Statistics are for losers

[sledger]

Have been watching quite a bit of NFL for the last few weeks, and came across an interview with Bill Belichick a few years ago in which he suggested that stats were something only losers would be interested in.

Different sport obviously, but can we learn about the true nature of statistics in cricket by comparing it to the attitude taken to stats in the NFL?

Undoubtedly imo.

 

[indiaholic]

NFL players have too much brain damage to understand stats imo.

 

[Bijed]

I'm not 100% sure what mean by "learn about the true nature of statistics" tbh, but I'll give my thoughts on Mr Belichick's idea that stats were something only losers would be interested in.

If he means players, then yes, a player playing to improve a specific statistic at the expense of their team's chances of winning a match is a bad thing indeed.

If he means fans, then I have no problem whatsoever with finding stats interesting to the extreme extent that some people do. No problem either with them forming an opinion of a player or team based on stats alone as it's up to them how and why they enjoy the game

 

[Shri]

NFL players have too much brain damage to understand stats imo.

They wouldn't be able to comprehend Ashwin.

 

[sledger]

I'm not 100% sure what mean by "learn about the true nature of statistics" tbh, but I'll give my thoughts on Mr Belichick's idea that stats were something only losers would be interested in.

If he means players, then yes, a player playing to improve a specific statistic at the expense of their team's chances of winning a match is a bad thing indeed.

If he means fans, then I have no problem whatsoever with finding stats interesting to the extreme extent that some people do. No problem either with them forming an opinion of a player or team based on stats alone as it's up to them how and why they enjoy the game

Well, to me statistics, as a very concept, are poorly understood in the cricketing context.

They are something upon which a very heavy influence is placed and pervade general discourse, but yet this heavy emphasis impedes the answering of a number of interesting questions as to the true nature and value of stats in this context.

Precisely whose stats should be tracked? Precisely what data is it that should be tracked? What should the extent should that tracking be? And, what underlying theoretical justifications can be identified for doing so? What can we hope to achieve by doing this? Is the tracking of stats intended as a means of improving performance etc...? Or is the tracking of stats something that is worthwhile in and of itself.

By taking a comparative approach and analysing the usage and role of stats in other sports, I think we could shed some light as to the nature of stats in cricket, and how we should approach them.

 

[BoyBrumby]

To paraphrase AE Housman,

"And gentlemen who treat statistics as rational men treat all evidence, and test it by reason and by the knowledge which they have acquired, these are blamed for rashness and capriciousness by gentlemen who use statistics as drunkards use lamp-posts,—not to light them on their way but to dissimulate their instability.”

 

[Bijed]

So, amongst other things, are you asking whether it would be worthwhile to collect more specific and detailed statistics about a player, as in this thread? I believe that there is merit in that, though how much is certainly debatable. As to the extent to which we should go, in theory I would say there is no limit to how far , but of course in reality in the absence of some very dedicated people it comes down to a cost-benefit trade-off.

If you're going to select a team, you need statistics to let you do that and the bare minimum of statistics you need will always be available (well, the raw data to calculate them) because you collect that data simply by keeping score of a match to a fairly basic standard. Whether the use of more bespoke statistics allows for materially better selectorial decisions remains to be seen. So, yes, cricket could learn from other sports here but I would expect it to derive less value just because of the nature of the sport.

There's more I could say on the matter but right now I don't feel capable of putting my thoughts together in a coherent manner.

 

[Howe_zat]

So, amongst other things, are you asking whether it would be worthwhile to collect more specific and detailed statistics about a player, as in this thread? I believe that there is merit in that, though how much is certainly debatable. As to the extent to which we should go, in theory I would say there is no limit to how far , but of course in reality in the absence of some very dedicated people it comes down to a cost-benefit trade-off.

If you're going to select a team, you need statistics to let you do that and the bare minimum of statistics you need will always be available (well, the raw data to calculate them) because you collect that data simply by keeping score of a match to a fairly basic standard. Whether the use of more bespoke statistics allows for materially better selectorial decisions remains to be seen. So, yes, cricket could learn from other sports here but I would expect it to derive less value just because of the nature of the sport.

There's more I could say on the matter but right now I don't feel capable of putting my thoughts together in a coherent manner.

You are giving sledger perhaps a little more airtime than is necessary.

 

[Howe_zat]

@sledger - your thread title hurt my feelings. Apologise immediately.

 

[srbhkshk]

This is actually a broad problem in data science, not only in sports. People look at irrelevant or incomplete stats and conclude that stats are misleading and useless when they fail to match reality. You just need better models, obviously that's easier said then done.

 

[sledger]

This is actually a broad problem in data science, not only in sports. People look at irrelevant or incomplete stats and conclude that stats are misleading and useless when they fail to match reality. You just need better models, obviously that's easier said then done.

I think these are important issues, but that in many ways to ask these questions is to put the cart before the horse.

 

[sledger]

@sledger - your thread title hurt my feelings. Apologise immediately.

You apologise for trying to shut me up when I started talking about Pratters on CW commentary.

 

[sledger]

So, amongst other things, are you asking whether it would be worthwhile to collect more specific and detailed statistics about a player, as in this thread? I believe that there is merit in that, though how much is certainly debatable. As to the extent to which we should go, in theory I would say there is no limit to how far , but of course in reality in the absence of some very dedicated people it comes down to a cost-benefit trade-off.

If you're going to select a team, you need statistics to let you do that and the bare minimum of statistics you need will always be available (well, the raw data to calculate them) because you collect that data simply by keeping score of a match to a fairly basic standard. Whether the use of more bespoke statistics allows for materially better selectorial decisions remains to be seen. So, yes, cricket could learn from other sports here but I would expect it to derive less value just because of the nature of the sport.

There's more I could say on the matter but right now I don't feel capable of putting my thoughts together in a coherent manner.

This is interesting. What are you driving at with this?

 

[Magrat Garlick]

This is actually a broad problem in data science, not only in sports. People look at irrelevant or incomplete stats and conclude that stats are misleading and useless when they fail to match reality. You just need better models, obviously that's easier said then done.

GIGO applies, too. The best model can't save a dataset compiled by illiterates.

 

[mr_mister]

In cricket, stats are important

 

[sledger]

Maybe, but why? And what does it truly mean to say that something is important?

 

[srbhkshk]

GIGO applies, too. The best model can't save a dataset compiled by illiterates.

True, although it seems to me that this is less of a problem for cricket in general, some pretty decent data sets are available (If only we knew what to do with them. :P )

 

[Burgey]

True, although it seems to me that this is less of a problem for cricket in general, some pretty decent data sets are available (If only we knew what to do with them. :P )

Guys, can we take this to the Nerdorium please?

 

[Magrat Garlick]

True, although it seems to me that this is less of a problem for cricket in general, some pretty decent data sets are available (If only we knew what to do with them. :P )

Well, the most promising data sets are proprietary (the simple outcome stats are useful as a starting point, but really don't get us much further than the lamp-post). Unfortunately, the people capable of paying for the data don't have the sense to know what to with it.

 

[Burgey]

Well, the most promising data sets are proprietary (the simple outcome stats are useful as a starting point, but really don't get us much further than the lamp-post). Unfortunately, the people capable of paying for the data don't have the sense to know what to with it.

Guys, can we take this to the Nerdorium please?

.