T20 Player Value: Part III

This is the third post in a series, in which I outline my approach to assessing player value. The first explains the overall objective: to measure the expected contribution of each player in runs. The second then details four main adjustments that I make to historic performances to remove any obvious biases in the data

This post walks-through an example and then adds a further three considerations on top of the ones explained previously: weighting, regression to the mean, and ageing


Inevitably, I chose Chris Gayle as my example. The following few paragraphs walk through the process of estimating how much a team can expect to gain from featuring the West Indian star in their line-up

Across all the matches in which he has played in my database Gayle has added an average of 8.8 runs per innings as a batsman. That is enormous and no contextual adjustment is ever going to change that. However, that figure is boosted by playing at bat-favourable grounds. Playing for RCB at M Chinnaswamy Stadium is an obvious example – the most bat-favourable stadium in my analysis. On average, he gained 0.27 runs per innings from the venues where he performed

On the other hand, he has played in some of the best leagues in the world. The IPL, CPL, and PSL feature the strongest bowling attacks of all the domestic leagues and Gayle has also featured in 49 T20Is and led the Windies deep into the later stages of World Cups. We grant him an additional 0.12 runs per innings to account for that

With the ball, he is net negative in comparison to the average T20 bowler (-1.2 runs). This figure represents how much Gayle would cost his team, on average, if asked to bowl the full 4 overs. Once again, we can adjust that figure based on the venues (+0.12) and the leagues he has played in (-0.04)

Table / chart showing how we rate Chris Gayle contributions in runs including venue, league, batting position, workload adjustments

The batting position and bowler workload adjustments for Gayle are a bit larger. As an opening batsman, he has the privilege to play mainly in the PowerPlay overs and this benefits him to the tune of approximately 1.08 runs per innings. As a part-time bowler, he gets penalised 1.23 runs in comparison to a 4-over bowler, on the basis that he benefits from bowling more when matchups are tilted in his favour

For a player like Chris Gayle, the adjustments make little difference. His baseline numbers are so big and he has played in such a variety of venues and competitions that most adjustments net out to around zero.  Young players like Tom Curran, for example, who have played almost entirely in one league and predominantly at one ground, the adjustments make a bigger difference. Hopefully, the Cape Town Knight Riders have not over-estimated his skills, taking him with the third round of the Global T20 draft

Weighting – Recency bias

The evolution of batting in T20 has often held Gayle as the ultimate batsman. In his early thirties, starting slowly, sometimes taking an entire over or more before getting off the mark, and hoping to avoid an early dismissal, he would eventually, in a flurry of fours and sixes, unleash. But eventually age caught him. The failures of Royal Challengers Bangalore in 2017 sometimes seemed directly tied to Gayle’s own struggles. Freddie Wilde has written about how the West Indian has since reinvented himself as a more mortal batsman, with a more conventional style

Chart from  CricViz  showing a sharp uptick in Gayle’s ‘running ball percentage’ in the latest CPL

Chart from CricViz showing a sharp uptick in Gayle’s ‘running ball percentage’ in the latest CPL

Using his career average may not be the best representation of Chris Gayle’s current value. Whilst still good, he has clearly declined from his universe-bossing peak. There are a few potential alternatives to using the career average, none of them are perfect. The truth is that in T20, due to the high degree of variance, more data almost always beats less, but more recent, data. My compromise is use all the data available but to upweight more recent performances

Regression to the mean

Washington Sundar had a phenomenal debut season in the IPL with Rising Pune SuperGiant. He had the second-best economy in the league (min. 15 overs). He became the youngest player to win a man of the match award for his wickets against Mumbai Indians in the qualifier for the final. And he was only 17. He is set to become a bowling superstar

However, it would be foolish to have 100% confidence in Sundar just yet. We only have 10 data points for Sundar’s entire T20 career – not enough to have confidence that he will continue at such a high level. This is another application regression to the mean – the idea that exceptionally low or exceptionally high performance is unsustainable in the long term

To account for this, I apply a simple formula that is based on the number of games available for each player to pull their estimated value back towards average. Without applying the formula, I have Sundar rated as 3.25 runs above average, similar to Imad Wasim or Shakib Al Hasan. After applying the formula, his rating falls to 2.18 runs above average, similar to Amit Misra.

All that being said, Washington Sundar is still extremely young and would be expected to improve, as a talented young spinner. Maybe we have Rising Pune Supergiant ready-made superstar, maybe they have a one-hit wonder, or most likely, they have an excellent bowler who has yet to reach the height of his powers


Indeed, players develop and players age. Young players tend to get better and older players tend to get worse. Unfortunately, every individual is different and players do not all follow the exact same ageing curve. This makes it extremely difficult to adjust for age. Looking at the last few seasons of data, rather than an entire career’s worth, often gives a more credible valuation than using a universal ageing curve

This has not stopped me investigating this phenomenon and constructing countless ageing curves. Here is my first post where I document the basic approach in more detail, and here is another pointing out some of the approach’s flaws. And here are some of the insights from my work:

  • Batsman have a short period of improvement in their teens and early twenties, enjoy a long peak until their early thirties and then start to decline
  • Bowlers, on the other hand, seem to continually get better with the bat and produce increasingly (much) more value with their end-of-innings cameos
  • With bowling, spinners enjoy a similar trajectory as batsman. They improve as they learn the trade early in their careers and then start to decline, potentially due to reduced motivation or the batsmen learning their secrets
  • And finally, pace bowlers get worse. No matter how I cut the analysis, pace bowlers seemed to demonstrate a strong, practically linear, downward trend. Injuries will factor into this and it obviously doesn’t affect all bowlers equally
Age 1.PNG
Age 2.PNG
Age 3.PNG
Age 4.PNG

One potential implication from these ageing curves is that an older pace bowler with a history of top-level performance is a more reliable investment than a younger bowler with more upside. But I would be cautious in drawing too many hard conclusions. The reality is that player performance changes over time, unpredictably, and this further highlights the prudence to weight analysis towards more recent performances

Part IV