Your Time is Gonna Come (Should We Discount the Golden Oldies?)

Hats off to (Rob) Eastaway

The ICC player ratings were developed in 1986-87, actually at the suggestion of Ted Dexter, who was fascinated by the world golf rankings and which story is entertainingly told by Rob Eastaway on his website . Rob is a rare thing, a cool mathematician, who has written a number of books, including the very popular “What’s a Googly?”, famously gifted to President George by John Major at Camp David. Rob also has a podcast, Puzzling Maths, having been a maths teacher for much of his adult life.

One of the most interesting parts of Rob’s website is the link to his interview with Jonathan Agnew on TMS. As mentioned above, Eastaway was approached by Ted Dexter in 1986 after the former had written in The Cricketer about a computer simulation of a cricket match developed by his friend, Gordon Vince. Dexter was keen to develop a player rating system, and this would be first published as the Deloitte Ratings of Test Cricketers in June of that year.

I’ve used the ICC player ratings extensively when putting together articles for Cricketweb (most significantly this one) and also as the basis of certain assessment categories used in the books Masterly Batting and Supreme Bowling, with Patrick Ferriday. One thing I noted was that, in general, modern players were rated higher than Golden Age players, and had always assumed that was because the modern era players were better.

However, it also seemed that some of the ratings of the older era players when compared with their peers were possibly not what I would have expected. I decided to contact Rob Eastaway and we had a series of emails where I discovered a number of features related to the scaling of the algorithm used to calculate the ratings that aren’t apparent from the ICC’s FAQ. I am grateful to Rob for his time and input.

In summary, the player’s calculated rating is discounted on a sliding scale from 40% to 70% of his calculated peak for the first ten innings, then 70% to 100% up to the 40th innings. So the maximum rate possible after ten innings would be 700 (based on a theoretical maximum of 1000) and after 25 innings the maximum possible would be 850. From 40 innings upwards, the maximum possible would be 1000, which puts Bradman’s all-time peak of 961 into perspective.

How many more times?

I understand, and accept, that if modern players were apportioned their full rating from scratch then the ratings would be unlikely to represent the general consensus ranking of the best players of all time, as it might be seen to over-estimate the value of those players that start out very hot for a short time.

However, this may be considered as unfair to the early era players, when far fewer Tests were played. The numbers of Tests played increased by more than 400% between 1920 and 1980, so it would seem some kind of consideration should be given to those early era players. For example, Joe Root played 29 Test innings in 2021 alone; WG Grace played his 29th innings in his 16th year of Test cricket, and Ranji played 24 innings total over his Test career.

So I set about recalculating the ratings of a number of players from around the Golden Age based on the information that Rob had shared with me. I applied an upscaling of each players’ progressive rating, this being based on the ratings graphs available online for each player, up to the 40th innings as noted earlier, thus allowing each player to keep 100% of their calculated rating from the first innings. Note that I’m not recalculating the peak value of the player, and in any case I don’t have access to the nuts and bolts of the algorithm, rather I’m simply removing the discounting applied within the first 40 innings.

Upon completion, I decided to try an experiment, so I asked my colleague and noted cricket historian, Martin Chandler, to rank nine players that played before WW1 or didn’t play many Tests, as shown below:

• WG Grace
• AG Steel
• Andrew Stoddart
• Tom Hayward
• KS Ranjitsinhji
• Warren Bardsley
• Ernest Tyldesley
• Eddie Paynter

I then compared Martin’s ranking of these nine batsmen to that of the ICC; The rankings of the nine are shown below (actually eight, as I forgot to include Steel in Martin’s list):

I found almost no correlation (0.048), between Martin’s ranking and the ICC. I then compared the peak rating of the players based on their unscaled ratings, and found this ranking had a strong correlation (+0.738) with Martin’s selection – this (admittedly small) sample supported my feeling that it’s possible that early era players are somewhat short-changed by the scaling back of players who played less than 40 innings.

What is and what should never be?

Please understand that I’m not trying to say that Martin’s rating should be held in higher regard than the ICC’s rating system (though Martin may do), however he is as well-read about cricket as anyone I know, and I was trying to gauge how well the scaled ICC ratings reflect our perception of where the players of yesteryear rank in the pantheon of cricket all-time greats.

For example, Ranji’s highest unscaled peak rating is 996 (current ICC peak is 689), as compared to Bradman’s all-time best 961 (though Bradman will also likely enjoy a higher rate once unscaled), and Grace (679) rates alongside Hobbs’ official rating with 948 (though again Hobbs’ unscaled rate may be higher than before). I plan to look at the unscaled ratings of post-Golden Age players’ in the next feature.

Below are the highest unscaled ratings achieved by a selection of early era players, alongside their peak ICC ratings; the selection focuses on those whose unscaled rating is significantly different from their current ICC peak rating. Note that, as a guide, I’ve also indicated more recent players whose peak ICC rating is around the same as the early era players’ unscaled and ICC peak ratings:

Nobody’s fault but mine

The rating of “Tip” Foster merits some discussion, as the ratings algorithm is supposed to cap at 1000, however this may be either because the 1000 cap is not applied after only one innings or because the system isn’t expecting a player’s first ever Test innings to be a world record high score (Foster made 287 on debut against Australia at the SCG in December, 1903). A third option is that I got the maths wrong, and now is probably a good time to add the comment that any errors are mine and are not related to the work done by Mr Eastaway and the good folks that produce the ICC Ratings. By the way, I wasn’t intending to single out Bevan Congdon in the table above, but for some reason a number of high-profile Golden Age players have ICC ratings that correspond to the New Zealander all-rounder’s bating rating.

Looking through the comparisons above, it is somewhat shocking to see someone like Stanley Jackson elevated from a rating comparable to New Zealander John Reid, as against an unscaled rating comparable to the legendary Jack Hobbs. Or Frank Iredale, who may not have been a Len Hutton with the bat but was possibly not a Chris Cairns either. But the most significant difference is surely that of “Plum” Warner, currently rated alongside Jacob Oram but whose unscaled rating of 972 is higher than Bradman’s, albeit after only two innings, though it was still as high as 946 after four innings.

What is particularly noteworthy is that some peaks were reached very early (Grace after his first match) and others after a significant amount of innings (Joe Darling 21 and Arthur Shrewsbury 24), which is the kind of disparity that the scaling is intended to address, so in the tables that follow I’ve shown how the rankings for pre-WW1 players change after varying numbers of minimum innings:

There are players here, e.g. Faulkner, Hill and Trumper, that were not discussed earlier, because their highest rating is not significantly impacted by removing the scaling factor, i.e. their ratings were already very high. Note that I left Hobbs out of this exercise as more than half of his Tests were played post-WW1. It is interesting to see how the rankings are changed by the varying cut-off points beyond which scaling is applied. Looking at how these ratings and rankings vary depending on the chosen threshold, there’s certainly a suggestion that England batsmen of the time peaked early and didn’t maintain the highest standard, though, once again, this is a small sample size.

Ramble on

Also, how good was Faulkner? Possibly the greatest all-rounder to play the game until Sobers came along, Faulkner was ranked number one for batting and number four for bowling during his Test career, and in this feature I concluded that he was the most all-round all-rounder in history. The ICC player ratings website doesn’t make it easy to check other all-rounders as there’s no best-ever rating, as there is for batsmen and bowlers, however there may be a bit of a quirk due to the war, as a search on the date-specific bowler ratings for the end of 1914 shows Faulkner at number one, which could be a combination of the gradual reduction of a player’s rating over time when he plays no matches, coupled with the fact that higher rated bowlers such as Sydney Barnes had dropped off the ratings altogether due to prolonged Test inactivity. Despite Faulkner being listed at number one, his best-ever ranking is still listed as number four.

Thank You

In conclusion, I believe the scaling is necessary for modern day cricketers that play in so many more Tests, however it’s possible that this same scaling is not really fair to the players of pre-World War One. I wonder how Ted Dexter would have felt to see his Sussex antecedent Ranji, elevated from being comparable to Sherwin Campbell to being rated alongside the best ever?

My original concern was that the scaling applied to ICC batting ratings was a little unfair to early era players, though I presume the same discounting is applied to bowlers, so a future piece may look at that too – and, in homage to Faulkner, maybe I should look at all-rounders too.

It is possible that the scaled ratings are a more accurate reflection than is the public’s perception of the true worth of earlier era players. This possibility notwithstanding, I thought it would be interesting nonetheless to apply the same process to all Test batsmen, so next time I’ll review post-WW1 players and re-evaluate them as if there was no scaling applied, when I’ll also look at who has the highest peak ever if different innings thresholds are applied – you might be surprised.