Pappus' plane - cricket stats

Over at The Bat and Ball Brimborion, Andrew Samson had a look at the best partnership breakers in Test cricket, ranked by the average size of the partnership broken by each bowler. With a qualification of 20 wickets, Norman Yardley was the clear leader, with an average partnership broken of 87,29, well clear of Viv Richards at 67,69. (Samson's numbers are slightly different from mine. I've manually checked Yardley's and my figure agrees with the Cricinfo scorecards.)

In comments, zscore noted that bowlers with few wickets per Test (such as all the players in Samson's top five) were likely to have high average partnerships broken, since part-timers don't bowl often when wickets are falling regularly.

You can take this idea a bit further: part-timers are also likely to have a higher average than specialist bowlers. We would expect that the higher the bowling average, the higher the average partnership broken. If you plot a graph of this for all bowlers with at least 20 Test wickets, you see a strikingly linear relationship.



A quality we would like in a partnership breaker is not just breaking large partnerships, but doing so without conceding many runs. A new method of ranking presents itself: the vertical distance above the regression line in that graph. If you're above the line, it means that the partnerships you break are larger compared to bowlers with your bowling average.

The huge and very obvious outlier is Norman Yardley, so he still comes in at number one.

name              matches wkts  avg     avg part  above reg line
Yardley,Norman    20      21    33,67   87,29     55,36
Sarwan,Ramnaresh  67      23    46,74   63,70     22,63
Gordon,Norman     5       20    40,35   59,00     22,40
McMillan,Craig    55      28    44,89   60,96     21,19
Stuart,Colin      6       20    31,40   51,15     20,81
Smith,Collie      26      48    33,85   51,67     19,61
Harris,Paul       8       24    24,50   44,33     18,81
Hick,Graeme       65      23    56,78   66,26     18,18
Burtt,Tom         10      33    35,45   51,03     17,85
Francis,George    10      23    33,17   49,13     17,55
Astle,Nathan      81      51    42,02   54,59     16,82
----
Nadkarni,Bapu     41      88    29,08   44,72     16,00
Vincent,Cyril     25      84    31,32   45,46     15,18
Price,Ray         18      69    35,87   47,84     14,37
Waugh,Mark        128     59    41,17   51,36     14,19
Mann,Tufty        19      58    33,10   45,47     13,93
Cook,Nick         15      52    32,48   44,52     13,42
Ratnayeke,Ravi    22      56    35,21   45,68     12,67
Gayle,Chris       68      59    39,59   46,37     10,30
Patel,Dipak       37      75    42,05   47,97     10,19
Kaneria,Danish    51      220   33,90   41,73     9,64

The rows after the "----" are the continuation of the table with a qualification of 50 wickets.

The figures include all Tests up to Test 1853, and not the just-completed Boxing Day Tests. In particular, Paul Harris has dropped down a lot, as he took a bunch of wickets in the Windies' second innings collapse.

Of course we also want to know the other end of the table — who has a really bad average and only breaks small partnerships? Anti-leading the way is the West Indian orthodox bowler Inshan Ali from the 1970's, whose 34 wickets came at 47,68 runs apiece, but whose average partnership broken was just 21,26. He's the only bowler over 20 runs below the regression line. With a qualification of 50 wickets, the worst is Rumesh Ratnayake: an average of 35,11 and an average partnership broken 22,05, putting him 10,88 runs below the regression line.

Lastly, a graph showing the average partnership broken against wickets per Test:



The outlier with 4 wickets per Test and an average partnership broken of 59 is Norman Gordon, who played five Tests just before World War II.