The Best Cricketer Ever

SF Barnes.

No, really, it\’s not Don Bradman. It\’s SF Barnes.

11 thoughts on “The Best Cricketer Ever”

  1. I’m probably taking this far too seriously, but:

    1 – The “Wickets/Match” factor is suspect – the strength of the other bowlers in the team will affect your value – so if you are a good bowler in a good bowling line-up, you’ll take an average number of wickets/match; a good bowler in a poor line-up will take an above-average number of wickets/match. There’s at most 20 wickets to go around (on average, 14.7 wickets are taken by the bowlers in any given side in a Test match – run-outs, not-outs, declarations, not needing to bat making up the remaining 5.3 that you’d expect) so it’s a zero-sum game between bowlers. Batsmen scoring centuries doesn’t prevent the other batsmen in the team scoring highly.
    This means that if you’re in a team with Shane Warne and Glenn McGrath, your W/M will be less than if you’re in a team with no-hoper bowlers – without your own ability changing at all.
    2. Conditions in the game have changed (no sticky wickets today, better ground preparation, rule changes, protective armour/helmets for batsmen), so the bowling averages can’t really be directly compared to the “average over the last 130 years”
    3. The “comparison with the norm” figure given relies on an arbitrary value of wickets/match anyway – and plugging in a spreadsheet shows that the comparative ERS/M vs the norm for two very good bowlers (Barnes and Lohmann) changes pretty noticeably for a change in that arbitrary value (at 3-30, Barnes leads Lohmann by 161 to 154; at 4-30, they’re level on 144)

    Anyway … plugging in the average bowling average* for the first 50 years of Test cricket for England and Australia only (using those two teams for the longest baseline), we have a bowling average of 25.4 for that era and those representative teams (for England alone, it’s 24.1)
    Assuming that the wickets (if all was average) would be shared between 4.5 bowlers (4 bowlers is a bit more common than 5 bowlers, but there are occasional part-time bowlers used), then 3.3 W/M should be average.

    With 3.3 – 25.4 instead of 3-30, we have:
    Barnes 124, Lohmann 123 (Best Australian bowler would be CTB Turner on 96 – excluding John Ferris**)

    So Barnes is slightly better than Lohmann out of the population of Australian and English bowlers of that era. But out of English bowlers only, using 3.3 – 24.1, they’re dead level again, on 114.5 each.

    *”average bowling average” looks clunky, but you know what I mean …
    ** John Ferris played only 9 tests (8 for Australia then one for England, so both sides can claim him) and got a 3-30 score of 165 and a 3.3-25.4 score of 130, beating Barnes and Lohmann both, but with a very small number of matches played)

    Personally, I’d prefer a bowling average multiplied by strike rate as a fair value (the lower the better) – a low strike rate implies that each ball has a better chance of taking a wicket (which is comparable to the Wickets/Match measure but not a zero sum game and dependant on the quality of the other bowlers in the team).
    An even better measure would be to derive average bowling averages and strike rates for given eras (and defining exactly what constitutes an era would be challenging and highly arguable between champions of various bowlers – shades of Gordon Brown defining economic cycles for his Golden Rule) and dividing the specific bowler’s average and strike rate by the era average and strike rate would provide a better comparison.

    Okay – I definitely have too much time on my hands this evening 🙂

  2. SF Barnes played many fewer Tests than he deserved, owing to the fact that he knew the authorities were twats, and refused to do what he was told.

    An average of 16 is still unbelievably cool, though.

  3. (Auther of the paper)

    Thanks very much to Andy Cooke for showing how this method of comparative analysis can be refined and made more specific.

    He is quite correct that the norm comparison is crucial to validity – but a direct comparison between two bowlers overcomes this.

    Naturally, any single summary statistic is going to represent a great simplification – but to be useful I think a bowlers measure needs to be in the same ‘currency’ as a batsman’s measure (ie. *runs*).

    AC’s suggestion of measuring bowling ability by bowling average (in runs conceded per wicket), multiplied by strike rate (in balls bowled per wicket) gives a measure in runs-and-balls per wicket which is somewhat hard to comprehend.

    I also think it is vital to include a measure like the number of wickets taken per match to measure a bowler’s contribution to the win. The ‘great’ bowlers almost all took more wickets per match than the merely-good bowlers.

    Indeed. taking a sufficient number of wickets is the *primary* job of a bowler in winning test matches. While there is a certain zero-sumness about this, in practice (for example) it doesn’t seem to matter – Jason Gillespie’s test record is a very good 3.6 wickets per match at 26. If he hadn’t been playing with Warne and McGrath he might have taken more wickets per match, but probably also at a higher average.

    In the end, the validity of ERS/M is not self-evident or deductive (neither is the validity of the batting average calculation) – but depends on using the measure in practice.

    ERS/M does not need to be a perfect quantitative comparative statistic – the batting average is not perfect, no summary statistic is perfect. It just needs to be more useful than the present situation.

  4. Bruce, Thanks for commenting. I’ve just had a thought on the strike-rate issue: Would it not be possible to examine the average number of balls bowled per match by a front-line bowler, and then use the strike-rate to obtain an independent number of wickets per innings (e.g. if a bowler sends down an average of 50 6-ball overs per match, that would be 300 balls – how many wickets should he take in 300 balls)?

    You’re correct in the direct comparison measure, of course.

    And, of course, I could well quibble with the validity of the batting average itself as is widely used – the conditions over the eras have changed (What would Bradman have scored in 1878-1898, for example, or in 1988-2008?) – so I’d also have the same arguments against someone bringing up the batting average for the first time.

  5. Andy – this sounds very interesting.

    Maybe you should write a letter-in-response to OR Insight, where the article was published – or try sending them a formal paper since the editor seems interested in these issues (if linked to a ‘business’ rationale)?

    To avoid boring those less entranced by the magic of cricket stats perhaps we could continue the conversation over at The Doosra?

    http://the-doosra.blogspot.com/2006_11_01_archive.html

  6. Splendid! About time SF received wider recognition.

    Leaving his test figures aside, his League analyses are even more amazing – as can be seen in Duckworth’s ‘S.F. Barnes, Master Bowler’. On one occasion playing for North Wales he tore through a touring Test side, taking 8 for 40, bowling 30 overs unchanged. He was over 50 at the time.
    Truly a Great Man.

Leave a Reply

Your email address will not be published. Required fields are marked *