Teams October 7, 2010

Test series analysis - part 1

A comprehensive series-based analysis of results

West Indies: Most consecutive series undefeated © Getty Images

Couple of weeks back I did an analysis of the Test match results by periods. I am confident that the readers derived a lot of new insights into the performance of teams. I had also done a graphical analysis of the teams over the years. Amit Patel had suggested that the analysis be done by series. A simple but very sound suggestion since the series wins are the ultimate objective of any teams. Individual tests are only the means to this end. Hence I decided to do a comprehensive Series-based analysis.

Ha! Easier said than done. This turned out to be one of the most intriguing tasks I have ever undertaken, comparable to the Night-watchmen analysis. This is primarily because the Test Series has a myriad of variations and as I opened one door I came across intriguing possibilities in front of me. Unlike the Test match analysis this also offered a lot of insights beyond the base performance graphs. Let us now move on.

What is a series? There have been 150+ one-Test series, and moving upwards to 34 six-Test series. I have defined a Test series as any bilateral contest between two teams. Even the one-Test series has been considered as a Test series. The minimal nature of the contest has been taken care of in the point allocation. Until now there have been 605 series played during the past 133 years.

Out of these 605, three have been triangular tournaments. The first was held during 1912 between England, South Africa and Australia. There were 9 matches and no Final. I have treated this as 3 bi-lateral series. A 3-match England-Australia series with England as home team. A 3-match England-South Africa series with England as home team. Finally a 3-match South Africa-Australia series on neutral grounds. Then the individual Series dynamics take over.

The second was the Asian championships during 1998 involving Pakistan, India and Sri Lanka. There were three round-robin matches followed by a Final. The match between India and Pakistan in India has been treated as a one-Test series with India as home team. Two matches were then played in Sri Lanka. The matches between Sri Lanka and Pakistan/India have been treated as one-Test series with Sri Lanka as home team. Finally the Final between Pakistan and Sri Lanka in Pakistan has been treated as a one-Test series with Pakistan as home team.

The third was the Asian championships during 2001 involving Pakistan, Bangladesh and Sri Lanka. There were two round-robin matches followed by a Final. The match between Bangladesh and Pakistan in Bangladesh has been treated as a one-Test series with Bangladesh as home team. The match between Sri Lanka and Bangladesh has been treated as one-Test series with Sri Lanka as home team. Finally the Final between Pakistan and Sri Lanka in Pakistan, played after a gap of 6 months, has been treated as a one-Test series with Pakistan as home team.

Other than the 1912 triangular tournament referred to above, 4 Test series have been played on neutral locations. These are Pakistan vs West Indies at Sharjah during 2002, Pakistan vs Australia at Sri Lanka during 2002, Pakistan vs Australia at Sharjah during 2002 and recently, Pakistan vs Australia at England during 2010. These four series carry "neutral" tags for both teams concerned. Sharjah could be termed "home" for Pakistan from many points of view. However it is fair to designate that as neutral. Thankfully, even though Pakistan cricket has gone through troubling times periodically, the neutral venues have ensured that we do not lose out on watching one of the most exciting teams in world cricket.

Now for the allocation of points for series results. I have given below the basis for points allocation.

1. Scoring method: I will not adopt the 2-1-0 method which I adopted for the Test matches. That was acceptable there since those were only single Test matches. They might be part of a longer series. However the impact was confined to the specific 3+ days. On the other hand, a series is a much greater contest and could be conducted over 3 months. The Test matches were played over different grounds and weather conditions. It would be unfair to use a simple 2-1-0 method. After all a series can range from a 0-0 draw in a 1-Test series to 5-1 in a six-test series. Hence I have adopted a more complex method of allocating series points, described below.

2. Series wins: In general, a 5-x series win will get more points than a 4-x series win, a 4-x win more points than a 3-x win, a 3-x win more points than a 2-x win and a 2-x win more points than a 1-0 win. This certainly makes sense since this rewards the extent and quantum of win. the minimum points for a series win is 3.0 for a 1-0 win in a 1-Test series at home. The maximum points, theoretically since this has not yet happened, is 15.0 points for a 6-0 away win. It is also necessary to mention that there has never been a 6-0 series win. The table is shown below.

4. Series draws: I will allocate more points, for both teams, for a 2-2 draw than a 1-1 draw. Similarly a 1-1 draw will carry more points than a 0-0 draw. Obviously a 2-2 draw is possible only in a 4+ match Test series. This makes more sense since the score draws deserve higher consideration than score-less draws. And so for other draws. There has never been a 3-3 draw in Test history. The table is shown below.

3. Series losses: A series might be lost 0-6, 1-5, 2-4, 2-3, 1-2 et al. Not all these losses are the same. The teams which have fought hard to win at least one or two Tests deserve some consideration. Hence I have allocated some token points for Test match wins in losing series. In other words, a 3-0 win will give x points to the winning team and 0 point for the losing team. A 3-2 win will give the same x points to the winning team and y points to the losing team. So the differential points will be reduced for closer wins. The table is shown below.

5. Win points in series wins: The series wins for 1-Test rubbers will get the lowest points, for 2-test rubbers more, for 3-Test rubbers still higher and maximum for 4/5/6-test rubbers. Again understandable since it is more difficult to achieve the wins in longer rubbers.

6. Dead rubbers: There is a misconception regarding dead rubbers. People complain about dead rubbers only when the leading team takes it easy and loses. If a team is leading 3-0 and wins, now everyone appreciates the ruthlessness of the winning team and exhorts them to go for a clean sweep. Finally no team wants to lose a Test, whether they are leading 3-0 or trailing 0-3. Having said that, I have given lower weight for the dead rubber results. Some fine tuning still needs to be done for this.

7. Away bonus: Away results, wins and draws, will carry an additional weight of 25%. Less than a third of the series (31.1%) are won away. It is far more difficult to win a series away than a Test away. Hence this significant weight. Since the weight is applied on the points secured, short series away wins carry correspondingly lower points.

8. Neutral bonus: Neutral results, wins and draws, will carry an additional weight of 12.5% for both teams.

9. Team Strength adjustment: The relative strengths of the teams are finally applied to the series points secured. The factor varies from 125% (for Bangladesh performance against Australia in 2003 et al) to 75% (for Australia performance against Bangladesh in 2003 et al). About 10 series qualify for these extreme adjustments. Then the weaker teams start improving and the adjustment becomes 122%/78% and so on. The relative Team Strength indices are used to arrive at this factor.

10. Series Index: The total points secured for the concerned period is divided by the number of series to get an Index value which indicates the position of the team in the concerned period. An index value of 5.0 is indicative of a very good period for the team. An average value of nearing 6.0 indicates significant domination and nearer 7.0, possibly complete and total domination.

A. Points allocation for winning team for series wins

Max    1-win  2-wins  3-wins  4-wins  5-wins  6-wins (Not yet there)
1        3      -       -       -       -        -
2        4      5       -       -       -        -
3        5      6       7       -       -        -
4        6      7       8       9       -        -
5        6      7       8       9      10        -
6        6      7       8       9      10       12

B. Points allocation for both teams for series draws

Max     0-0     1-1    2-2
Tests  Draw    Draw   Draw

1 1.0 - - 2 1.5 2.0 - 3 1.5 2.0 - 4 2.0 2.5 3.0 5 2.0 2.5 3.0 6 2.0 2.5 3.0

C. Points allocation for losing teams in series losses

Max     Matches won by losing team
Tests  0-win   1-win   2-wins    (3 wins is not possible)
1       0        -       -
2       0        -       -
3       0      0.50      -
4       0      0.50      -
5       0      0.50     1.00
6       0      0.50     1.00
Before we go on to the tables and graphs, let me identify some outstanding team performances in the form of great streaks. This is a great by-product of this series analysis. And series streaks have a far greater value than Test streaks.

First two wonderful streaks have been identified and presented here. I find it difficult to distinguish between the two. Both are outstanding examples of team performances.

The first is the longest unbeaten streak of test series without losing. Let us trace the path.

West Indies lost the 3-test series to New Zealand by 1-0 (that too a 1-wkt loss).

Then, during the next 14 years, West Indies went on to play 29 series, 17 away, and remained unbeaten, yes, you read it correctly, unbeaten. They won 20 series and drew 9 series. And let us remember, no weak teams. This is the definition of domination, matched by only one team afterwards. Just for information, West Indies gathered 174.5 points during this streak and averaged 6.02 points per series.

The streak came to an end during 1994 when West Indies lost 1-2 to Australia at home.

Now for the other, equally mind-blowing streak.

Australia lost to Sri Lanka 0-1 during 1999 away.

Then, during the next 9 years, Australia went on to play 33 series, 12 away, and had 29 wins, 2 draws (New Zealand and India) and 2 losses. This is the alternate definition of domination, matched by only one team before. Just for information, Australia gathered 196.3 points during this streak and averaged 5.94 points per series. The two losses were the famous 2001 Indian win and the equally famous 2005 Ashes win, both by narrow 1-2 margins.

The streak came to an end during 2008 when Australia lost 0-2 to India away.

The most number of continuous series wins was by Australia during the period 2005-2008 when they had 9 consecutive wins. This streak was book-ended by the 2005 Ashes loss and the 0-2 away loss to India during 2008. The index for this streak was 6.29.

Australia had a 8-series streak of wins just before this one. England also had a 8-series streak way back in 1882. The best sequence for West Indies was a 7-series sequence of wins starting 1983.

Just for information, India has a sequence of 5 wins and 2 draws at the current point. With a win/draw against Australia, this will be extended to 8 series and possibly 9 when the New Zealanders come. But remember that these are not wins.

Since the article has already become quite a long one, I am going to keep the other tables and graphs to the follow-up article. Here I am only going to show the table for all tests combined.

Team       Total  Win Draw Lost Aw-W Aw-D Points Index

Australia 197 113 29 55 45 12 870.1 4.42 England 239 118 39 82 45 24 888.8 3.72 West Indies 123 54 21 48 24 12 443.0 3.60 South Africa 106 49 17 40 20 11 338.4 3.19 Pakistan 124 48 34 42 18 20 338.0 2.73 India 133 47 30 56 15 13 339.1 2.55 Sri Lanka 87 31 20 36 8 10 188.6 2.17 New Zealand 136 29 35 72 11 11 225.5 1.66 Zimbabwe 43 5 7 31 2 2 32.6 0.76 Bangladesh 35 2 0 33 1 0 9.6 0.27

No surprises here. Australia leads with 4.42 points, followed by England with 3.73 points. West Indies are in close third position with 3.60 points. South Africa and Pakistan complete the top-5. To cross off the final 't', the 1-test series between Australia and ICC has been considered as a home series for Australia. There is no contra-away series since ICC as a country does not exist.

I will show the period-wise tables, team tables and supporting graphs in the next part which will follow in a few days. All the tables will also be made available then.

For those of you who have started sharpening their key-board skills to say that these points are arbitrary and not objectively determined, I have a one-sentence answer. The same points, arbitrary and subjective they may be, are applied across all teams and all periods, over 1971 test matches.

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

  • testli5504537 on October 11, 2010, 9:19 GMT

    A more comprehensive statistical analysis Ive yet to read. Its understandable that ICC has to keep its rankings current, but its nice to get all the cricket ever played on one table and then locate your team in it. I have loved it. I am also happy to note that anti-Team-Pakistan factors both within and without the country still may have a year or two's work ahead of them to throw Pakistan out of the all-time top 5. That gives the "guardian angel" time to execute the miraculous rescue. And hopefully Pakistan won't immediately be thrown out of the cricketing map =)Much blessings for the likes of AH Kardar and Imran Khan.. [[ A name would be nice. Ananth: ]] .

  • testli5504537 on October 11, 2010, 4:57 GMT

    You should also make a slight adjustment for a 2 test series like a 0.8 or 0.75 factor in their counting.

    Another method according to me is to eliminate counting the number of series altogether. Just Divide the number of points scored by a team by the maximum it could have scored. This under the current table is 3, 5, 7, 9, 10, 12 for a 1, 2, 3, 4, 5, 6 test series. After doing this there is no need to give weights to the series.

    A country which has only played 1 or 2 test series is not penalised or favoured here because it is judged against the maximum points available to be scored. I have found this method to be more fair than the weight of 0.5 suggested by me before.

    Another thing I like about using this method is that we assign points only once to the series played. After then its all objective. In weighting there are 2 levels of subjectivity in the points allocation and weighting which may be questioned. In fact the weighting has already been questioned by a reader. [[ Raghav That is an excellent sugegstion. Also remember that that is what I did in the Test matches analysis and I feel there was more acceptance. What I will do is to do both and let us see and compare the results. After all we are it is a free world. Ananth: ]]

  • testli5504537 on October 11, 2010, 3:50 GMT

    Hey Ananth, great analysis, truly an interesting read.Just one concern. How were the test matches that were abandoned because of a bad pitch handled? There was one last year between West Indies and England and there was also one some time ago between India and Sri Lanka. Since there was hardly any play in those matches, should the weaker team actually get credit for drawing those? [[ Gaurav Why split hairs over a trivial factor. Once the toss has been done the match is a test match. Ananth: ]]

    Overall, a job well done. Looking forward to part 2.

  • testli5504537 on October 10, 2010, 17:33 GMT

    With regard to response to Rachit, your adjustment will still not solve the problem as someone may say why .5 and not .33 (since most of the test series have been 3 game, not 2 game series). Basically you are trying to compare apples with oranges in a sense that different length of the series can not all be put together and given subjective markings as pointed by others. The only fair way to do this is to first separate series based on number of tests it contained and look at home and away record in different time frames. In statistics there is something called 'Selection Bias' and 'Confounders' that skew the analysis. The one Selection Bias (by bias I don't ever mean you are biased, I mean only in the strictest statistical term so please spare the outrage) is evident, the teams that played more 5 test series are better off than the ones who played 3 test series. Also Australia and England early on played mainly at home and to each other so that creates a different set of problems. [[ Rohit Your points are well made. The problems with 3 teams for 50 years, entrance of weak teams, period of stability, again entry of weak team is an omni-present problem. Kinly wait for Part-2. You need not look at the 133 year tables. I will be presenting the period wise separation as well as the progress of teams over the periods. That will offer much better insights If it would help I could do a splt pre-ww2 and post-ww2 also. At least 7 teams were there from 1952.. Ananth: ]]

  • testli5504537 on October 10, 2010, 11:59 GMT

    Awesome analysis ... but there is one small doubt i have ... u give 3 points for a 1-0 win in a 1 match series and 6 for a 1-0 win in a 5 match series ... while this is completely fair as a 5 match series is tough to win, teams like Sri Lanka will be at a disadvantage since they have played a lot of 1 match series ... and teams like australia and england will have an (if i can say that) unfair advantage ... so a lot of 1-0 results will get sri lanka 3 points while australia or england will get 6 points for the same result .. this will definitely mean that sri lanka will not be able to compete wrt the average index ... for example, indiaq and sri lanka have a series win/sweris played ratio of .35 .... but the indexes are 2.55 and 2.17 .. while u may say that india has won series more handsomely than sri lanka, i am sure that a lot of this difference can be explained by the fact that sri lanka played a majority of there series as 1 match series .. or max 2 matches ... [[ Rachit Pl see my comment on the changes I am going to effect. I will now consider a 1-test series as half a series for the purpose of index calculation. Also the points band will be narrower. This should certainly address a number of points you have raised. Ananth: ]]

  • testli5504537 on October 10, 2010, 9:15 GMT

    Ananth, Shamaz Majid Sure. Just hoping to clear some of misconceptions:….Shamaz, 1) The question of hating pakistan etc does not exist. It is sad that even educated folk like you have fallen prey to such invidious propaganda. There is simply a concern because of certain activities widely perceived to be emanating out of pakistan- and not only by indians, that is all. Otherwise most indians would not even bother about pakistan, there are much better and constructive things to do- Trust me on this.

    2) I actually think it runs a bit the other way round, in fact. For eg. Wasim Akram is widely respected in india and you will hardly ever hear an adverse comment about him by indian cricket fans. However, I can guarantee you that every single blog or forum which contains something about tendulkar will have a virulent anti tendulkar brigade comprising 99% pakikstanis. Point to ponder. In any case, as Ananth says, it is futile going to and fro about this. Back to the cricket! [[ Abhi As one who has felt the heat from both sides a lot, let me assure you that there are any number of fanatics on both sides. Over the past two years I have been quite ruthless in handling such comments. Even now I have used the direct mailid (I did not want to expose my mailid) to warn both types of readers that I could easily blacklist such readers from the entire blog. I resorted to this only for hate mails. When Umar Akmal fails, I am indeed sad that such outstanding talent is not flowering. There is lack of mentoring. Younis Khan should be doing that. But where is he. I think Indian cricket has a stability unfortunately lacking in Pakistani cricket. And the reason is probably a miniscule part of the populace. I fervently hope that somone cleans up the Palistani "Aegean stables". IPL can afford lose more franchises than Punjab and Jaipur. For that matter world cricket can afford to lose IPL. But world cricket cannot afford to lose Pakistani cricketers. Ananth: ]]

  • testli5504537 on October 10, 2010, 5:43 GMT

    Important note: Based on the reader responses, I will tweak the analysis in the following manner and come out with the final ratings, tables and graphs in the follow-up article. This is expected to be published no later than Friday, 15 October. 1. 1-test series will be taken as 0.5 series for purposes of determining the Index points. 2. There will be a tighter band of points, ranging from 2 to 8, as compared to the current values of 3 to 12. 3. A 3-1 win will carry lesser points than 3-0, a 2-1 win will carry lesser points than 2-0 and so on. 4. Higher credit will be given for coming from behind. However only if a team wins 2 remaining matches to either draw or win a series. This will not be applicable for 1-test comeback situations.

  • testli5504537 on October 10, 2010, 3:25 GMT

    I completely agree with Anantha on the rating based on strength of the opposition. Treating everyone exactly the same will create huge discrepancy and would be a step in the wrong direction.

  • testli5504537 on October 10, 2010, 1:42 GMT

    (Contd). Similarly, the Indian team that beat England in 2007 was probably depleted due to Sehwag's absence (hence a lower team strength) but Sehwag was in the worst patch of his career and his presence may have been worse for India under the circumstances. Probably such occurences are few and far between. But was just curious to know if they will change anything. My apologies iof I completely mis understood the definition of team strength.

    Also am looking forward to see a similar analysis in ODIs. Here my guess is the changes of the best team will be more dynamic. [[ Anand I have two Tesm Strength metrics. One based on Career-to-date figures and another one on Career figures. For this macro level analysis, I have used the career lbase4d numbers. This has the advantage of a settled number for players already retired. It also takes the temporary lapses in form in its stride. Any blips in players' career would be smoothed over. However the absence of a player is something else. The absence of McGrath and Warne would hit the team, to varying degrees. I have no idea how much we are going to feel the absence of Laxman in this test match. He is only a very good batsman overall but a giant against Australians. However his career numbers would only be used for determining the Team strength. India would have dropped significantly from Mohalli to Bangalore and that is also true in real life. Ananth: ]]

  • testli5504537 on October 10, 2010, 1:25 GMT

    Ananth: A thought occured to me. In your team strength factor, how would it change things if "current form" is taken into account? By current form, I mean the players' current form. Of course raises a flag on how current is current. But if we put some reasonable number to that, say the last 10 innings or so.. Or could be the player's form against a particular opponent. This may not yield the correct comparison of teams because I understand that even if Pointing scores signel digits in his next ten innings, a Australian team without him is a depleted one. The reason for this thought is there is a talk about Australi being depleted in 2003-2004 against India because of the absence of McGrath and Warne. While I agree that McGrath's absence was a handicap, give Warne's record against India am not sure how much of a factor his absence was. (Continued in the next message)

  • No featured comments at the moment.