cricrate: new cricket ratings website

[flibbertyjibber]

Also just from a quick overview, the English 97 batsmen weren't that great.. Not that this Aussie line-up is much better, but Smith was near a career high.

I'd take the England 97 batting side over the current Aussie one by a long way.

 

[marc71178]

These drop rates are coming from actual textual evidence. It's misassigned/missing very rarely. So these drop rates are facts. It's a fact that Steven Smith has dropped 27% of his Test catches. It's also a fact that he's only dropped 6% of his ODI chances - which is where his fielding reputation comes from. Same applies to Bravo.

But that's the problem. There's information missing and it's a very subjective area.

Also, how can you say something is missing rarely without watching the games? You can't, which makes the whole stat a load of rubbish.

 

[weldone]

I'd take the England 97 batting side over the current Aussie one by a long way.

ya nah

 

[viriya]

But that's the problem. There's information missing and it's a very subjective area.

Also, how can you say something is missing rarely without watching the games? You can't, which makes the whole stat a load of rubbish.

Because I've checked whole matches to validate accuracy.. That's almost like saying because you didn't see the batsmen score the runs, average is a rubbish stat. It's a fact that Smith dropped 27% of his Test chances.. it can possibly be slightly higher (say 28%), but it's not lower.

I'm using my eyes instead of spreadsheets..

 

[Dan]

Smith does drop more catches than you'd expect, tbh. Mostly because he anticipates so well and is able to fling himself at things mere mortals would get nowhere near.

There's myriad flaws to any fielding ratings; obviously it's going to be skewed against people who are offered a lot of difficult chances -- and are usually in those positions because they're the better fielders who are more likely to catch them!

 

[viriya]

Smith does drop more catches than you'd expect, tbh. Mostly because he anticipates so well and is able to fling himself at things mere mortals would get nowhere near.

There's myriad flaws to any fielding ratings; obviously it's going to be skewed against people who are offered a lot of difficult chances -- and are usually in those positions because they're the better fielders who are more likely to catch them!

The formula ignores difficult chances.. with Smith's Test chances, I went through every single one of them and they weren't difficult chances.

Relevant thread:
http://www.cricketweb.net/forum/cricket-chat/65351-fielding-statistics-new-approach.html

 

[OverratedSanity]

I don't know how you'd even get data from Mahela and Dravid's early careers for ball by ball "drops" and catch attampts.

 

[viriya]

I don't know how you'd even get data from Mahela and Dravid's early careers for ball by ball "drops" and catch attampts.

I don't - that's what I meant when I said it was unfair on them because I only have reliable data from 2006/07.

 

[Daemon]

Yeah, the numbers are fine. They seem pretty accurate. The problem is that they don't mean anything.

 

[viriya]

Yeah, the numbers are fine. They seem pretty accurate. The problem is that they don't mean anything.

I think if you just look at careers that started after 2007, this gives you a very accurate picture of who the best and worst fielders are. It also tells you that the average drop rate among all players is around 15%, and if you are below 10% you can be considered a great catcher, and above 25%, a poor catcher.

In terms of how much value fielding brings to the table, win shares gives some idea, it's not significant enough for a fielder generally to be picked for his fielding in Tests/ODIs, but an exceptional fielder probably can make a much larger impact in T20s - I haven't calculated win shares for that format yet. Currently working on some betting algorithms.

 

[hendrix]

I agree with Viriya that it's real data, but I disagree that the data is a particularly good indicator of the level of fielding.

Someone like Dhoni is going to be rewarded for not going for catches that he should be attempting. It punishes the athletic fielders.

It's probably only useful for comparing fielders at first slip where it's all about the hands and most opportunities don't involve much athleticism.

 

[viriya]

I agree with Viriya that it's real data, but I disagree that the data is a particularly good indicator of the level of fielding.

Someone like Dhoni is going to be rewarded for not going for catches that he should be attempting. It punishes the athletic fielders.

It's probably only useful for comparing fielders at first slip where it's all about the hands and most opportunities don't involve much athleticism.

Technically a fielder who goes for difficult chances and misses should be unaffected, but there is a possibility that it just looks like an easy drop just because he went for it I guess.

 

[BackFootPunch]

Really think this would need some kind of measure of difficulty or distance moved pre-catch to really make it useful.

 

[RossTaylorsBox]

Google's deep learning algorithm will soon make all these stats irrelevant, as well as putting dog heads on everyone's bodies.

 

[marc71178]

Really think this would need some kind of measure of difficulty or distance moved pre-catch to really make it useful.

Exactly my point - a drop is a subjective event, as much as a chance is.

 

[viriya]

Updated after that crazy test:

- Angelo moves up to #2, Sanga drops out of top 5, Chandimal jumps 29 spots to #28 in current Test batting rankings with a near top 100 batting performance (2775 vs #100 rated at 2817):
cricrate | Current Top Test Batsmen
- Herath returns to #1 in the current Test bowling rankings, Ashwin joins the top 10:
cricrate | Current Top Test Bowlers
- Herath with the #97 rated Test bowling performance:
cricrate | Best Test Bowling Performances
- Rahane with the #10 rated Test fielding performance:
cricrate | Best/Worst Test Fielding Performances

 

[TNT]

It just misses the top 100.. rated 2803 with the #100 rated 2808. But I agree it should be in the list. I'm going to make a change to increase the "bowling quality" factor which should push it higher (Laxman as well).

Its not a rating anymore, its a method of validating an opinion.

 

[viriya]

How would it be my rating system unless it is influenced by my opinion on what factors are important?

 

[Prince EWS]

How would it be my rating system unless it is influenced by my opinion on what factors are important?

There's a big difference between deciding how important each factor is relative to the other factors and and deciding what you'd like the final results to look like then tweaking the weightings if they don't match up.

The point should be to take the calculations/estimations you think you'd do personally better than any sort of algorithm (like determining how important each factor is) and then getting the algorithm to do the precise things a human brain isn't as good at (like actually applying those weightings consistently). If you're just trying to get a certain order of results you might as well just jot them down in order of personal preference and save yourself the maths.

I've got a ratings system that has changed a lot over time too, but I try not to let the actual results change my weightings, otherwise it'd be rather pointless.

 

[viriya]

There's a big difference between deciding how important each factor is relative to the other factors and and deciding what you'd like the final results to look like then tweaking the weightings if they don't match up.

The point should be to take the calculations/estimations you think you'd do personally better than any sort of algorithm (like determining how important each factor is) and then getting the algorithm to do the precise things a human brain isn't as good at (like actually applying those weightings consistently). If you're just trying to get a certain order of results you might as well just jot them down in order of personal preference and save yourself the maths.

I've got a ratings system that has changed a lot over time too, but I try not to let the actual results change my weightings, otherwise it'd be rather pointless.

What makes you change your weightings then? If the results don't make sense, you stay with it because the algorithm is always right?

This is a rating system, not some kind of betting algorithm where there is a clear definition of what is correct and what is not.