The Role of Age in Red Ball Batting Performance
Do older batters really do better? and other questions relevant to investment in and by players.
I did a follow up to this with a dive into aging for white ball batters. You can find the write-up on that here.
Aging in sport can be a double-edged sword. On one hand, aging is associated with deteriorating physical abilities — stamina, reflexes, coordination, strength etc. while on the other hand, older athletes develop a better understanding of the game and their game, and gain the benefit of tactical level-headedness that is brought about by experience i.e. repeated and prolonged exposure to a range of oppositions and scenarios.
Here, I try to understand how cricketers age, when they peak, and how they decline, with a focus on red ball batting. I examine some conventional wisdom around this and some analytical write-ups that I found while doing my research and then attempt to plug the holes that I found in these by leaning on some prior work from folks in Baseball. I then apply this new methodology to explore typical aging trends and understand how we can leverage this information in Test and First-class Cricket.
While there are several in-depth bits of interest in the piece, here are some of the highlights:
- The impression that batters get better as they age into their 30s is not broadly true.
- Age related performance trends are relative to the level of competition. The typical batter peaks in their early 20s, while really good batters relative to the level of competition peak in their late 20s to early 30s and decline at a slower rate.
- Looking at the cumulative batting average is not enough to assess the timing and risk of investing in a batter.
- Selectors do well at identifying young talent and inducting them into the Test XIs. There is a large enough gap between the domestic First-class & Test levels that makes this easier.
- Aging plays out very differently against pace and spin.
Why is it interesting to look at aging?
Fans are perplexed when their idols’ record doesn’t justify their own adulation anymore. They don’t know whether they should be hopeful or eager for a regression back to their perception of normal. Critics have a similar in magnitude, but reciprocal sentiment towards changes in player records over time.
Teams should care to understand how players age because it can help them answer questions around how to invest in players and what to expect from them — Will a player who seems to be doing better and better keep improving indefinitely if they continue to get the help needed? What is the upside potential and risk of giving the same resources and attention to players in different age brackets & how can they be allocated effectively? What is the implication on investment areas at places like the NCA? How do we think of team composition to ensure smooth transitions without impact on performance at the team level? Teams and players themselves would also appreciate an understanding of how likely it is for a player to stage a comeback, what to expect from a comeback and the heroics needed to pull it off. What about a role change?
Aging in Cricket has been somewhat misinterpreted and misrepresented
Cricket pundits tend to claim that batters ripen with age and hit their peaks at around 30 years of age. Here’s some numerical analysis that I was able to find, suggesting that Test batters peak at 32 years of age & Sheffield Shield batters peak after 30. Here is one that claims that “the life span of the modern batsman is still very good at least up until the age of 35” & this piece takes a similar approach but without proclaiming a conclusion with the same clarity as some of the others.
These investigations broadly look at batting averages by age and conclude that older batters do really well. Let’s replicate this. I did this in two ways — looking at overall average by age and looking at the batting average of the average batter by age.
What this tells us is that older Test batters have years with higher averages. It doesn’t tell us that batters peak in their 30s.
There is some strong survivor bias at play here. Batters who are still playing in their 30s are inherently good batters, otherwise they would have been weeded out earlier in their careers. At the same time, the batting pool at the younger ages have a significant number of vanilla batters, pulling those numbers down. Only 44% of batters in the pool above play beyond 30 years of age. These batters have career averages that are ~7 runs higher than those whose careers end before 30.
In other words, better batters have longer careers as we’d expect them to. Crucially, this doesn’t necessarily mean that they play at their personal best when they are older. They do have more opportunities at a higher one-off peak year later in their career. Also, while we didn’t do so here, we tend to look at the career batting average of a batter even as we look at them over time. This is a cumulative measure and hence is a lagging indicator of changes in performance trends as we will soon see in more detail.
One way to incorporate player specific peaks could be to look at the age at which each player has their highest average to find performance peaks. However, this is contingent on the the opposition and conditions that they faced at each age. For someone who played in a couple of home series in one year and a couple of away series in the next year, one would expect to see a decline in unadjusted performance. Moreover, overall Test averages themselves can change by year, owing to some extent on these factors.
Someone who had a consistent batting average of 40 through the 90s and early 2000s significantly over-performed their peers in the 90s compared to the 2000s.
Another issue with looking at peaks per player is that short careers are the norm. So, we can’t tell how each player would have done across a broad range of ages. Moreover, different players have different ranges of baseline performance and hence aging curves.
In doing some research, I found that Omar Chaudhuri, now an executive at a sports intelligence company, Twenty First Group, alluded to these issues here, but I am not convinced of the aptness of the method he chose to address them. He created age buckets and looked at the top batters in each age group as well as batters who played across each of the age groups. In my understanding, this attempts to apply the biased selection to each group and looks into continuity across careers. However, this is still subject to variations in trends of overall averages across years. Most importantly, this reduces the number of players to a very small sample size.
An approach that mitigates bias
In this section I go into some details on methodology. If you are mainly interested in outcomes, skip to the next section — “The Aging Curve”.
No matter what one does, it is hard to overcome selection bias completely in the analysis because it is inherent to the underlying system which relies on selection based on past performance. Ideally we would want a large number of players who play against a similar opposition in similar conditions for a wide age range, say 15 through 50, while scoring trends in the game remain stagnant as well. This is obviously not possible. However, we can try to have the data be representative of all batters while looking to see when the typical player may be peaking and how they decline. The methodology I use here is inspired by and in some cases follows prior work done in baseball referenced in 1, 2, 3, 4, 5, 6.
The first thing we can do is to switch from using raw averages to averages above the expectation. For coming up with an expectation, I decided to account for the calendar year, batting position, innings in the match and continent. This normalizes differences in conditions, scoring trends etc. across years and increases signal.
For example, here is sample of Graeme Hick’s innings at age 29 and the groupings used to calculate expected and actual average. For the expected average, we use all players with the respective grouping and not just Hick’s performances.
The second thing we can do is to look at improvement instead of raw performance by taking year over year deltas per player. This normalizes the measure across players and allows us to pool data from all players , since different players will have different baseline averages. The definition of a peak is the time when a player stops improving and starts declining. Taking deltas also means that players without consecutive yrs get removed (at least for those yrs). This filters down to consistently picked players but it still allows us to use data from players with shorter careers.
For example,
At this point, we have a set of “improvements” per player-year. However, different players play different number of innings and the number of innings is also different across the two years for which we took the delta. We take this into account as we weight the regression of the relationship between age and improvement.
The improvements can now be strung back to reconstruct the raw value. Remember, this value is the batting average over expectation per year. We will also recenter this to the value at the age of the average player weighted by innings count. Given a batter in the top 6 in an innings, their average age is 29. So we will have the value at age 29 as 0 and the rest of the curve will tell us the delta in batting average from this average player.
The Aging Curve
This is the basic approach that we will be using. To quickly recap, we aligned all batters to the same scale and removed noise in performance, converted it to an improvement (YoY) paradigm, leveraged data across players and strung it back to go back to the performance by year paradigm. We then normalized the resulting curve to the average age of a batter in the data set. For clarity of insights, in some cases, we may normalize to a different point, which will move the entire curve vertically. We will call that out as such when we do so.
Note, that the curve we have here shows batting average in the respective age-year and not the cumulative batting average. The batting average that we are used to tracking as part of player stats is the cumulative batting average which is a lagging indicator for the purposes of identifying peak in skill and age based decline which is what we are interested in here. Besides, when picking a team in the present or when making investment decisions for games n years from now, it is a better idea to consider the value that a player is likely to bring to the table now and at the end of n years respectively. While we won’t be using this, the below chart shows what the cumulative batting average weighted by the median number of innings at each age would look like for the aging curve that we obtained above.
Should this be closer to perception than it is?
The curve we have obtained above is very different from the general perception. But, our perception may be influenced by successful players. So let’s look at the top batters. I took the top 100 batters by career batting avg with at least 20 Test innings. The top players in this list are Jacques Kallis, Kumar Sangakkara, Steve Smith, Graeme Pollock, & Adam Voges between 55 and 61 runs of avg and the bottom players in the list of 100 are Sourav Ganguly, Chris Gayle, Desmond Haynes, Ian Chappell, & Roy Fredericks all around 42 runs batting avg.
This shows a few things
- Top batters continue improving and peak at a later age. By 7 years! However, they are still on a decline beyond 30 yrs of age.
- The top batters have a plateau rather than a peak. Their peak is longer lasting (~3–5yrs)
- Top players perform better than the average player at their peak age by ~4 runs and then improve on that later in their career an additional 3–4 runs per dismissal.
- Top batters decline more gradually, thus widening the gap to the average player of the same age later in their career. They don’t underperform the average active batter in the game until much later in their careers.
A forward looking perspective
So far, we have looked at the aging curve in retrospect. How do we know which curve a player is going to follow? How do we know whether an older player has left their peak behind?
If we look at the batting averages of players at age 25 and use that to group players, we can see when each group peaks and see that better players peak later. Additionally, batters with an avg higher than 35 at 25, also decline at a slower rate than the average batter.
This shows that while older players regardless of their performance have usually left their peak behind, the key is to look for YoY improvement in performance in the early years of a batter’s career. Batters rush against age to get to as high a peak as possible. The longer they can keep improving, the higher the peak and the longer the buffer to fall. The typical player playing at a batting average of 30–35 at age 25 has already been on the decline for 3 years. It is just that their cumulative batting average hasn’t caught up to this yet. They will start performing below average in a couple of years. Contrast this to the typical player who peaked at 24. Their peak is not that much higher, but they will start performing below average almost 4–5 years later even though they are also on a decline. The difference in the typical rate of decline exacerbates the difference between these two players.
It is important to understand that what we have seen here, is what happens to the typical player. When we ask questions about a particular player, the answer becomes a bit more elusive. Of the approximately 700 players that I looked at here, more than 150 had their single best year after the age of 30. This could be a flash in the pan, a late comeback or consistent improvement, in order of likelihood.
Player Selection from the Domestic Circuit
Given that players are usually on the decline in Test Cricket, timely player indictment at the Test level is important. If we compare the average player who is picked before 22, which is about 20% of players, to the average player who is picked after 22, we see that for the average player who joins after 22, there is no improvement ramp.
Importantly, we can see here that selectors are doing a good job identifying young talent. The typical player who is identified to be introduced early is really good, so they realize their potential early and do well in the long run. Given that domestic First-class setups are the breeding grounds and funnels for playing at the Test level, looking at the aging process in the First-class set up should shed some more light on this.
The red and the black curve shows the difference in aging between the two levels for the same set of players, while the blue curve shows aging for players who don’t make it to the Test level.
Here are my takeaways from this:
- There is a huge gap between the Test level and the FC level of competition — worth about 25 runs per inning above the average for the respective expectations including levels and other factors that we’ve accounted for in determining the expectation.
- Similar to what we saw with the top Test batters, the better players at the First-class level (those who make it to Tests) peak at a much later age (in fact peaking exactly at 29).
- For par-skilled players within a level of competition (Test players playing Test matches and FC players of lower skill playing FC matches), age related decline typically starts in the early 20s.
- The higher the relative level of competition, the steeper the age related decline i.e. the larger the advantage a younger player has.
This tells us that as a rule of thumb, the typical batter peaks in their early 20s, while really good batters relative to the level of competition peak in their late 20s to early 30s and decline at a slower rate.
The separation between the red and blue curves above, between Test players and non-Test players in FC Cricket, especially the fact that this separation starts early in playing careers makes this a useful aid for identifying promising Test candidates. Given how we have seen that cumulative batting averages, while being good indicators of consistency, can be lagging indicators of age related decline and given how early improvements in performance can be a differentiating factor for players in the long run, the latter is a great complementary tool at our disposal when making decisions on selection for players young and old. In the most rudimentary form, this calls for a weighted moving average or a windowed moving average type metric for batting average. We won’t get into the distraction of developing that here, but let’s take a very brief look at Suryakumar Yadav’s selection in the Test squad over Sarfaraz Khan as a case study.
While the Sarfaraz vs. SKY debate may not be consequential since there may not be an actual spot on the playing XI against Australia, the selection in itself is an interesting statement. Let’s look at their performances in First-class Cricket in the context that we have been looking at here.
Many have argued without looking at this chart that SKY’s selection is driven by his performances in white ball Cricket. They may not be wrong, but we see here that short form Cricket seems to have led to a reinvention of his game with the red ball as well. What had looked like an above average First-class career fizzling out is suddenly showing glimpses of being above that of an average Test player in First-class Cricket, and this performance is being achieved on the back of a different brand of batting at a much faster scoring rate. Sarfaraz, on the other hand has been consistently glorious since 2020. His brand has been closer to traditional red ball batting and so if all stakeholders — players, selectors and management, have clarity around what is on offer from each player, there is enough light between the two to make a good decision. What they should be keeping an eye on is any decline in Sarfaraz’s performance from here since he hasn’t necessarily been improving. This is tricky since he may have reached a ceiling from where the only way to go is sideways or down.
Aging Against Pace & Spin
We’ve taken a look at how age affects overall performance. If we think one level deeper about how aging impacts our bodies, where do we expect age to have the most impact as a batter? The quickness and coordination of hands, eyes and feet? Ideally, we would investigate the impact of these using biomechanical markers across a range of players, but given that we don’t have that luxury yet, let’s take a look at how aging impacts performance against pace and spin and infer what we can from that. The premise here is that facing pace and spin require different primary skills and physiological mechanisms as far as reaction time and coordination are concerned.
Against pace, where batting relies a bit more on reflexes, the decline starts early but the curve is flatter both on the improvement side and on the decline. Batting against spin is arguably slightly more of a cat and mouse game that has room for strategic growth and batters improve as they learn to read quality spin at the highest level over the first few years. The steeper decline against spin may be a result of a decline in swift footwork with age.
To recap, here are the top 3 takeaways that we have with respect to aging:
- The average Test batter peaks in their early 20s and declines from there. They continue to improve in First-class Cricket, peaking in their late 20s and early 30s unlike the average First-class batter who also peaks in their early 20s.
- For evaluating players to invest in for the Test squad, YoY improvement and per year/season batting average trends relative to the competition are a useful complementary tool to cumulative batting averages which are lagging indicators of declining performance.
- Aging trends can look different based on the physiological skill involved as well based on the measure of performance (Here, we used the batting average as a measure of red ball batting performance). This should also extend to playing conditions and different formats.
Going back to the article using 35+ year old players in the 2010s doing well as an example of a general trend of “Older, Wiser, Deadlier” batters in the sport, older players have actually declined a bit faster if anything in this, the T20 era. This doesn’t negate the fact that we have seen the Sangakkaras, the Misbahs and the Younis Khans in this era, but they are not the typical player; they are exceptional players pulling off heroic feats.
If you enjoyed this piece, check out more of my work at Boundary Line and follow along here & on twitter @amol_desai
