The ATG Teams General arguing/discussing thread

[stephen]

I remember some friends and I getting massive turn by copying Murali's action but obviously chucking it. It was a lot of fun but you can see why people thought he was a chucker. But yeah it definitely was much closer to Warne in execution than to Harbhajan.

 

[mr_mister]

Hmm I’d put Verity in that class too. Still counts as an off spinner despite being leftie right?

Maybe Tayfield too, ie one of the most forgotten players ever

 

[mr_mister]

And it's interesting how much easier wronguns are to bowl than leg breaks

 

[Spark]

I remember some friends and I getting massive turn by copying Murali's action but obviously chucking it. It was a lot of fun but you can see why people thought he was a chucker. But yeah it definitely was much closer to Warne in execution than to Harbhajan.

I think every junior cricketer who grew up in the mid-00s (including me) tried that, or had a teammate who tried it. It's certainly fun getting it to go absolutely sideways by basically reverse flicking it using your wrist (and elbow).

 

[Migara]

My point is - leg-spin is harder to bowl than off-spin. Jeepers.

My point is orthodox off spin may be but not unorthodox off spin.

 

[Migara]

I think every junior cricketer who grew up in the mid-00s (including me) tried that, or had a teammate who tried it. It's certainly fun getting it to go absolutely sideways by basically reverse flicking it using your wrist (and elbow).

And extra fun when you do it once in a while in the middle of a volley of orthodox off breaks.

 

[Spark]

I've never done it on small slices and I only use stats like this when a player has a decent sample size. Somewhere in the vicinity of ten tests played in a country or against an opponent is good.

I fundamentally disagree. The error bars on a 20 innings sample are massive.

 

[Red]

I fundamentally disagree. The error bars on a 20 innings sample are massive.

Error bars? Okay.

2 or 3 series in or against another country is plenty to make a judgement.

 

[Spark]

Would you make a definitive judgment on a batsman whose career was 10 Tests long? Of course not, you'd say that it's far too soon to really tell. Why it's considered fine to make conclusive judgments purely on statistical grounds after 20 innings, which is an objectively miniscule sample size, is beyond me.

Even after 200 innings you can easily have a +/- of about 3-5 runs on a batsman's batting average IIRC. You'd expect that to be an order of magnitude higher after just 20. There's just very little meaningful to be gleaned just by looking at an average of, say, 30, when the +/- on that average could easily be upwards of 50%.

EDIT: This is some of the testing I did some time back on this exact question, and why I have become increasingly sceptical of the precision of batting average as a measure of quality.



This is after 200 innings, and the "underlying" average should be 48.95. Look at how much variation there is in the "actual" batting averages after 200 completed innings, a 10-run spread.

(I gave Cribb this Mathematica notebook a while back hoping he would look at it. Hopefully this post prods him into doing so )

 

[OverratedSanity]

I think you can get a decent idea of how a batsman did across 20 innings. Just don't only use average. People also compare averages of two batsmen who played two different series against said team years apart under completely different circumstances. I'm not sure how you're meant to compare the two. But people go oh well, a averaged more than b so he must've done better. It's kind of silly imo.

 

[honestbharani]

I agree with your broader premise, Sparky but the way test cricket is, you usually are doing well if you get 10-15 tests in a foreign country. Assuming a career of 100 tests (which is still much more than an average career), I feel 10 or so tests in an away location is practically all you get to judge how well anyone went there.

 

[Spark]

I think you can get a decent idea of how a batsman did across 20 innings. Just don't only use average. People also compare averages of two batsmen who played two different series against said team years apart under completely different circumstances. I'm not sure how you're meant to compare the two. But people go oh well, a averaged more than b so he must've done better. It's kind of silly imo.

I agree with your broader premise, Sparky but the way test cricket is, you usually are doing well if you get 10-15 tests in a foreign country. Assuming a career of 100 tests (which is still much more than an average career), I feel 10 or so tests in an away location is practically all you get to judge how well anyone went there.

To be clear I'm not saying you can't make a definitive judgment after 10 Tests. But you should not be doing it purely on statistical grounds, i.e. using batting average. You can, and should, be making arguments based on "mechanistic" grounds, i.e. technique against certain bowling and on certain pitches, temperament in unfamiliar conditions, performance under pressure (which you usually are away from home) etc etc. Not batting average.

EDIT: And yes this is even getting before the massive differences in pitch conditions, bowling attacks, general form, match situations, series situations etc etc which are just completely ignored by a direct average vs average comparison.

The way I think about it this. Batting average is very good as the zeroth order term in your expansion. Maybe even first order. But the higher order corrections from all ^^^ are really quite big and quite important, and ignoring them is not defensible as a statistical exercise or a cricket-watching exercise.

 

[Daemon]

Disagree Spark. I can make a definitive judgement based on my 2019 batting average of 6.3 that I’m a rubbish batsman

 

[Prince EWS]

Disagree Spark. I can make a definitive judgement based on my 2019 batting average of 6.3 that I’m a rubbish batsman

That's because even the upper bound of the error range would make you rubbish tbf.

 

[Daemon]

Hey, averaging 10 was my new years resolution

 

[Burgey]

You should play tennis mate.

 

[honestbharani]

To be clear I'm not saying you can't make a definitive judgment after 10 Tests. But you should not be doing it purely on statistical grounds, i.e. using batting average. You can, and should, be making arguments based on "mechanistic" grounds, i.e. technique against certain bowling and on certain pitches, temperament in unfamiliar conditions, performance under pressure (which you usually are away from home) etc etc. Not batting average.

EDIT: And yes this is even getting before the massive differences in pitch conditions, bowling attacks, general form, match situations, series situations etc etc which are just completely ignored by a direct average vs average comparison.

The way I think about it this. Batting average is very good as the zeroth order term in your expansion. Maybe even first order. But the higher order corrections from all ^^^ are really quite big and quite important, and ignoring them is not defensible as a statistical exercise or a cricket-watching exercise.

Well, unless you are saying we should not use averages at all to compare batsmen, I am not sure how else we can discuss this. And to be sure, I am not comparing players I did not see when I am bringing up their stats like folks here with Sobers' etc. If I am talking about Ponting in India or Lara in NZ, I am talking as someone who has actually seen them play there and when that part is true, I do think the stats can be brought up as supporting evidence as long as the context is there.

Anyways I get the feeling that you and I are mostly on the same page on this and are just splitting hairs.

 

[Coronis]

If you guys had to choose say 5 cricketers, based on their impact on the game of cricket for any period you want to choose, who would they be?

 

[Spark]

Well, unless you are saying we should not use averages at all to compare batsmen, I am not sure how else we can discuss this. And to be sure, I am not comparing players I did not see when I am bringing up their stats like folks here with Sobers' etc. If I am talking about Ponting in India or Lara in NZ, I am talking as someone who has actually seen them play there and when that part is true, I do think the stats can be brought up as supporting evidence as long as the context is there.

Anyways I get the feeling that you and I are mostly on the same page on this and are just splitting hairs.

This is not what I'm saying at all. I'm saying that career average over a big big sample size (100-200 innings) is a good starting point for analysis. But it should be by no means the final word because of the significant amount of statistical uncertainty and lack of corrective factors from context etc (many of which are tackled by projects such as DoG's and Cribb's).

That's for an entire career. Cut the number of innings you're considering by a full order of magnitude and the uncertainty will blow up by a commensurate amount. In my view that renders batting an average an extremely poor metric of, well, anything. That doesn't mean you can't say anything about a 20 innings sample, indeed you might be able to say a great deal, but you should be going beyond batting average, well beyond, to do so. Ponting in India is a decent example of this: anyone who knows anything about cricket and watched those games would see that he had fatal issues with Harbhajan in Indian conditions, and that should be very much held against him given the importance of Indian tours and the importance of Ponting in all those batting lineups. But that's not a "well he averaged x in y Tests" argument, which is my point.

 

[srbhkshk]

If you guys had to choose say 5 cricketers, based on their impact on the game of cricket for any period you want to choose, who would they be?

Bradman , Sobers , Marshall , Imran , Tendulkar.