A number of readers had asked me to do a complete analysis of the Test team strengths. I had done some work on this earlier. However there is a need to throw out that lot and completely do this from scratch since the following related needs have been expressed at various times.
1. Completely integrate the career-to-date (c-t-d) values into the Team strength analysis.
2. Build in period based adjustments.
3. Allow no dilution into the process, especially the Bowling strength determination where a fifth weak bowler might completely distort the index values.
4. Give some weight for Bowling strike rates since these are Test matches.
1. Use only c-t-d values. Make adjustments during the early phase of the player's career. Essential for players like Mike Hussey, Brett Lee et al, whose first third of career was way better than the next two-third. I have been quite tough in this regard. During the first 50 innings or until 100 wickets have been captured, I have capped the c-t-d values at the career average, if it goes higher. Perfect example is Hussey. He had an average of 86.33 at the end of his 30th innings. But this has been capped at 52.50, which is his career average. Brett Lee had captured his first 50 wickets at 22.82. This has been increased to 30.82, which is his career average. I know it is quite tough on these players. However this has ensured that there are no spikes.
2. Determine the best 7 batsmen and use these batsmen figures to determine the Batting strength. This is to take care of night-watchman situations and genuine cases where the no.8 batsman in the batting order is better than the no.7 batsmen. The lower four batsmen are thus excluded. They might turn to be useful but do not really add to the strength.
3. After a lot deliberation I decided to do the Bowling strength determination with the best four bowlers only and not bring into consideration the fifth bowler. Traditionally most strong teams have had 6 batsmen, a keeper and 4 bowlers. The fifth bowler only provided additional support but the team's bowling success really depended on their top four bowlers.
4. There is no separate weight for all-rounders since the top all-rounders would find their place into either the top-7 batting or top-4 bowling or both. If Imran Khan bats at no.7, he brings to the table a 37+ batting average and sub-23 bowling average. That is his strength and will be reflected in the team strength index.
5. Do a period-wise adjustment. This is the one area where I have done something radically different and a complete change to the existing process. Until now I had done adjustment based on the adjustment factor for the decade (or period) in which the Test was played in. I was aware that this had the following major shortcomings.
- The adjustment is done based on the decade/period the Test was played in. However the player could have played during that decade, the previous and in some cased the previous one. So the adjustments are not perfect.
- The adjustment factor is the same whether the batsman is Tendulkar (22 year-career), Dravid (16 years), Yuvraj Singh (8 years) or Raina (1 year). Not exactly good and has to be improved significantly.
As I sat for hours on end watching the WC simulations rolling by in the desktop, I kept on fiddling with ideas and then one day I had a spark. I kept looking at Peer comparisons and then suddenly discovered that I had the solution staring at me. Why not adjust each batsman's career-to-date values dynamically and independently, with his own peer values. It was a natural process to zero down to an adjustment based on the Peer value for the batsman himself, in other words, from his debut test to the current test.
Easier said than done. However I set about creating database segments containing data values, such as batting average and bowling average for all teams, for each batsman, for each test he played in. Needless to say, the Batting averages were only for the best 7 batsmen in each test. Also the adjustment for each player will kick in only after he has played 10 matches since there is insufficient data in the early stages. The adjustment is done by determining the ratio of 35.99 (the all-time average of the best 7 batsmen in each innings for 1989 tests) to the concerned batsman's C-t-d Peer average. If the C-t-d Peer average is higher than 35.99 then it has been a batsman-friendly x-tests era and the ratio would be below 1.000. If the C-t-d Peer average is below 35.99 then it has been a bowler-friendly x-tests era and the ratio would be above 1.000. A similar working for the bowlers, for whom the all-time bowling average is 31.76.
It was impossible to split the, already complex, players's c-t-d peer bowling average into Pace and Spin. Hence I have done this based on a composite bowling average and done the bowling type adjustment at a later stage. The spinners have their bowling averaged lowered by a fixed factor.
It has worked beautifully. This allows for the changes which take place during a player's career. If there was a glut of runs during a phase of 2/3 years, it would be reflected instantly.
6. In view of the importance of Bowling strike rates in Test cricket I have computed the Team strike rates for the four best bowlers separately and multiplied the Team Bowling index by a pro-rated value based on this.
7. The adjustments are done separately for Pace bowlers and Spinners.
The last but very important point. After hours, nay days of struggling to make an equitable adjustment and exasperation, I decided to bite the bullet and exclude 64 Test matches played before 1900. The problem was mainly with bowlers. The 1800s were downright crazy. 10 bowlers, who had captured 50+ Test wickets, had averages below 20 and this distorted everything else. However it must be mentioned that very relevant players such as Clem Hill (1896), Trumper (1899) and S.F.Barnes (1901) are included. The only serious players we would miss out are Lohmann, S.E.Gregory and W.G.Grace. Lohmann, with his 100+ at 10+ single-handedly wrecked all analysis. Based on numbers, Lohmann was the greatest bowler ever, by a few miles, may his soul rest in peace.
Finally a note on the tables. Teams like the 1945 Australians, 2005 Australians, 1990 West Indies would have multiple entries in the table since quite a few of these teams were quite strong. Now that I would be using Career-todate values there would be changes from match to match even if the eleven remains the same. Hence I have extracted one representative and best team amongst this group and presented here a unique team table. This means even if there are 25 Australian teams of 2005-06 era, having almost the same team combinations, I will select one amongst these 25. However the team selected will be a real life team from a played Test. In other words there would be only one 1948 Australian team, one 2005 Australia team and so on. At the same time if two West Indian teams had radically different bowling line-ups, say 1980 and 1990, both have been included. Of course the complete table contains all the entries and can be downloaded.
While selecting the teams out of this collection of teams, I have followed the common-sense based principle that two bowling teams which have two of the four bowlers changed and the batting team which had 3 of the batsmen (out of seven) changed, will be considered different teams. The selection had to be manually done by me. While I have tried to be careful, it is not certain that I have included all teams qualifying. If readers note any misses, they are requested to inform me so I could include the same. I had also to do quite a bit of cutting and pasting. Hence there might be minor errors.
I have used two further criteria in selecting these teams. One is that the selected team should be in the top-100 in the concerned table. The other is that there should be a rough correlation to the population of teams in the top-100 while looking for as much representation as possible.
Let us now look at the tables. First the top-10 Batting teams of all time.
1. Australia: 49.81 MtId: 1661-2003 CtdAvg PeerAvg Adj FinalAvge Gilchrist A.C 58.80 36.92 0.975 57.31 Hayden M.L 52.01 35.86 1.004 52.20 Waugh S.R 51.07 35.90 1.002 51.19 Ponting R.T 51.12 36.01 0.999 51.09 Martyn D.R 46.38 35.89 1.003 46.51 Langer J.L 45.86 35.90 1.002 45.97 Lehmann D.S 44.95 36.39 0.989 44.45
Now comes the mighty Australians in Bradman's farewell test during 1948. Let us not forget that they had Lindwall, with his 2 centuries yet to come, at no.8. This, despite Bradman's 101.39, his average at the beginning of the test, being reduced by 6%. England, of vintage 1928, with Hobbs, Sutcliffe and Hammond, clocks in next. A slightly different Australians of 1934 are the next team. As compared to the 1948 team, this team had only Bradman.
In sixth place is the 2010 Indian team. As in the first team, this team had Dhoni at no.7, the statement which defines the batting strength admirably. The West Indian team of 1963, with Sobers, Kanhai, Butcher and Solomon, is in seventh place. In eigth place is the recent Australian team, with Hussey, Clarke and Symonds.
The table is rounded off by two West Indian teams of different ages. The 1955 West Indian team, with Sobers and the three W's is in ninth place. The table is rounded off by the Richards-led West Indies team of 1984. Not one of these 10 teams is out of place and almost all top batsmen of the world, barring Gavaskar, Kallis, Chappell, May, Compton, are represented.
Now for the best bowling teams.
1. West Indies: 49.89 MtId: 1158-1990 Marshall M.D 20.72 31.97 0.993 20.58 46.8 Bishop I.R 24.29 34.49 0.921 22.37 52.3 Ambrose C.E.L 23.97 33.11 0.959 22.99 54.6 Walsh C.A 23.91 32.84 0.967 23.13 57.9
Then comes the first of two English bowlins attack in this list, the 1956 foursome of Tyson, Laker, Wardle and Statham. Just behind them is the wholly different Caribbean pace quartet of Garner, Croft, Roberts and Holding, of vintage 1981.
The Australian batsmen have dominated the tables. However they also had top class attacks. The modern Australian attack is the one during 2005 and had McGrath, Warne, Gillespie and MacGill. Two attacking fast bowlers, a world class spinner and an excellent medium pacer complete the second English team of Botham, Underwood, Willis and Hendrick.
The fearsome South African pace attack of Pollock, Steyn, Ntini and Nel comes in next. Now the most balanced attack in this list, each a giant bowler, of Waqar, Imran, Akram and Qadir. The top-10 list is rounded off by the classical Australian attack of 1975, comparing of Lillee, Mallet, Walker and Thomson.
The Indian attacks miss out since they never had four really world class bowlers together. Even with adjustments, spin-dominated attacks, with averages between 25 and 30 are unlikely to fare well. For the record, the best Indian bowling attack was the one which played Test# 1782 (during 2006), with 42.39 points. The attack comprised of Kumble, Zaheer, Harbhajan and R.P.Singh. Same applies to New Zealand and Sri Lanka, Hadlee and Muralitharan notwithstanding.
Now the top-10 teams of all time.
MtId: 1744-2005 Australia : 95.34 (48.28+47.06) Langer J.L Hayden M.L Ponting R.T Martyn D.R Clarke M.J Gillespie J.N Katich S.M Gilchrist A.C Warne S.K Kasprowicz M.S McGrath G.D
For the record, the strongest teams from the other countries are given below.
South Africa Test# 1860 (2008) against West Indies, with 86.24 points. India Test# 1782 (2006) against Pakistan, with 84.96 points. Pakistan Test# 1443 (1999) against India, with 84.92 points. Sri Lanka Test# 1691 (2004) against Australia, with 83.28 points. New Zealand Test# 1700 (2004) against England, with 78.14 points. Zimbabwe Test# 1511 (2000) against New Zealand, with 66.64 points. Bangladesh Test# 1905 (2009) against Sri Lanka, with 54.60 points.
I would like to inform the readers that I will be taking a month off to handle range of commitments I have during the World Cup. As things stand, I will be back after the completion of World Cup.
Anantha Narayanan has written for ESPNcricinfo and CastrolCricket and worked with a number of companies on their cricket performance ratings-related systems