Career Impact – the Final Ratings

Career Impact – the Final Ratings

Levelling the Field

Eight years ago I posted my first feature for CricketWeb, Series Points; Series Points was my first effort at rating all Test players regardless of discipline, so that batsmen, bowlers and wicket-keepers could be compared on a level playing field based on everything they do.

The basic idea was that the points awarded by the ICC Test team rating system could be allocated to all of the participating players based on their performance with the bat, ball and in the field. It was first necessary to score each Test going back to 1877 as the ICC team ratings were only introduced in 2003. Once that was done, the relevant team totals could then be shared out between the players; producing those numbers for all Test series took a long time – in fact much of hte work for these studies is collect together the data then converting to a format which can be utilised. Over a series of articles I posted the leading players based on this rating system for each decade.

The Golden Age Chimera

After these ratings were combined to produce an all-time ranking, it became apparent that early bowlers were being favoured, since seven of the top 20 of all time were bowlers who began their careers in the 19th century. While I believe it’s possible that bowlers may contribute more to success than batsmen, presuming we have a way to compare players overall as I was trying to do with Series Points, I still found it difficult to accept that the first 130 or so Tests played produced seven bowlers deserving of being rated in the top 20 players of all time (and five of the top 13), while the next 2,000 Tests produced only 13 players of that calibre.

Understand that I didn’t care who the bowlers were, rather I cared that a certain group of players seemed to be unfairly rated, i.e. the group identified as 19th century bowlers. So some investigation was necessary to see if there was a reason why this group was being favoured; if not, then it would mean these players were in fact the best of all time. It is possible that those bowlers do have more actual value than later players, and it turns out that was the case due to increased opportunity.

If we look at the percentage of overs bowled by the opening pair through history, we find that in the 19th century they enjoyed 55% of the bowler share, whereas in later times it was as low as 38% – this should be accounted for. So a sliding adjustment was applied from 1877 to 1914. The effect of this adjustment was that after its application only two early era bowlers remained in the top 20; this is still a high percentage representation, but less extreme than 35% as before. I discussed this issue in detail in the feature Series Points – Slight Return.

Critique of Impure Reason

Still, although we now had a method of rating players regardless of discipline, rather than being restricted to having to compare only batsmen with batsmen and bowlers with bowlers, e.g. as the ICC player ratings do, there were some aspects of Series Points with which I wasn’t particularly satisfied. Series Points is a reasonable measure of a player’s average contribution across a series, and by extension his Test career; on the positive side it is context-independent, as the players contribution is not impacted by how well the rest of the team did. And as mentioned, it is also discipline-independent. But, Series Points doesn’t measure a player’s actual contribution in any match. It’s possible, though highly unlikely, that a player performed averagely in team wins but exceptionally in team losses. So his ACTUAL contribution to wins would be incorrectly represented. And whether or not we think this matters, and notwithstanding the kind comments of CricketWeb forum members, nevertheless it still scorched my slippers.

Match Impact – the End of an Era Study

So I came up with Match Impact. Match Impact is based on a player’s actual impact in terms of his contribution to his team’s chances of winning by looking at his contribution to the defining events of an innings – batting partnerships, wickets, dismissals. This impact is measured by a comparison of the win probability of a team prior to the partnership or wicket, and the win probability after it. I’ve discussed the method associated with this rating in a number of previous features, such as Win Probability and Schrodinger’s Bat, and several articles in between, but basically a batting team’s win probability increases as a partnership advances, while a bowing team’s win probability increases with the fall of a wicket.

As with Series Points, there was a great deal of ground work to do before I could start the interesting part of rating the players. In order to determine the win probability of a team at any given point in a match it was first necessary to convert all of the falls of wickets of the 2,200-odd Test matches into a win probability. So for example if a team was 150/5 in the fourth innings chasing a target of 230, what was their win probability? Or 75/1 chasing 295? Or 250/8 in the first innings? Once the FOW database was completed it was then possible to find the win probability at any point in a Test by looking at all other matches where the same situation occurred and checking how many ended up in a win, draw or loss for the batting team.

Then it was necessary to convert all of the Test match details into partnership and wicket events, with before and after win probabilities and identify the players involved. At that point, I could collect each match impact event for all players then total them. That work was finally completed last week after almost exactly three years. The final table of career impact ratings for each player/match has over four million data entries – that’s a lot of data and the difficult part is making sense of it.

How Big is an Elephant?

And therein lies the problem. Just as we have an issue with ranking batsmen on aggregates, which would show Sachin Tendulkar at number one, as compared to averages, which would show Don Bradman top, the same “how big is an elephant” quandary exists here. There are many different ways to measure the match impact elephant, such as career impact totals, career impact averages per match, moving average over five Tests, time spent with moving average over a given threshold value, as well as all of the above adjusted for opposition strength, or ratings in matches against only good quality opponents, etc.

Deciding which measure to use was discussed in detail in the Schrodinger’s Bat feature (see link above), where I proposed using the geometric mean. But for now I’ve decided just to post the ratings based on each of the above categories rather than attempting to combine them using a method which may or may not be the correct one. First, to help illustrate a number of interesting things which I’ve discovered through this study, I’m going to call on the great minds of the CricketWeb forum and discuss the various ratings of the CW top 50 (revision 1), which allows me to include players from all eras as exemplars.

Ladies and Gentlemen, the CW50, v1.0

The CW50 v1.0, to refresh everyone’s memory (and also confirm Bradman was not in fact No. 8) is reproduced below:-

1 Don Bradman
2 WG Grace
3 Garry Sobers
4 Sachin Tendulkar
5 Jack Hobbs
6 Imran Khan
7 Shane Warne
8 Viv Richards
9 Malcolm Marshall
10 Richard Hadlee
11 SF Barnes
12 Muttiah Muralitharan
13 Glenn McGrath
14 Keith Miller
15 Brian Lara
16 Dennis Lillee
17 Wally Hammond
18 Adam Gilchrist
19 Curtly Ambrose
20 Ian Botham
21 Len Hutton
22 George Headley
23 Wasim Akram
24 Greg Chappell
25 Ricky Ponting
26 Jacques Kallis
27 Graeme Pollock
28 Bill O’Reilly
29 Sunil Gavaskar
30 Victor Trumper
31 Fred Trueman
32 Steve Waugh
33 Herbert Sutcliffe
34 Michael Holding
35 Kapil Dev
36 Shaun Pollock
37 Waqar Younis
38 Allan Border
39 Denis Compton
40 Ray Lindwall
41 Harold Larwood
42 Fred Spofforth
43 Barry Richards
44 Jim Laker
45 Andrew Flintoff
46 Allan Donald
47 Joel Garner
48 George Lohmann
49 Hedley Verity
50 Douglas Jardine

Choosing Their Battles

First, one of the more interesting things to come out of this study was resolving each player’s games played as regards strength of opposition. Graeme Pollock had the most tough matches of the CW50, with 96% of his matches coming against opposition with an ICC rating of over 100, which makes his very high average that much more impressive. George Lohmann meanwhile did not play against any teams with a rating over 100 – South Africa were very much minnows while Australia were not at their best during Lohmann’s time. Brian Lara also had a tough time of it, with almost 92% of his matches coming against tougher opposition – it might be interesting to perform a parallel study with adjustments of traditional statistics based on strength of opposition. I had previously thought that these things probably come out in the wash over a whole career, but with Lara facing such a high percentage as compared to say Tendulkar, who faced less than 70%, that may not be the case. Murali and Warne, same thing – Warne faced almost 80% higher quality opposition, whereas Murali faced less than 62%. Douglas Jardine, whose representation in the CW50 is based more on his leadership qualities, nevertheless has a pretty impressive Test batting average of 48, but this must be tempered with the fact that only a third of his opponents were what we are designating here as good quality.

Raising Their Game

Another aspect I want to highlight is players raising their game against the better sides – Joel Garner is the king of the CW50 in this regard, as his impact per match is almost twice as high as in matches against sides rating less than 100. Fred Trueman is next, with an 80% increase. At the other end of the scale and other than the aforementioned Lohmann, surprisingly Victor Trumper’s good quality impact score was only 45% of his lower quality average, while Fred Spofforth’s was just 22%. One of the more interesting anomalies with this measure is that the highest values, at least among the CW50, are almost all fast bowlers, e.g. Garner, Trueman, Donald, Kapil Dev, Imran, Lindwall are at the top. However, this may just be an artifact of the small sample size; a quick check of the all time list shows that, of the top 20 highest game-raisers, seven were batsmen, one of whom, Eddie Paynter, has the joint highest percentage at 204%, tied with Hugh Tayfield, while eight were fast bowlers.

Scaling the Highest Peaks

One measure I have utilized in previous match impact features is a moving average over five Tests (MA5) to compare impact visually (see Schrodinger’s Bat for a comparison of Gooch and Gower along these lines). The player who raised his game the best was Andrew Flintoff, whose peak rating improves from 41.0 to 48.3 against good quality opposition – that peak as you might expect was achieved during the 2005 Ashes series. Barnes’ peak meanwhile reduces from 50.9 (the highest) to 39.9 (tenth of the CW50) when opposition is taken into account.

In the next feature I’ll summarise the all-time ratings for all of the above impact measures.