Cricinfo - Blogs - It Figures - Comparing the two halves of players' careers

The usual criteria apply. This is just to ensure that the career is sufficiently long. I have taken 4000 runs and 45 Tests as the cut-off for batsmen and 150 wickets and 45 Tests as cut-off for bowlers. These two sets of twin conditions ensure that bowlers such as Barnes do not get into the picture. Most of the top keepers get in.

Only the batting average and bowling average are used for comparison. These two are the most trusted of all measures and will provide a very good platform for a clear understanding of a Test players' career.

Test Batsmen: Analysing the two career halves

SNo Cty Batsman         |<----Career---->|<--1st Half->|<-2nd Half>| % Chg
                        |Tests Runs  Avge|Mt Runs  Avge|Runs   Avge|
                        |                |             |           |
  1.Pak Younis Khan     |  63  5260 50.10|32-2033 39.10|3227  60.89| 55.7%
  2.Zim Flower A        |  63  4794 51.55|32-2013 41.94|2781  61.80| 47.4%
  3.Aus Redpath I.R     |  66  4737 43.46|33-1813 35.55|2924  50.41| 41.8%
  4.Nzl Wright J.G      |  82  5334 37.83|41-2123 31.22|3211  43.99| 40.9%
  5.Aus Chappell I.M    |  75  5345 42.42|38-2219 35.22|3126  49.62| 40.9%
...
 53.Eng Hobbs J.B       |  61  5410 56.95|31-2733 56.94|2677  56.96|  0.0%
...
 97.Aus Hayden M.L      | 103  8626 50.74|52-4714 58.92|3912  43.47|-26.2%
 98.Eng Smith R.A       |  62  4236 43.67|31-2255 51.25|1981  37.38|-27.1%
 99.Win Kallicharran A.I|  66  4399 44.43|33-2582 52.69|1817  36.34|-31.0%
100.Aus Gilchrist A.C   |  96  5570 47.61|48-3073 59.10|2497  38.42|-35.0%
101.Aus Harvey R.N      |  79  6149 48.42|40-3830 61.77|2319  35.68|-42.2%

Younis Khan has achieved the highest jump from the first half to second half, an astounding 55.7%. His average has improved from 39.10 to 60.89. Note that in his last 31 Tests he has scored at higher than 100 runs per Test.

Andy Flower has improved from 41.94 to 61.80, an increase of 47.4%, that too playing in a weak team. Ian Redpath, John Wright and Ian Chappell have also finished their careers very strongly.

For consistency one need not look beyond Jack Hobbs. He has only a second decimal difference in his second half average to the first half. Steve Waugh and Andrew Strauss are close to achieving this perfection.

Gilchrist's huge fall, from 59.10 to 38.42 is understandable considering that he had an explosive start and fell off drastically towards the end. What is surprising is the fall of Neil Harvey, who dropped his average from 60+ to 35. This is quite inexplicable. He scored 15 of his 21 hundreds in the first half of his career. Gilchrist, on the other hand, scored 9 of his 17 hundreds in the first half of his career. However he was dismissed for many single digit scores, quite a few 0s included, during the second half.

Note how Hayden, R Smith and Kallicharan have also fallen off.

To view the complete list, please click here.

Test Bowlers: Analysing the two career halves

No Cty Batsman          |<----Career---->|<-1st Half-->|<2nd Half>| % Chg
                        |Tests Wkts  Avge|Mt Wkts  Avge|Wkts  Avge|
                        |                |             |          |
 1.Eng Laker J.C        |   46  193 21.25|23-  78 29.95| 115 15.35| 48.8%
 2.Eng Bedser A.V       |   51  236 24.90|26- 100 33.87| 136 18.30| 46.0%
 3.Pak Iqbal Qasim      |   50  171 28.11|25-  65 35.78| 106 23.41| 34.6%
 4.Nzl Hadlee R.J       |   86  431 22.30|43- 192 26.17| 239 19.19| 26.7%
 5.Nzl Morrison D.K     |   48  160 34.68|24-  73 39.53|  87 30.61| 22.6%
 6.Slk Muralitharan M   |  129  783 22.22|65- 337 25.48| 446 19.76| 22.5%
...
38.Aus McKenzie G.D     |   60  246 29.79|30- 126 29.81| 120 29.77|  0.1%
...
66.Win Gibbs L.R        |   79  309 29.09|40- 176 24.56| 133 35.09|-42.9%
67.Pak Mushtaq Ahmed    |   52  185 32.97|26- 105 27.51|  80 40.14|-45.9%
68.Win Hall W.W         |   48  192 26.39|24- 119 22.15|  73 33.29|-50.3%
69.Eng Botham I.T       |  102  383 28.40|51- 231 23.46| 152 35.91|-53.1%
70.Eng Lock G.A.R       |   49  174 25.58|25- 104 20.13|  70 33.67|-67.2%

Laker moved from an average spinner to Lohmannish figures in the second half, no doubt aided by the 19 for 90 at Manchester. That is nearly 50% improvement. Similar with Alec Bedser, who had totally different career halves. What about Richard Hadllee, with sub-20 average in the second half of his career. Again Muralitharan's last 64 Tests have had sub-20 average and an average of 7, yes, you read it correctly, 7 wickets per Test.

McKenzie was like Hobbs, averaging almost the same figure in his two halves. Saqlain Mushtaq and McDermott are in the middle group.

Look at the last five, especially Ian Botham. He was a shadow of himself, increasing his average by over 50%. Lock's figures are still more astounding. An average of 20.13 moving to 33.67 and below 3 wickets per Test. Possibly he played the supporting role to Laker quite often as happened at Manchester in 1956.

To view the complete list, please click here.

This blog is going nowhere with readers following a single agenda, whatever be the subject matter of the article. I have had complaints from serious readers that the purpose of the articles is lost. Hence a firm reminder that only relevant comments will be published. Henceforth I will not and readers should not forget that the purpose of the blog is to come out with new analytical efforts. I myself have been guilty of side-tracking into irrelevant and/or non-cricketing issues. Remind me, gently or otherwise, to remove the offending comment or response.

Good Analysis. But somewhere, the numbers hide the story.

"For consistency one need not look beyond Jack Hobbs. He has only a second decimal difference in his second half average to the first half. Steve Waugh and Andrew Strauss are close to achieving this perfection."

Steve Waugh (split into roughly one third and two thirds)
M Inn Runs Avg
52 80 2503 36.27( First 52 tests)
116 180 8424 58.09 ( Last 116 tests)

A dramatic improvement of 60% over the approximate the last 2/3 of his career. He also scored 28 of his 32 hundreds in the later segment. This not a purple patch analysis as both parts are significantly large (equal to careers of many cricketers) However, split into 2 equal halves he is the epitome of consistency.

Can we have a analysis which divides player careers from watershed moments rather than 50-50. I know one will have to look at each player individually, but some interesting results may show up.
[[
Raghav
One thing I am certain of. Give you guys a centimetre and you will claim a metre. That is a valid point and let me add to the list of things to be done. Tough job since I have to find a way to build an algorithm to automatically determine such wide variations within a single career.
Can you look at the article I did a few months back on streaks, both at the high end and low end.
Ananth:
]]

Ananth,
To me the stand-out point of this analysis is Bedser, Laker and Harvey. The Australian targeted the debutant Laker in 1948 and he succumbed spectacularly under pressure (a bit like Warne against India in 1992). Bedser was only at the beginning of his career and was grossly over-bowled in both the 1946 and 1948 series. On the flipside, Harvey was the beneficiary of possibly the most severe bowling drought in test cricket history. There just wasn’t a decent bowler to be found outside Australia during those five lean years after WWII. And the batsmen, particularly the Australians, feasted. Arthur Morris would also be on your list if you relaxed the cut-off point to 3500 runs. Of course Bradman also benefited from this, as he averaged well over 100 after WWII despite all experts being in agreement that at this point he was way past his best.
[[
Deon (or Xolile - I presume that is a pseudonym),
I am sure you have understood that the only factors considered are those within a player himself. I am surprised to see Hayden dropping 26%. Was it visible.
Ananth:
]]

Ananth, Raghav Bihani

No need for such a complicated procedure. I’ve posted a comment on the previous blog towards the end. This covers most of the confusion, i.e. in the case of most batsmen the 2000s figures will be better than the 1990s.
However, in the case of S.Waugh his 90s avg is 53.06 (89 matches), 18 100s.
His 2000s avg. is 53.30(44 matches), 11 100s. But then his last year, 2003, yielded 996@76.61!!
So taking his “last” 116 matches to mean that they yielded a consistent 58.09 does not really denote the correct picture.
His best patch will be somewhere in between. So, as with most pre 2000s batsmen it will actually be a bunch of matches mid career, when at their peaks.
Only in the case of the 90/2000s batsmen will you find figures improving in the 2000s irrespective of period of career.
[[
Abhi
There is no confusion nor a complicated procedure being suggested. You seem to be generalizing. Different batsmen have different career half formations. Otherwise how do you explain the two totally contrasting half patterns of Hayden and Younis Khan.
Ananth:
]]

Abhi,

As Ananth has said this is not that complicated. And different players have different cycles. While Ananth has an excellent database and methods at his disposal, using simple cumulative and reverse cumulative averages one can do the exercise.

The problem is that here I am using my judgement for the watershed moments rather than a formula. Statistical analysis should be fully objective, which I cannot do with cricinfo ony.

Following is my demarkation for Ponting.
M Inn NO Runs Avg 100s
30 47 4 1661 38.62 4 (First 30)
76 130 21 7387 67.77 28 (Middle)
30 52 1 2297 45.03 6 (Last 30)

Notice that each period should be sufficiently large that it cannot be called a purple/lean patch. Thus one should avoid batches of 10-15 matches only where a player may have Bradmanesque figures. I think somewhere above 25 matches should be a cutoff to make a time period in a career.
[[
Raghav
I like the idea. However please have a look at the streaks article I had referred to earlier. It is titled "The highest peaks and lowest troughs in player careers" and it was published almost exactly an year back. there I have done a similar type of analysis.
Ananth:
]]

Excellent article on peaks and troughs. But I feel that 10-15 matches at superb figures (avg of 100+) are achieved by average players also. However maintaining even 60+ over 50 matches is done by the class acts.

However on another note, when i went through the article I saw Ponting had a lean patch during innings 61-70. This made me go back to his numbers and revise my demarkation as below.

M Inn NO Runs Avg 100s
45 71 8 2552 40.5 7 (First 45)
61 106 17 6496 72.99 25 (Middle)
30 52 1 2297 45.03 6 (Last 30)

This is again the big problem. Stats analysis should be objective and I being subjective made a mistake. The average below 40 for the first 30 matches got me.

This analysis (Instead of streaks/troughs this will be Golden Periods) should be for a very limited set of players (50 bat + 50 Bowl). Perhaps you can do the necessary calculation using your database (We always claim a metre given a cm).
[[
Raghav
This article was the equivalent of, say, the serve in Tennis. We will trade strokes and come to a conclusion. Both your and Abhi's work seems to me to indicate that the split should not be down the middle but rather a third and two-third. I will wait for other comments and, maybe, do a quick follow-up on these lines.
Thanks to you and Abhi.
Ananth:
]]

Ananth, luck needs no explanation. If Laker's '56 results were in the first half of his career, not the second, his numbers would be more than reversed (15.28 vs 31.99). The fact that he has significant variation over his career tells you no more and no less, than at one point in his career, he was blessed with a near perfect summer (weather having no small part to play in that result). Similar things could be said about just about every other difference mentioned above. The reason's might be to do with mid-career bulges, or changes in bowling attacks, or something else, but if so, those things should be tested empirically and properly. Examining the randomness of a player's career and proclaiming some insight into it is akin to reading chicken entrails.

What the analysis shows is that (assuming the two halves of a career should be equal) the standard deviation of a player's average over more than 20 tests is around 25%. Even an established average is a poor predictor.

Ananth,
Donald Bradman’s test career comprised of 80 innings. I recommend that we used this as the definition of “standard test career”. Players with more than 80 innings could be said to have had long careers, and those with fewer than 80 innings could be said to have had short careers.
We can then look at “careers within careers” for players that played 80 or more innings. This would simply involve calculating the 80-innings rolling career average for each player – and then looking at the maximum, the minimum and the difference between the two as a %.
For me it would be extremely insightful to see who ends up in second place behind Sir Donald.
[[
Deon
52 tests/80 innings seems to be the perfect cut-off point. I use 5000 test runs only to exclude some tail-enders. A combined checking of 5000 runs/52 tests will probably will have around 100 players.
Your idea and Arjun's combined together will mean I look for the best 52-test stretch in any player's career. Will do. For the follow-up article.
Ananth:
]]

They used to show a graph on TV years ago that had bars showing every innings a batman played with his average superimposed over the top in the form of a line. The average was adjusted after each innings so you could see very quickly how a career had progressed - for instance the classic situation where the average started off low, then rose throughout the career and tailed off near the end.

I mention this because a graph of this nature imparts the sort of information we're discussing here in a very efficient manner - the nature of a player's career quickly becomes apparent, and there is no need to decide where to split a career in half.

Perhaps to be more accurate the graph could show the average for the last 10 or 20 matches - in the same way we show the runs-per-over for the last 5 overs as a gauge of the batting side's momentum.
[[
Aaron
The problem with graphic images is that it is perfect when the commentators, or anyone else, for that matter, want to show one player's career on the screen. What do we do when we want to compare over 100 players and find out the extremes.
Ananth:
]]

Dear Ananth, Nice article once again!

Brian Lara's career has three segments
1. In first 31 test matches in 52 innings, he scored 3048 runs with the help of 7 centuries and 16 half centuries with batting average 60.96.

2. Then in 49 test matches he scored 3485 runs with help of 8 centuries and 17 half centuries with batting average 40.05.

So in first 80 test matches he scored 6533 runs with the help of 15 centuries and 33 half centuries with batting average 47.68.

3. In his last 51 test matches in 91 innings, he scored 5420 runs with the help of 19 centuries and 15 half centuries with batting average 60.89.

Totally he scored 11953 runs with help of 34 centuries and 48 half centuries with batting average 52.88.
He is excellent up to .24 of his test matches and then his career is below to his standard from .24 to .61 and then from .61 to onward excellent again.

It means that his career has two peaks and one decline and his .37 career is not up to his standard.