December 6, 2008

The optimum age for a cricketer

Analysis performances to determine the optimum age for batsmen and bowlers
21

Our CSW database has the capacity to analyse data by age, so I decided to use it to investigate what age(s), if any, provided significantly better performances.

My sample was Sheffield Shield data since 1977, when the newest state, Tasmania, entered the competition. This provided reasonably homogenous data, with little of the cultural variation that might be obtained using Test match data. The sample thus analysed over nearly 950 matches, involving 800 players.

The results are in the table below:

Age-wise averages with bat and ball in Sheffield Shield since 1977
Age Batting average Bowling
<20 28.34 34.01
20 26.66 35.78
21 29.08 36.45
22 30.55 36.92
23 30.31 36.10
24 31.23 34.93
25 30.77 33.22
26 31.67 33.56
27 30.99 32.75
28 31.53 32.86
29 32.41 31.45
30 33.96 30.58
31 32.35 32.19
32 34.63 30.42
33 31.64 29.93
34 30.71 31.19
35 32.72 33.32
>35 33.06 34.97

What, if anything, can be deduced from the results? Well, clearly, both batting and bowling averages improve as one gets older, but the extent to which this happens after the age of 30 surprises me. The picture of the young, virile cricketer in his early-20s emerging triumphant over the aging has-been is not sustained by this data, and it would seem that experience outweighs youthful exuberance more times than it doesn't. For both batsmen and bowlers, the ages of 32 - 33 are the vintage years, and perhaps we are too keen to write players off as they move through their early-30s.

I would be most interested to read of the comments by others in relation to this data, and data from other spheres of cricket.

Comments have now been closed for this article

  • santhosh on December 20, 2008, 2:00 GMT

    Well, good analysis! Whilst I would like to retort sanguinely only after a chary construal on the basis of age, location played at, opponent team aganist played, career period lasted, total matches played by representing team and in a calender year and overall average runs scored in a match as well by individual innings by batting / bowling first or second respectively. However rudimentry of this critcial analysis is to dwell accentuate!

  • Prasanth on December 7, 2008, 8:56 GMT

    This analysis holds good for cricketers from Australia alone. In other countries like India and Pakistan, youngsters make debut and for some strange reason do well as compared to their Aussie counterparts. Case in point, the U-19 WCs, and U-19 competitions etc. Just a thought.

  • Baundule on December 7, 2008, 4:01 GMT

    Good analaysis; but a better way to look at it would be first selecting a group of players who had a career span from, say 20 to 35 years and then compare their performances on the age basis. For example, if would be interesting how Brian Lara performed when he was 22 or 25 compared to when he has been 30+. The same goes for bowlers like Shane Warne, Murali or McGrath.

    An average over a sample of such players would probably be a better indication of what happens.

  • maddy on December 7, 2008, 3:28 GMT

    It would be great if this analysis is restricted to just players who have played for their country or also include players who have had a fairly long first class cricket. By this way we can eliminate the mediocre ones and also the ones who didnt play after 22-23 etc....

  • Jamal Khan on December 7, 2008, 2:37 GMT

    I have seen that it is different for batsmen and bowlers. Specially a genunie fast bowler is better when is between 24-30 years old. Once he starts aging his bowling avg. starts getting bad. When Waqar Younis started he was the best bowler i have ever seen in cricket. His yorkers were so fast that it was impossible for batsmen to block them. But when he became older, same yorkers became full toss since he was unable to bent his back as much as he should have to make it a yorker.

  • Marcus on December 6, 2008, 23:24 GMT

    Dr. Manish Agrawal- many Australians make their Test debuts late, but this data is about Sheffield Shield cricket, and in fact most Australian cricketers make their first-class debuts quite young, and play Test cricket with seven or eight years' experience behind them.

    While I agree that the data will be skewed by established older players performing better than inexperienced younger players, I can't help but wonder what some players might have achieved if they'd played longer, such as Graeme Pollock who played his last Test at 26.

  • David Barry on December 6, 2008, 23:21 GMT

    One thing that should work better is to ask, How do really good players' performances change with age? This way you can just consider all players with at least (say) 50 Tests and see how their performances vary over those 50 Tests. I did some here for bowlers:

    http://pappubahry.blogspot.com/2008/08/bowlers-as-they-get-more-experienced.html

  • David Barry on December 6, 2008, 23:16 GMT

    Aging curves are difficult. The selective sampling issues raised by commenters above can have HUGE effects on the results.

    A better method is to compare each player's results in one season with his results in the next season. Do this for all players, take ratios, combine together over the course of a career, and you've got an aging curve.

    The problem is another selective sampling problem - those players who play in the next season probably had a good previous season, and having a good season is partly luck and partly talent. So you're selectively sampling players who were lucky in the first season (and so you'll observe bigger declines than there really are).

    To fix this problem, you need to regress the first season's performances to the overall mean - I don't know of anyone doing this rigorously in cricket yet. And this step is very important - in certain aspects of baseball performance, this can change the apparent peak performance age by TEN YEARS!

  • alex on December 6, 2008, 22:25 GMT

    Surely this is skewed by better players having longer careers and continuing to their 30's, where as worse ones will have been dropped by then.

  • Kartik on December 6, 2008, 22:22 GMT

    I distinctly notice that white and black cricket players often do perform at full quality even after the age of 36. Border, Waugh, Hayden, Warne, McGrath, Lara, Richards, Greenidge, Walsh, etc. all come to mind.

    Indian players usually drop of totally after the age of 34, even if they are batsmen. Gavaskar and Jayasuriya are perhaps the only partial exceptions.

    Hence, the top-tier Australian cricketer usually retires while still in solid form. Note how well S. Waugh, McGrath, and Warne did even in their last series. The Indian cricketer, on the other hand, usually only retires when their form has been bad for long enough to be dropped, and this is usually at a young age.

    Dravid and Tendulkar both dropped off in form after the age of 33 or so. Contrast that to Lara and Hayden's performance after age 33.

  • santhosh on December 20, 2008, 2:00 GMT

    Well, good analysis! Whilst I would like to retort sanguinely only after a chary construal on the basis of age, location played at, opponent team aganist played, career period lasted, total matches played by representing team and in a calender year and overall average runs scored in a match as well by individual innings by batting / bowling first or second respectively. However rudimentry of this critcial analysis is to dwell accentuate!

  • Prasanth on December 7, 2008, 8:56 GMT

    This analysis holds good for cricketers from Australia alone. In other countries like India and Pakistan, youngsters make debut and for some strange reason do well as compared to their Aussie counterparts. Case in point, the U-19 WCs, and U-19 competitions etc. Just a thought.

  • Baundule on December 7, 2008, 4:01 GMT

    Good analaysis; but a better way to look at it would be first selecting a group of players who had a career span from, say 20 to 35 years and then compare their performances on the age basis. For example, if would be interesting how Brian Lara performed when he was 22 or 25 compared to when he has been 30+. The same goes for bowlers like Shane Warne, Murali or McGrath.

    An average over a sample of such players would probably be a better indication of what happens.

  • maddy on December 7, 2008, 3:28 GMT

    It would be great if this analysis is restricted to just players who have played for their country or also include players who have had a fairly long first class cricket. By this way we can eliminate the mediocre ones and also the ones who didnt play after 22-23 etc....

  • Jamal Khan on December 7, 2008, 2:37 GMT

    I have seen that it is different for batsmen and bowlers. Specially a genunie fast bowler is better when is between 24-30 years old. Once he starts aging his bowling avg. starts getting bad. When Waqar Younis started he was the best bowler i have ever seen in cricket. His yorkers were so fast that it was impossible for batsmen to block them. But when he became older, same yorkers became full toss since he was unable to bent his back as much as he should have to make it a yorker.

  • Marcus on December 6, 2008, 23:24 GMT

    Dr. Manish Agrawal- many Australians make their Test debuts late, but this data is about Sheffield Shield cricket, and in fact most Australian cricketers make their first-class debuts quite young, and play Test cricket with seven or eight years' experience behind them.

    While I agree that the data will be skewed by established older players performing better than inexperienced younger players, I can't help but wonder what some players might have achieved if they'd played longer, such as Graeme Pollock who played his last Test at 26.

  • David Barry on December 6, 2008, 23:21 GMT

    One thing that should work better is to ask, How do really good players' performances change with age? This way you can just consider all players with at least (say) 50 Tests and see how their performances vary over those 50 Tests. I did some here for bowlers:

    http://pappubahry.blogspot.com/2008/08/bowlers-as-they-get-more-experienced.html

  • David Barry on December 6, 2008, 23:16 GMT

    Aging curves are difficult. The selective sampling issues raised by commenters above can have HUGE effects on the results.

    A better method is to compare each player's results in one season with his results in the next season. Do this for all players, take ratios, combine together over the course of a career, and you've got an aging curve.

    The problem is another selective sampling problem - those players who play in the next season probably had a good previous season, and having a good season is partly luck and partly talent. So you're selectively sampling players who were lucky in the first season (and so you'll observe bigger declines than there really are).

    To fix this problem, you need to regress the first season's performances to the overall mean - I don't know of anyone doing this rigorously in cricket yet. And this step is very important - in certain aspects of baseball performance, this can change the apparent peak performance age by TEN YEARS!

  • alex on December 6, 2008, 22:25 GMT

    Surely this is skewed by better players having longer careers and continuing to their 30's, where as worse ones will have been dropped by then.

  • Kartik on December 6, 2008, 22:22 GMT

    I distinctly notice that white and black cricket players often do perform at full quality even after the age of 36. Border, Waugh, Hayden, Warne, McGrath, Lara, Richards, Greenidge, Walsh, etc. all come to mind.

    Indian players usually drop of totally after the age of 34, even if they are batsmen. Gavaskar and Jayasuriya are perhaps the only partial exceptions.

    Hence, the top-tier Australian cricketer usually retires while still in solid form. Note how well S. Waugh, McGrath, and Warne did even in their last series. The Indian cricketer, on the other hand, usually only retires when their form has been bad for long enough to be dropped, and this is usually at a young age.

    Dravid and Tendulkar both dropped off in form after the age of 33 or so. Contrast that to Lara and Hayden's performance after age 33.

  • Ness on December 6, 2008, 21:41 GMT

    I think its because the only people that are playing at the age of 30+ are usually very good and have been in the team for a long time.

  • Ravi Mattu on December 6, 2008, 20:34 GMT

    The problem here is that you have censored data. The odds are that the team selectors are doing a decent job of only retaining those cricketers who are over the hill. A better statistical test is to look at each player who played till age 35 0r 33 and then average their scores only.

  • wulffie on December 6, 2008, 20:34 GMT

    I wonder if the higher averages in the older age groups are skewed by the sample size and selection criter1a.. ie if you survived to play cricket to an older age it was because you were an inherently better cricketer than one who played in their youth for a brief perod of time??

  • Khan on December 6, 2008, 20:16 GMT

    I can personally testify to the accuracy of this data. I have definitely improved in my batting and bowling averages as i have grown older. I am 28 now and I think twice before slogging a ball over the midwicket region when my team is 4 wickets down. Early in my cricketing life, it was much difficult to be patient. As a bowler I have learned not to experiment too much which has led to more wickets and a better economy rate. All of this has come with age. I totally agree that in most cases experience triumphs over youthful exuberance. Thanks for sharing the data...

  • Praneeth on December 6, 2008, 18:50 GMT

    One reason you see the batting averages high for a people of 32-33s is because batsmen/bowlers who are playing for their international team at that age are established players who are very good. If they didn't have a good average, they wouldn't be on the team in the first place. Basically, the older players are good players and that is why their average is high. The youngsters average is low for the same reason. A lot of them are rookies and most of them don't go on to have successful careers. For example, take the Indian team. 30+s who played for India over the last 5 years is limited to Dravid, Sehwag, Tendulkar, Laxman, Ganguly, Kumble, Khan. Whereas the list of <25 is long. And most of them played only about 10-15 games before they were discarded. That is why the averages for that age group are low. I hope this makes sense.

  • Maneesh on December 6, 2008, 18:27 GMT

    Firstly, great thought and analysis

    The younger batsmen are bound to have lower averages 1. They would mostly start in the middle or lower order and work their way upwards 2. Many more entrants therefore, failures will weigh the averages down. At 30+ you are pretty sure of getting set batsmen in the game.

    The bowlers however dont have the same consistency and younger bowlers could be far better than older ones due to raw energy which can work for bowlers I would split the data by Pace vs spin

  • Bearon on December 6, 2008, 18:17 GMT

    As you get older you're only kept in your side if you're the very best. If not then you're generally dropped in favour of younger players. It is also easier to get into the side if you're younger and so there are more possibilities for failures. If you're a really old player and you get to play for the first time then you have to be extraordinarily good.

  • Dr Manish Agrawal on December 6, 2008, 18:05 GMT

    This data shows the performance of only Australians, who as we know tend to start their careers late. Not many Australians have made their debuts before the age of 25. Infact some of the stalwarts like Bevan and Hussey made their test debuts into their 30s. I think the players in the subcontinent tend to peak early. We have too many young prodigies in India and Pakistan, and if you were to compile a data for all the Indian players having played in tests, you will find that the peak performance will be at the age of 24-28 yrs.

  • Prabhu on December 6, 2008, 18:04 GMT

    I would like a similar analysis but only for batsmen with an average above a cutoff (say 30), and bowlers with an average below a cutoff (say 35), and those who have played a minimum of 20 matches or so. That way, we can identify of proper batsmen and bowlers

  • Amirali on December 6, 2008, 16:59 GMT

    There is possibly an inherent statistical bias that the good players are more likely to keep been selected into their thirties, while the mediocre performers are dropped by their late twenties. Hence those mediocre players would weigh down the averages for the younger generation disproportionately. A longitudinal look at a group of players' performance over the span of their careers might be more illustrative.

  • MOCKBA on December 6, 2008, 15:14 GMT

    I did a similar study based on NZ test bowling averages, it clearly showed (with the heavily influence of a Mr Hadlee), that bowling averages improve with age. That up until about 26 averages are pretty poor, and then improve after that.

    In my study and this, what you are faced with is survival bias. Only the best cricketers continue to play after a certain age. The poorer cricketers get replaced by younger players, therefore any player over the age of 30 is overall a better quality player (and will end with a better average) than the average player that is 23 (where only a percentage of which will be quality players).

    If I had such a database readily available to me, I would exclude all players that did not have a 10 year career, and then repeat the analysis.

  • No featured comments at the moment.

  • MOCKBA on December 6, 2008, 15:14 GMT

    I did a similar study based on NZ test bowling averages, it clearly showed (with the heavily influence of a Mr Hadlee), that bowling averages improve with age. That up until about 26 averages are pretty poor, and then improve after that.

    In my study and this, what you are faced with is survival bias. Only the best cricketers continue to play after a certain age. The poorer cricketers get replaced by younger players, therefore any player over the age of 30 is overall a better quality player (and will end with a better average) than the average player that is 23 (where only a percentage of which will be quality players).

    If I had such a database readily available to me, I would exclude all players that did not have a 10 year career, and then repeat the analysis.

  • Amirali on December 6, 2008, 16:59 GMT

    There is possibly an inherent statistical bias that the good players are more likely to keep been selected into their thirties, while the mediocre performers are dropped by their late twenties. Hence those mediocre players would weigh down the averages for the younger generation disproportionately. A longitudinal look at a group of players' performance over the span of their careers might be more illustrative.

  • Prabhu on December 6, 2008, 18:04 GMT

    I would like a similar analysis but only for batsmen with an average above a cutoff (say 30), and bowlers with an average below a cutoff (say 35), and those who have played a minimum of 20 matches or so. That way, we can identify of proper batsmen and bowlers

  • Dr Manish Agrawal on December 6, 2008, 18:05 GMT

    This data shows the performance of only Australians, who as we know tend to start their careers late. Not many Australians have made their debuts before the age of 25. Infact some of the stalwarts like Bevan and Hussey made their test debuts into their 30s. I think the players in the subcontinent tend to peak early. We have too many young prodigies in India and Pakistan, and if you were to compile a data for all the Indian players having played in tests, you will find that the peak performance will be at the age of 24-28 yrs.

  • Bearon on December 6, 2008, 18:17 GMT

    As you get older you're only kept in your side if you're the very best. If not then you're generally dropped in favour of younger players. It is also easier to get into the side if you're younger and so there are more possibilities for failures. If you're a really old player and you get to play for the first time then you have to be extraordinarily good.

  • Maneesh on December 6, 2008, 18:27 GMT

    Firstly, great thought and analysis

    The younger batsmen are bound to have lower averages 1. They would mostly start in the middle or lower order and work their way upwards 2. Many more entrants therefore, failures will weigh the averages down. At 30+ you are pretty sure of getting set batsmen in the game.

    The bowlers however dont have the same consistency and younger bowlers could be far better than older ones due to raw energy which can work for bowlers I would split the data by Pace vs spin

  • Praneeth on December 6, 2008, 18:50 GMT

    One reason you see the batting averages high for a people of 32-33s is because batsmen/bowlers who are playing for their international team at that age are established players who are very good. If they didn't have a good average, they wouldn't be on the team in the first place. Basically, the older players are good players and that is why their average is high. The youngsters average is low for the same reason. A lot of them are rookies and most of them don't go on to have successful careers. For example, take the Indian team. 30+s who played for India over the last 5 years is limited to Dravid, Sehwag, Tendulkar, Laxman, Ganguly, Kumble, Khan. Whereas the list of <25 is long. And most of them played only about 10-15 games before they were discarded. That is why the averages for that age group are low. I hope this makes sense.

  • Khan on December 6, 2008, 20:16 GMT

    I can personally testify to the accuracy of this data. I have definitely improved in my batting and bowling averages as i have grown older. I am 28 now and I think twice before slogging a ball over the midwicket region when my team is 4 wickets down. Early in my cricketing life, it was much difficult to be patient. As a bowler I have learned not to experiment too much which has led to more wickets and a better economy rate. All of this has come with age. I totally agree that in most cases experience triumphs over youthful exuberance. Thanks for sharing the data...

  • wulffie on December 6, 2008, 20:34 GMT

    I wonder if the higher averages in the older age groups are skewed by the sample size and selection criter1a.. ie if you survived to play cricket to an older age it was because you were an inherently better cricketer than one who played in their youth for a brief perod of time??

  • Ravi Mattu on December 6, 2008, 20:34 GMT

    The problem here is that you have censored data. The odds are that the team selectors are doing a decent job of only retaining those cricketers who are over the hill. A better statistical test is to look at each player who played till age 35 0r 33 and then average their scores only.