Fielding Statistics: A New Approach

[RossTaylorsBox]

Seems reasonable. Hypothetically, what would happen if someone like Guptill was able to run fast enough to get under that ball but then spill the chance. Would count as a regulation drop right?

 

[viriya]

Seems reasonable. Hypothetically, what would happen if someone like Guptill was able to run fast enough to get under that ball but then spill the chance. Would count as a regulation drop right?

The ball died on him though. If Guptill gets there on time and makes it, it would most likely be considered a great catch. If he drops it I think it would still be considered a tough one.. There is a slight chance that he might make it look like an easy drop, but you could argue that if he gets there with enough time and drops an easier catch, he should be penalized.

I'm not arguing that my setup is perfect - those grey scenarios would be very rare even if they do occur.

 

[viriya]

That Kohli drop would be ignored because of the "difficult chance" text in the commentary.

Actually it would've been ignored regardless since there's no "dropped" or "spilled" type text.

 

[Athlai]

Yeah I'm with Daemon, I'm not a fan of viriya's methodology here either but you're being right ****s.

Also, you're ridiculing him for using an "arbitrary" 20% modifier? Guess what those kind of made-up modifiers are used in every single ratings system.

The difference is that most systems actually trial a variety of "arbitrary" modifiers to actually find the one that most closely represents the desired or expected result.

This has not taken place. Hence throwing darts blind. You need to do a bit more than set a random modifier and hope for the best. It's why I actually quite like Pews system. Systems are my modified in comparison to another system. Everything relates to something, and it's why CPL can actually list the top 100 batsman in the world without accidentally putting Tim Southee or someone in that list.

 

[viriya]

The difference is that most systems actually trial a variety of "arbitrary" modifiers to actually find the one that most closely represents the desired or expected result.

This has not taken place. Hence throwing darts blind. You need to do a bit more than set a random modifier and hope for the best. It's why I actually quite like Pews system. Systems are my modified in comparison to another system. Everything relates to something, and it's why CPL can actually list the top 100 batsman in the world without accidentally putting Tim Southee or someone in that list.

You do realize that I just introduced never before seen fielding stats right? The ratings system has been around for a grand total of a week. I trialled 30% and ignoring them altogether before settling on 20% based on the premise that 5 regulation catches usually makes people notice a fielder's performance.

I realize you don't like the arbitrariness of the weight, but give me a break ffs. I'm not gonna drop a perfect system out of my ass - I'm asking for feedback so "we" can improve the system. At least RossTaylorsBox is suggesting an alternative.

 

[Athlai]

You do realize that I just introduced never before seen fielding stats right? The ratings system has been around for a grand total of a week. I trialled 30% and ignoring them altogether before settling on 20% I realize you don't like the arbitrariness of the weight, but give me a break ffs. I'm not gonna drop a perfect system out of my ass - I'm asking for feedback so "we" can improve the system. At least RossTaylorsBox is suggesting an alternative.

My argument against these stats is that they're missing far too much information to be useful. My comment above was not directed at you as in my opinion there is no "weighting" that will justify the stats you're arguing for. I think this is a house built on sand scenario, until the game captures fielding statistics to a better degree, no meaningful result can be drawn from them.

I appreciate your effort here but I think it's a hopeless venture. Fielding is a far more complicated thing to measure than batting or bowling in cricket. Unlike baseball there are few set positions, fielders are often required to make up an incredible amount of ground and the type of catches they're expected to take are incredibly varied.

Certain fielders like Guptill will put them-self in an opportunity to field the ball that other players wouldn't get close to. Whether this is a catch, stop or missed opportunity. The ground speed of a player is important, their reflexes are important, the ball speed is important. These just aren't things you will get anything but the most negligible amount of scope on with your method. The feedback I would give is to actually actively look to capture this information yourself, but it's a fairly hopeless venture for a single person without the resources of a broadcasting network.

So I'd just say you shouldn't attempt it at all and refocus your efforts on more accurately trying to weight the runs of Test and FC cricket.

 

[viriya]

Certain fielders like Guptill will put them-self in an opportunity to field the ball that other players wouldn't get close to. Whether this is a catch, stop or missed opportunity. The ground speed of a player is important, their reflexes are important, the ball speed is important. These just aren't things you will get anything but the most negligible amount of scope on with your method. The feedback I would give is to actually actively look to capture this information yourself, but it's a fairly hopeless venture for a single person without the resources of a broadcasting network.

You are overcomplicating the issue. Guptill is a good enough fielder that his brilliance shows in the results. He is already rated #14 in Tests and #20 in ODIs career-wise (albeit since 2005) after not so many matches. With a great ~5% drop rate and a high great catch frequency, he'd keep climbing if he keeps it up.

 

[Athlai]

Okay lets approach this on a different angle, what does your system say on McCullum's drop yesterday, cricinfo described it as "easy" yes?

 

[viriya]

Okay lets approach this on a different angle, what does your system say on McCullum's drop yesterday, cricinfo described it as "easy" yes?

That was the relatively rare case where they ****ed up. It didn't describe it as "easy" but it didn't say it was a "tough chance" or the equivalent which it should have. I've followed the commentary for long enough that I trust that is the exception and not the rule.

 

[Athlai]

It isn't "all that rare" unless you actually sample how they're commentary compares to the real life result over a series of catches. Try it for every WC game you watch. I'm quite frequently disgusted with cricinfo's text commentary on certain things, it's definitely biased to certain teams and players.

I've also followed cricinfo commentary for years and will frequently have both cricinfo and live play on my screens because I like to have more numbers on screen than the TV broadcast provides. Just relying on cricinfo without actually accounting for any sort of actual margin of error is something you should avoid.

That said my opinion is still to give up on the whole thing, so take that as you will.

 

[viriya]

At the end of the day fielding statistics will have to rely on subjective opinion. Even if I were to hire an army of monkeys to watch every since match ever and keep track of fielding events that would be a disclaimer that would need to be made.

And that's also why there's no such thing as fielding stats in cricket today. I think my attempt shows that it has value - drop rates would be very useful as a quick way to compare fielders imo. Would never have imagined such a huge difference between players for that stat.

 

[Athlai]

Drop rates will be moderately valuable admittedly.

 

[viriya]

So far for the Ind v Pak match, the significant fielding events would be Akmal's Kohli drop and Shezad's direct hit to get Dhawan out.

 

[NUFAN]

Drop rates is excellent and can debunk the saying "he'd catch that 99 times out of 100".

My major gripe is the points value assigned for each catch. If a player with a low rating is having a great game and is on 72 (68) if someone takes a great catch they should be rewarded more then the overall rating suggests.

Is there any potential to work out a formula for match situation?

 

[RossTaylorsBox]

Any possibility of getting the data that you've parsed (i.e. player, catch difficulty, caught/dropped)? I want to play around with the numbers and compare results.

 

[viriya]

Drop rates is excellent and can debunk the saying "he'd catch that 99 times out of 100".

My major gripe is the points value assigned for each catch. If a player with a low rating is having a great game and is on 72 (68) if someone takes a great catch they should be rewarded more then the overall rating suggests.

Is there any potential to work out a formula for match situation?

I already consider match situation for batting/bowling ratings. It's plausible to do the same for fielding, but I'm going to hold off on more complex measures until I get the basics solid (data parsing issues mostly).

 

[viriya]

Any possibility of getting the data that you've parsed (i.e. player, catch difficulty, caught/dropped)? I want to play around with the numbers and compare results.

Gonna hold off until at least I feel like I've resolved most of the parsing issues. I can send you specific stuff later - I don't store catch difficulty btw - if it's a tough chance it's a non-event.

 

[viriya]

Ratio for dropped sitters (excluding keepers) would be a bit higher I reckon. About 10 per regulation catch? You'd expect maybe one per innings.

Turns out the great catch to regulation catch ratio is ~3% for both Tests and ODIs. Low as I expected, but I'm sure I'm missing some great ones with parsing issues, so probably 5% is a good number to go with. Feels too low but curious to see the results with that.

 

[RossTaylorsBox]

Gonna hold off until at least I feel like I've resolved most of the parsing issues. I can send you specific stuff later - I don't store catch difficulty btw - if it's a tough chance it's a non-event.

Ah OK, no problem. I was thinking that it would be interesting to find out how successful a player is at catching "tough" chances (i.e. great catches).

 

[RossTaylorsBox]

Turns out the great catch to regulation catch ratio is ~3% for both Tests and ODIs. Low as I expected, but I'm sure I'm missing some great ones with parsing issues, so probably 5% is a good number to go with. Feels too low but curious to see the results with that.

Is that with or without keepers? Even 1 in 20 catches being great ones seems quite low, but who knows.