Fielding Statistics: A New Approach

[Riggins]

Only spot checks.. not practical for me to review 8 years worth of Test and ODI data (300MB of text).

I'll check the Steve Smith drops to see if there are any misattributions/issues. Any cases where the data seems off that you can point out will only make things better.. Rerunning everything only takes a couple hours so fixes can be implemented quickly.

I just didn't know how events like this would be treated.

12.3
Harris to Vijay, no run, whoa! Vijay looks to flick it to the leg side, but it hits the pads and lobs back down the pitch. For a moment it looks like it was off the inside edge, but there was nothing on it. Drops short of Harris anyway

Johnson to Vijay, no run, that lands short of the slip cordon! It was a well directed short one.. rearing towards the batsman's face, hits the gloves and loses the momentum to drop ahead of Haddin


2.5
Johnson to Vijay, no run, this one targets the rib-cage! Vijay is surprised but handles it pretty well, takes his bottom hand off and drops it down towards Marsh at short leg

 

[NUFAN]

Steve Smith has dropped 10 catches in test matches?

I found 9 drops and 23 catches = 28.13%

The Ashes - 5th Test
Australia v England 2011
124.2
Smith to Bell, no run, dropped! Oh boy, Smith has given this some air and Bell drives it hard off the meat of the bat back at the bowler, who has spilled it at waist height. It was really flying, but Smith, a good fielder, didn't have to move. Should have taken it

4th Test
India v Australia 2013
47.2
Lyon to Dhoni, 2 runs, dropped. Sharp turn and bounce from a length, takes the glove, to the right of Smith at backward short leg. He takes it too casually, and lets it slip through. Hits him on the wrist, he has actually been done by the slowness of it off the glove

3rd Test
England v Australia 2013
130.4
Lyon to Prior, 2 runs, dropped, now Prior is down the pitch and swinging, clubs it towards midwicket where Steve Smith can't quite hold onto what would have been a stunning grab

4th Test
England v Australia 2013
93.2
Lyon to Swann, 1 run, dropped, it was in the air for an age as Swann skipped out to slice a skier, Smith was under it at long on, he judged it alright oo, but it has popped out in the end

4th Test
Australia v England 2013
6.4
Harris to Carberry, no run, edged! Did it carry? Yes, dropped at third slip by Smith diving across, one-handed, to his left. Carberry pushing forward at a good-length ball outside off. Thick edge. It might have reached Clarke at second but Smith committed

2nd Test
South Africa v Australia 2014
47.2
Harris to de Kock, no run, dropped second ball. Another leading edge, much to the left of short cover, Smith goes for it, but finds out is going away from him, and manages to get one hand to it. It is high too

2nd Test
Australia v Pakistan 2014
55.5
Lyon to Azhar Ali, no run, dropped, but only a half-chance Azhar lunges forward to a flighted ball and looks to work it leg side. Hits it uppishly to short leg, who couldn't react quickly enough to catch it. Went off the face of his bat. Smith was the fielder

4th Test
Australia v India 2015
52.5
Watson to Rahul, no run, dropped, it is Smith who runs back from slip after this top edged pull, the sun is in his eyes, and it bursts through his hands into the ground, Smith points up as if to suggest something put him off, was it the spider cam, it was right above him

4th Test
Australia v India
82.1
Starc to Kohli, 3 runs, Smith has dropped another catch, this time at second slip, poor shot from Kohli, pushing way away from the body at a length ball, came over Smith's head at second slip, he had to get on his toes, but was it takeable

 

[arana]

What if a fielder catches in the hands, then slides and then drops the ball ... ?

 

[viriya]

I just didn't know how events like this would be treated.

12.3
Harris to Vijay, no run, whoa! Vijay looks to flick it to the leg side, but it hits the pads and lobs back down the pitch. For a moment it looks like it was off the inside edge, but there was nothing on it. Drops short of Harris anyway

Johnson to Vijay, no run, that lands short of the slip cordon! It was a well directed short one.. rearing towards the batsman's face, hits the gloves and loses the momentum to drop ahead of Haddin


2.5
Johnson to Vijay, no run, this one targets the rib-cage! Vijay is surprised but handles it pretty well, takes his bottom hand off and drops it down towards Marsh at short leg

Yes I have a lot of checks to make sure that just because the word "drop" is in the commentary it's not automatically considered a drop.

 

[viriya]

What if a fielder catches in the hands, then slides and then drops the ball ... ?

An effort is made to ignore tough chances when calculating drop rates.

 

[cnerd123]

It's based on text commentary data.

Methodology: cricrate | Test Fielding Ratings

If you want to discuss this further: http://www.cricketweb.net/forum/cricket-chat/65351-fielding-statistics-new-approach-14.html

The problems with this, ignoring methodology, is that it assumes that:
A) The guy writing the ball by ball commentary has a strong understanding of wicketkeeping and fielding (ie, has the ability to tell apart a catch that looks good and a catch that came about as a result of good fielding)
B) All the ball by ball commentary being done is thorough and indepth, in terms of evaluating all fielding (or atleast catches)
C) The catches are described in such a way that lets them be ranked against other cathes from other matches commentated on by other people. One person's 'good catch' could be another person's 'catch'.

 

[cnerd123]

Also its a huge methodology problem to define fielding ourely in terms of catches and runouts, given the bulk of the work done on the field is ground fielding.

 

[viriya]

Also its a huge methodology problem to define fielding ourely in terms of catches and runouts, given the bulk of the work done on the field is ground fielding.

Runs saved/given away, stumpings/missed stumpings are also considered.

 

[viriya]

The problems with this, ignoring methodology, is that it assumes that:
A) The guy writing the ball by ball commentary has a strong understanding of wicketkeeping and fielding (ie, has the ability to tell apart a catch that looks good and a catch that came about as a result of good fielding)
B) All the ball by ball commentary being done is thorough and indepth, in terms of evaluating all fielding (or atleast catches)
C) The catches are described in such a way that lets them be ranked against other cathes from other matches commentated on by other people. One person's 'good catch' could be another person's 'catch'.

These are valid concerns but overblown based on some of my checks, I'm sure it affects some matches but the majority of fielding events are not impossible to evaluate.

Further evidence that this is not a significant issue is that the generally agreed upon great fielders come out on top with this method.

Another thing to consider is that there is absolutely no other statistical method of rating fielders currently.

 

[Coronis]

These are valid concerns but overblown based on some of my checks, I'm sure it affects some matches but the majority of fielding events are not impossible to evaluate.

Further evidence that this is not a significant issue is that the generally agreed upon great fielders come out on top with this method.

Another thing to consider is that there is absolutely no other statistical method of rating fielders currently.

Just because there is no other method doesn't mean this method is good, or accurate in any way.

 

[viriya]

Just because there is no other method doesn't mean this method is good, or accurate in any way.

Just because you haven't looked at the data doesn't mean you know if this method is good or accurate.

 

[cnerd123]

Runs saved/given away, stumpings/missed stumpings are also considered.

Yea, but that's still less than the majority of what fielding entails. You can't evaluate, for example, the way the Indian batsmen in the recent series were cut off from scoring singles when the ball was within the ring due to the speed of the SA fielders, because there will be no significant mention of it in the ball-by-ball commentary, because the Indian batsmen didn't even attempt to run most of those singles as they knew the quality of the fielding would prevent them from doing so.

These are valid concerns but overblown based on some of my checks, I'm sure it affects some matches but the majority of fielding events are not impossible to evaluate.

Further evidence that this is not a significant issue is that the generally agreed upon great fielders come out on top with this method.

Another thing to consider is that there is absolutely no other statistical method of rating fielders currently.

Well, I didn't question the methodology, and 'there is no other method available currently' isn't a sufficient defence of why this method is any good.

My problem is with the data being used. Unlike with Runs and Wickets, we don't have any objective, numerical data with regards to fielding. We can evaluate a cricketer's runs and wickets -which are factual measures of what happened in the game- and then analyse it based on our own subjective values and biases and preferences. But by using ball-by-ball commentary, your entire raw data set is subject to the values, biases and preferences of the commentator at that point.

For example; there is an Instagram account which shares cricket highlights, and they recently shared a montage of clips of De Kock's wicketkeeping, saying that he is the best young keeper in the world in the moment. I vehemently disagree. I see his spectacular catches, and see them being a result of his poor footwork that cause him to have to leap and reach for deliveries that a better keeper would have had well covered. If the owner of that account and I were to both commentate on the same game, he'd be lauding de Kock's catching ability while I'd be trying to highlight how he gets into bad positions and thus needs to rely on his athleticism to complete a catch. If you used my ball-by-ball commentary, De Kock would fall down your ratings. If you used his, De Kock would climb

Similarly, catches that are 'made to look easy' don't get a whole lot of discussion, even tho they may have been a result of spectacular fielding, because they don't look as spectacular as one that requires a player throwing himself around to reach the ball, even if the latter had a massive misjudgement upfront.

There are also biases in reporting/commentary - a catch taken by a 'generally agreed upon great fielder' does sometimes receive more praise than it deserves.

This is a big problem for me. The data you are using rewards fielders and keepers who look good, not those who actually are good. Runs and Wickets are objective; you can argue till you are blue in the face that a player who averaged 55 over his career was actually a FTB who was never found out, but any form of statistical measurement or evaluation will still place him highly. But whether a fielder or keeper places highly in your rankings, right now, is affected by perceptions of that fielder's ability.

There are also effects of being a good fielder that are passive, and not active like misfields/cutting off a boundary/being involved in catches and runouts. The simple presence of a particular fielder in a particular area can discourage a batsman from playing a certain stroke, which could play a contributing factor to his dismissal later on. An example is a tall man like Ashley Giles or Kevin Pietersen at gully for England's tall, hit the deck fast bowlers, in order to discourage cuts shots in the air There is no way of measuring this effect, yet it's clear to all those viewing, and plays a role in how we evaluate a player's fielding ability.

 

[viriya]

Yea, but that's still less than the majority of what fielding entails. You can't evaluate, for example, the way the Indian batsmen in the recent series were cut off from scoring singles when the ball was within the ring due to the speed of the SA fielders, because there will be no significant mention of it in the ball-by-ball commentary, because the Indian batsmen didn't even attempt to run most of those singles as they knew the quality of the fielding would prevent them from doing so.

These are intricacies that are missed in evaluating batsmen and bowlers too.. No one discounts the runs you made just because you gave a lot of chances, no one cares that you bowled a long hop as long as you got a wicket. Just because you can't statistically evaluate the complete picture (at least for now), doesn't mean that what you can evaluate is not useful.

Well, I didn't question the methodology, and 'there is no other method available currently' isn't a sufficient defence of why this method is any good.

My problem is with the data being used. Unlike with Runs and Wickets, we don't have any objective, numerical data with regards to fielding. We can evaluate a cricketer's runs and wickets -which are factual measures of what happened in the game- and then analyse it based on our own subjective values and biases and preferences. But by using ball-by-ball commentary, your entire raw data set is subject to the values, biases and preferences of the commentator at that point.

For example; there is an Instagram account which shares cricket highlights, and they recently shared a montage of clips of De Kock's wicketkeeping, saying that he is the best young keeper in the world in the moment. I vehemently disagree. I see his spectacular catches, and see them being a result of his poor footwork that cause him to have to leap and reach for deliveries that a better keeper would have had well covered. If the owner of that account and I were to both commentate on the same game, he'd be lauding de Kock's catching ability while I'd be trying to highlight how he gets into bad positions and thus needs to rely on his athleticism to complete a catch. If you used my ball-by-ball commentary, De Kock would fall down your ratings. If you used his, De Kock would climb

Similarly, catches that are 'made to look easy' don't get a whole lot of discussion, even tho they may have been a result of spectacular fielding, because they don't look as spectacular as one that requires a player throwing himself around to reach the ball, even if the latter had a massive misjudgement upfront.

There are also biases in reporting/commentary - a catch taken by a 'generally agreed upon great fielder' does sometimes receive more praise than it deserves.

This is a big problem for me. The data you are using rewards fielders and keepers who look good, not those who actually are good. Runs and Wickets are objective; you can argue till you are blue in the face that a player who averaged 55 over his career was actually a FTB who was never found out, but any form of statistical measurement or evaluation will still place him highly. But whether a fielder or keeper places highly in your rankings, right now, is affected by perceptions of that fielder's ability.

There are also effects of being a good fielder that are passive, and not active like misfields/cutting off a boundary/being involved in catches and runouts. The simple presence of a particular fielder in a particular area can discourage a batsman from playing a certain stroke, which could play a contributing factor to his dismissal later on. An example is a tall man like Ashley Giles or Kevin Pietersen at gully for England's tall, hit the deck fast bowlers, in order to discourage cuts shots in the air There is no way of measuring this effect, yet it's clear to all those viewing, and plays a role in how we evaluate a player's fielding ability.

This is the whole reason that there are no insightful fielding statistics - people just point out the subjectivity and think it's better to avoid attempting in the first place. But in baseball, fielding is evaluated with video review of games and assigning errors accordingly - completely subjective, but found to be massively helpful in evaluating the value of fielding in the game, and so improving ground fielding as a result - something cricket is/was late to do because there are no numbers to back how much fielding is important.

I agree that biases play a part, and based on some initial results there was a suggestion that indian fielders were overrated in cricinfo commentary because of national bias. This might actually be the case (and I believe it to be so), but it wasn't that significant of a difference that it completely invalidated the results. The clearly better fielders still came up to the top. You say I reward fielders who "look good" and not those who are "actually good", and there is a point to that except you are going to the extreme. Fielders who are actually good always look good in the numbers - there might be some false overrating of fielders, but those who are good always "look good" in commentary.

 

[Coronis]

Just because you haven't looked at the data doesn't mean you know if this method is good or accurate.

That wasn't the point. You shouldn't be using "this is the only statistical method of rating fielders" to promote it as a good method.

 

[cnerd123]

These are intricacies that are missed in evaluating batsmen and bowlers too.. No one discounts the runs you made just because you gave a lot of chances, no one cares that you bowled a long hop as long as you got a wicket. Just because you can't statistically evaluate the complete picture (at least for now), doesn't mean that what you can evaluate is not useful.



This is the whole reason that there are no insightful fielding statistics - people just point out the subjectivity and think it's better to avoid attempting in the first place. But in baseball, fielding is evaluated with video review of games and assigning errors accordingly - completely subjective, but found to be massively helpful in evaluating the value of fielding in the game, and so improving ground fielding as a result - something cricket is/was late to do because there are no numbers to back how much fielding is important.

I agree that biases play a part, and based on some initial results there was a suggestion that indian fielders were overrated in cricinfo commentary because of national bias. This might actually be the case (and I believe it to be so), but it wasn't that significant of a difference that it completely invalidated the results. The clearly better fielders still came up to the top. You say I reward fielders who "look good" and not those who are "actually good", and there is a point to that except you are going to the extreme. Fielders who are actually good always look good in the numbers - there might be some false overrating of fielders, but those who are good always "look good" in commentary.

Baseball is an interesting comparison. They actually have cameras all over the field tracking every play, and I believe the people who evaluate the footage are trained in how to do so.

I get your point that this is the 'best possible way' and that the errors may not be that large, but personally I just wouldn't try any sort of fielding analysis at all. Or if I did, I wouldn't suggest it's meaningful. It's hard to take a system seriously where the raw data is so subjective. You can't really gain any new insights - if the commentary carries X bias, the resulting analysis will only serve to reaffirm X bias. The results just serve to confirm the assumptions we all already hold; the players whom everyone thinks are fantastic will be more likely to rise to the top partly because everyone thinks they are fantastic.

This lack of objective raw data is a huge obstacle in analysing fielders IMO. I appreciate that you have tried, and it may well be the best possible way to analyse past games apart from re-watching them all...but it just isn't good enough for me.

 

[AndyZaltzHair]

These are intricacies that are missed in evaluating batsmen and bowlers too.. No one discounts the runs you made just because you gave a lot of chances, no one cares that you bowled a long hop as long as you got a wicket. Just because you can't statistically evaluate the complete picture (at least for now), doesn't mean that what you can evaluate is not useful.

The statistical methods of recording runs and wickets are solid. There can't be no question about it that X player made Y run in a match or X player got Y wicket in a match. Matches after matches, these are tallied. If someone wants to measure a player by comparing average etc from these tallies after a set number of matches, it's perfectly ok because a a player might have scored 1 fluke hundred but he won't score 10 fluke hundreds. Similarly a bowler might have taken 10 fluke wickets but he would not get 100 fluke wickets. The sheer number outweighs those gaps. Therefore I am ok with the current methodology of evaluating batsmen and bowlers purely from statistical point of view. For the fielders and WK, there is no such method. As I have already said the gaps are too huge. It's good that least you are trying something. But at the same time can't help feeling these stats are giving wrong notion.

 

[marc71178]

Runs saved/given away, stumpings/missed stumpings are also considered.

Not accurately enough to give this any real value.

 

[dfrinku]

But in baseball, fielding is evaluated with video review of games and assigning errors accordingly - completely subjective, but found to be massively helpful in evaluating the value of fielding in the game, and so improving ground fielding as a result - something cricket is/was late to do because there are no numbers to back how much fielding is important.

If you actually did this then I would say it would be very worthwhile and you'd have my support Following the cricinfo text is a bit of a silly way to evaluate anything but especially fielding.

 

[viriya]

If you actually did this then I would say it would be very worthwhile and you'd have my support Following the cricinfo text is a bit of a silly way to evaluate anything but especially fielding.

Sure let me buy all historical video footage of matches and watch all of them for fielding events till my eyes bleed.

 

[weldone]

No one discounts the runs you made just because you gave a lot of chances

Don't speak on behalf of Richard