overs
Match 127: Sri Lanka vs India. 14 overs

Innings in which there were 15 overs with 4 runs or less.
Match 67: Bermuda vs Canada. 15 overs
Match 27: Kenya vs Sri Lanka. 15 overs
Match 64: Ireland vs Kenya 15 overs + 13 overs for Kenya. Total 28 overs

I have created a veritable treasure house of information in the form of an Excel sheet. This is a 796 by 26 worksheet that contains the runs conceded by over and the 4/5 derived values for each over in each innings. The analytically minded can download the Excel sheet and really go to town. To download/view this Excel sheet, please CLICK HERE.

In the next part of the article I will have a comprehensive look at the three over-groups. The start phase (1-6 overs), consolidation phase (7-15) and the finish phase (16-20). I will also look at the fascinating subject of maiden overs: my new definition, not the pseudo one currently under force.

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