New Feature: Batting Medians

[chasingthedon]

Having never been a fan of batting averages, here Dave proposes a possible alternative.

Cricket Web - Features: Batting Medians

 

[Joao]

So there you have it, Shane Watson > Viv, Sanga, Hammond and Hutton

 

[Migara]

I see a major blip in that analysis. This basically treats a long innings as a probability innings, which is not the case.

 

[weldone]

I see a major blip in that analysis. This basically treats a long innings as a probability innings, which is not the case.

Agreed. There are some cases where high numbers can be ignored as outliers, and hence median becomes a better measure than mean. But batsmanship is not one of those cases.

If Brian Lara scored 8 (9?) double hundreds and a bucket-full of sigle-digit scores that doesn't make those double centuries fluke only because most of the times he was out for low scores.

 

[weldone]

And the first example of Bannerman is an outright pathetic attempt to show that median can be a better measure than mean. There can be no measure for such 'SMALL SAMPLE SIZE'.

 

[miscer]

that's ridiculous because then u are under appreciating batsmen who score giant scores now and then (lara). Maybe if u want u can have batting average + standard deviation.

 

[fredfertang]

I'm no statistician and I have always assumed that batting averages are the time honoured way of measuring a batsman's worth as much for the fact that even I can readily understand and calculate them as for how accurately they enable comparisons between individual batsmen to be made.

In particular I too struggle with the idea that Bradman is as much as 50% and more better than all comers.

A graphic illustration of how poor a measure averages can be, that this analysis sorts out, is Bill Johnston's achievement in 1953 when, over as many as 17 innings, he averaged 102 (against a career average of 12)

Thus I really enjoyed reading this interesting and thought provoking piece and neither the fact that it too may have flaws, nor that I don't suppose for one moment that batting medians are going to become a part of the graphic that TV producers use, detracts from that.

 

[weldone]

...Bill Johnston's achievement in 1953 when, over as many as 17 innings, he averaged 102 (against a career average of 12)
...

Fred please understand when we talk about batting averages, number of innings in no way corresponds to the sample size. The sample size is 'number of dismissals'. So, Bill Johnston's 1953 might have 17 innings, but should have much lesser number of dismissals - thereby making the sample size insignificant for any analysis.

 

[weldone]

that's ridiculous because then u are under appreciating batsmen who score giant scores now and then (lara). Maybe if u want u can have batting average + standard deviation.

Exactly. Doing some analysis based on avg and stdev should be fine. avg would indicate how they performed on average, and stdev would indicate how consistent they were.

But median is in no way a good measure when we are talking about batsmanship.

 

[fredfertang]

Fred please understand when we talk about batting averages, number of innings in no way corresponds to the sample size. The sample size is 'number of dismissals'. So, Bill Johnston's 1953 might have 17 innings, but should have much lesser number of dismissals - thereby making the sample size insignificant for any analysis.

I do appreciate that, and towards the end of that tour the whole Australian side were "conspiring" to try and bring about a situation where Johnston averaged 100, so it was all artificial anyway - all I was really trying to say was that I thoroughly enjoyed the article, it got me thinking and I am grateful to Dave for putting in the hard yards to come up with it.

 

[ankitj]

It's interesting thing to look at, but like others, I won't rate players based on medians.

 

[Zinzan]

Agree, interesting stuff. It really is an underrated & underused statistic that I've never given much thought to

 

[Adamc]

The median just says, assuming a reasonable sample size, that there is an approximately 50% chance that the batsman will exceed that score. It doesn't take into account by how much the batsman is likely to exceed their median score. As such it might be a better measure of reliability, but it isn't a better measure of ability. That someone like Katich has a higher median than say Lara and Tendulkar is a good example of this.

 

[smash84]

The median just says, assuming a reasonable sample size, that there is an approximately 50% chance that the batsman will exceed that score. It doesn't take into account by how much the batsman is likely to exceed their median score. As such it might be a better measure of reliability, but it isn't a better measure of ability. That someone like Katich has a higher median than say Lara and Tendulkar is a good example of this.

 

[Adamc]

What?

 

[weldone]

It's interesting thing to look at, but like others, I won't rate players based on medians.

The median just says, assuming a reasonable sample size, that there is an approximately 50% chance that the batsman will exceed that score. It doesn't take into account by how much the batsman is likely to exceed their median score. As such it might be a better measure of reliability, but it isn't a better measure of ability. That someone like Katich has a higher median than say Lara and Tendulkar is a good example of this.

awta

 

[Migara]

Better idea would be a p25 - p75 value range expressed with average

 

[nick-o]

That was a very interesting read.

I'd be interested to see how it could be used to assess peaks within a batsman's career. Basically, to establish a baseline -- say 20 innings, or 20 matches -- and compare peak medians of batsmen whose overall figures are similar.

It would probably only work as a way of comparing contemporaries, but it would be interesting to see the peak medians over a baseline period for, for example, Viv, Miandad and Chappell, or for Lara, Tendulkar, Ponting and Kallis.

 

[Howe_zat]

An interesting read, Dave . As Adamc says, it seems to tells you more about the nature of a batsman's record rather than their overall value.

The main thing you can't get around here, in my veiw, is that by using medians you essentially eliminate the outliers - but batting is all about the outliers. Those are the innings we remember and celebrate, after all.

 

[ankitj]

AdamC is absolutely right mate