Prince EWS
Global Moderator
I was just having a look at some of the overall stats the new StatsGuru function can give us and I came across this data.
Inspired by Goughy's recent Test match threads on this subject, I thought I'd give this a try. I have the advantage of programming knowledge for mass data operations (and copious amounts of time during my uni holidays as well) so I'll even be able to produce mass lists instead of just doing individual players eventually.
NOTE:
It should be noted that I don't intend on suggesting that my finding actually prove anything; we all know that raw averages and strike rates, even weighted, don't tell the full story about the importance and difficulty of different performances or the effect playing different roles within teams will have on data. However, I thought it'd be an interesting exercise and a way to keep myself occupied.
Anyway, it's often been noted that ODIs have changed massively over time and hence it has become increasingly difficult to compare players across eras. Whilst standardising the performances won't be perfect, it's as close as we're going to get and personally I'm interested to see what kind of results I get. I'm starting with batting for the time being but once I get all my formulae set out and save all my data it'll be relatively easy to do it with bowlers as well. Given the shorter history of ODIs and more mass data handling tools at my disposal, I'll actually be weighting performances by the par averages and strike rates in each year rather than each decade. EDIT: The period of 1971 to 1978 will be classified as one year as the small number of games played in those years has skewed some of the data for individual years. 1979 was the first time more than 20 ODIs were played in a calender year.
To explain the "standardisation" process for those who aren't familar with it, it's about finding par or average values for statistics and then weighting different years to standardise the data. For example, if you averaged 20 with the bat in an a year where the par average was 30, that'd be the equivalent to averaging 40 in a year where the par average was 60.. or indeed 10 in a year where the par average was 15. The same applies to strike rates.
The main question, IMO, is how to express the relationship between average and strike rate (or economy, for bowlers). In Test cricket your average represents your entire statistical value - scoring rate is largely unimportant as long as you score enough runs per dismissal or concede few enough runs per wicket. In ODIs however, the ability to score quickly or restrict the batsmen from doing so is arguably just as important as one's average. One method CricInfo writers generally use is simply multiplying a batsman's average by his strike index. For example, a batsman with an average of 40 and a strike rate of 50 would be deemed to have an "average" of 20. The fact that this method suggests that such a batsman would be as useful as one averaging 10 and striking at 200 tells me it is far from a perfect measure. However, these are extremes which I doubt will surface too often if at all. Hence, at this stage, I'll be figuring out each batsman's standardised average and standardised strike rate and then applying this formula to it. I'm quite open to ideas on how to implement this, however.
A few things to note about the data
Open to predictions, comments, abuse, suggestions, empty care cups etc until I provide more data.
Inspired by Goughy's recent Test match threads on this subject, I thought I'd give this a try. I have the advantage of programming knowledge for mass data operations (and copious amounts of time during my uni holidays as well) so I'll even be able to produce mass lists instead of just doing individual players eventually.
NOTE:
It should be noted that I don't intend on suggesting that my finding actually prove anything; we all know that raw averages and strike rates, even weighted, don't tell the full story about the importance and difficulty of different performances or the effect playing different roles within teams will have on data. However, I thought it'd be an interesting exercise and a way to keep myself occupied.
Anyway, it's often been noted that ODIs have changed massively over time and hence it has become increasingly difficult to compare players across eras. Whilst standardising the performances won't be perfect, it's as close as we're going to get and personally I'm interested to see what kind of results I get. I'm starting with batting for the time being but once I get all my formulae set out and save all my data it'll be relatively easy to do it with bowlers as well. Given the shorter history of ODIs and more mass data handling tools at my disposal, I'll actually be weighting performances by the par averages and strike rates in each year rather than each decade. EDIT: The period of 1971 to 1978 will be classified as one year as the small number of games played in those years has skewed some of the data for individual years. 1979 was the first time more than 20 ODIs were played in a calender year.
To explain the "standardisation" process for those who aren't familar with it, it's about finding par or average values for statistics and then weighting different years to standardise the data. For example, if you averaged 20 with the bat in an a year where the par average was 30, that'd be the equivalent to averaging 40 in a year where the par average was 60.. or indeed 10 in a year where the par average was 15. The same applies to strike rates.
The main question, IMO, is how to express the relationship between average and strike rate (or economy, for bowlers). In Test cricket your average represents your entire statistical value - scoring rate is largely unimportant as long as you score enough runs per dismissal or concede few enough runs per wicket. In ODIs however, the ability to score quickly or restrict the batsmen from doing so is arguably just as important as one's average. One method CricInfo writers generally use is simply multiplying a batsman's average by his strike index. For example, a batsman with an average of 40 and a strike rate of 50 would be deemed to have an "average" of 20. The fact that this method suggests that such a batsman would be as useful as one averaging 10 and striking at 200 tells me it is far from a perfect measure. However, these are extremes which I doubt will surface too often if at all. Hence, at this stage, I'll be figuring out each batsman's standardised average and standardised strike rate and then applying this formula to it. I'm quite open to ideas on how to implement this, however.
A few things to note about the data
- Only games involving two top 8 teams will be considered. Bangladesh, Zimbabwe and the associates are only likely to distort the data. This is harsh on a few quality Zimbabweans who won't feature at all however I plan to do special separate data for the likes of Streak and the Flowers just to see where they'd rate.
- Only data on the top seven batsmen (batting order) will be used for standardising purposes. Tailenders are largely irrelevant for this exercise so I thought I'd remove the possible chance of a big discrepancy in their performances in some years tainting the data.
- Players must have played 25 ODIs to be considered.
Open to predictions, comments, abuse, suggestions, empty care cups etc until I provide more data.
Last edited:
