As someone who loves sport (all sport, recently, even the snooker and the golf) and who enjoys data, cricket stats have a big pull on me. It’s the most obvious ‘English sport’ equivalent to baseball, the game that gave the world Moneyball and sabremetrics (although not in that order).
As a newbie, there’s a lot I don’t know – although it’s cool to see some relative crossover with football with Cricket Savant and CricViz creating Expected Goals-esque models for runs and wickets. However, articles about adjusted strike rates/batter ratings by John Robertson and S Rajesh piqued my interest.
They vary in complexity, but the point of both is that circumstance (era, pitch, time of match a batter plays) affect the value of their averages and strike rate. Robertson’s focuses on Twenty20 and Rajesh’s on ODIs, which fitted nicely for me personally as the format of cricket I’ve followed most closely (and still only quite loosely) is Test cricket.
The aims of Test cricket are different to the aims of the one-day formats, with the ability to just stay at the crease a valued one, and one which – even by my limited knowledge – batters like Paul Collingwood and Shivnarine Chanderpaul built their careers on. Put more simply, I guess, one-day cricket is about getting wickets and runs; Test cricket is about getting wickets and runs, but also (or mainly) about using your time wisely.
Adjusted Batting rating
Therefore, to Rajesh’s adjusted batting rating, which utilises averages and strike rates, I added the number of balls a batter faced per innings. For each of these categories, batters are given a ratio against the average for that category in the time they played* in order to adjust for era.
*NB: With only the stats easily available from ESPN cricinfo, this is a bit of a fudge. The ‘era adjustment’ creates an average for every player who played part of their career within the timeframe of Player X’s career. Eg Player X played from 1990-2005, any player who played part of their career in that frame would be counted. Number of balls faced is only available from the 1980s or so onwards, and this method also fails to account for quality of opposition, which I believe Rajesh’s does do.
The number itself is constructed by taking their ratio, compared to the average, of their averages and strike rates, multiplying these together, and multiplying this by 100 to make it more easily consumable. Eg, if a player’s average and strike rate were both double the average for their era, their ratios would be 2.0 for each, therefore 2×2=4, multiply by 100, 400. To make the old and new ratings comparable, and make the ratings closer to 100 being a top score, I’ve divided the old ratings by 2.25 (1.5 times better in each category) and new ratings by 3.375 (the same reasoning).
The Actual Batting Lists
With my version of Rajesh’s ‘ODI’ adjusted batting rating, the top 20 list is as follows.*
*As balls faced is only available from the 80s onwards, players’ careers are cut at the 1980-mark
|Player||Nation||Start Year||End Year||Old’ Batting Rating|
|Q de Kock||SA||2014||2017||106.36|
Add in the balls per innings measure and the top 20 shakes up, with some batters more than others being affected. Here’s the new top 20, with the players’ ranks in both ratings (out of about 500 batters to play more than 25 innings since 1980), along with the members of the previous ranking’s top 10 who got bumped down more significantly.
(NB: As this rating has an extra factor, the scores are bigger – I didn’t bother to normalise them both to 100 which I could, and perhaps should, have done, and may well do when I get time)
|Player||Nation||Start Year||End Year||Batting Rating rank||New rating rank||‘Old’ Batting Rating||‘New’ Batting Rating|
|AB de Villiers||SA||2004||2016||36.00||18.00||87.42||96.60|
|Q de Kock||SA||2014||2017||8.00||44.00||106.36||83.38|
I will admit to not knowing a lot about most of these players. For English players, who I know better, Joe Root moves from 33 to 21 and Kevin Pietersen makes the reverse journey, from 24 to 32. Alastair Cook moves up from 108 in the first rating to 47 in the second and Geoffrey Boycott, who apparently played in 46 innings in the part of his career within the date parameters, shoots up from 195 to 88. Ian Botham and Andrew Flintoff both make reverse journeys, from 76 and 96 to 183 and 196 in the list.
Rajesh also talks in his part of the Cricket Monthly article about doing a similar rating for ODI bowling, using average and economy. This gets at who gives up the fewest runs, but it strikes me that in Test cricket it’s also useful to be able to take wickets quickly, otherwise you’ll end up drawing every match you play.
The same fudged method for era-adjustment applies here as it did for the batter rating, as does the problem about quality of opposition faced, and so my version of Rajesh’s ‘ODI’ adjusted bowling rating has this for the top 20.
|Player||Nation||Start year||End year||Old’ Bowler Rating|
|Sir RJ Hadlee||NZ||1980||1990||83.89|
Again, while I recognise a bunch of these names, I can’t speak to their world-beating (or not) quality, so onto the ratings with strike rate – how many balls bowled per wicket – added in.
|Player||Nation||Start year||End year||Bowling Rating Rank||New Rating Rank||Old’ Bowler Rating||New’ Bowler Rating|
|Sir RJ Hadlee||NZ||1980||1990||5.00||2.00||83.89||86.43|
This seems more stable than the changes to the batting rating, although the changes generally seem to favour quicks/hinder spinners. The biggest mover in the new top 20 is Waqar Younis, whose strike rate of a wicket every 43.4 balls is only bettered (in this timeframe and having bowled more than 500 overs) by Shane Bond and Dale Steyn.
Speaking of bowlers I know better, Anil Kumble and Graeme Swann both move down a few places while players like Darren Gough and Matthew Hoggard bound upwards from 108 and 121 to 51 and 78 on the list.
I’d obviously like to replicate Rajesh’s model more closely to properly adjust for era and opposition strength. There was also an interesting article from Opta aaaages ago about who took the most difficult wickets and allowed more or fewer runs than expected based on the quality of batsmen they faced.
The bowler ratings here had a similar feature to Rajesh’s, in that the values are lower than the batter ratings. My hypothesis is that everyone in the team bats (more often than not), whereas only a select few bowl with any regularity – bowlers bat, but batters don’t bowl. If the averages for batting ratings were calculated with all of the specialist bowlers taken out, leaving only specialist batters and all-rounders, then I imagine the batting rating values would go down a little due to the averages having changed (specialist bowlers can still have a batting rating, of course, they’ll just be judged against the ‘non-bowlers’ average). Unfortunately, I don’t have a way of knowing who the specialist bowlers are in the data in order to take them out of the average.
Again from Opta, an article about England’s batting conundrum in 2012 featured stats on the number of dot balls and swings and misses etc, which would be lovely to look into in more depth as presumably players who miss shots (or edge them) are more likely to get out, and might just have been getting lucky to stay in.
Given that both of the articles just mentioned are 5 years old, I could be waaaay behind the curve in terms of where cricket analytics – or cricalytics, as I will now call it – is right now. But this was kinda interesting to do, and I’m deeming this a siren call to cricalytics followers to let me follow you and absorb your knowledge. I’m also interested in the stats side of rugby, which seems little explored even by Opta, so if you know anything about that then please holler at me.
Like what you read? Got any thoughts? Leave a comment below