Statistics

[ramssuperbowl99]

1 minute ago, LETSGOBROWNIES said:

I knew I loved you, but I didn’t know I loved you this much.

I’ll break the news to my wife.

I've considered getting the front office position in Cleveland by just being a homeless guy.

[LETSGOBROWNIES]

8 minutes ago, ramssuperbowl99 said:

I've considered getting the front office position in Cleveland by just being a homeless guy.

But then you’d work for..... him.

[AlNFL19]

I’ve always thought it would be interesting to see the viability of a machine learning model that tracks how quickly a quarterback progresses from read to read. Obviously there are a lot of things keeping that from happening but I’m sure somebody could flesh it out and find something there with some decent results. 

[ET80]

1 hour ago, ramssuperbowl99 said:

@ET80 here's a real word example of how this could function. The Red Sox drafted an otherwise completely unassuming Mookie Betts because of neurological data.

https://www.bostonglobe.com/sports/2015/02/18/neuroscouting-may-give-red-sox-heads-prospects-potential/EFBHR3zNdThk1NboRpNMHL/story.html

Half of getting stats to inform decision making is finding something you can measure that's reliable and predictive. The Red Sox wanted to know about Mookie's reaction abilities, so they basically made a model of it and saw what the data looked like.

 

You could do the same thing with a QB. Have a WR flash up on the screen either with a DB all over him, or with separation, and have him hit one button for next read, and another button for throw. Measure the reaction time and see what percentage he gets right. Maybe do some where there is also a lineman that pops up near the bottom of the screen.

So often you have NFL teams talk about how hard it is to find a QB that can process information quickly. This simple experiment may or may not work, but it's a different way of thinking about answering the question instead of exclusively relying on college performance.

This is actually a fantastic read. Thanks for sharing.

[AlNFL19]

23 minutes ago, ET80 said:

This is actually a fantastic read. Thanks for sharing.

Agreed. Awesome stuff. 

[ramssuperbowl99]

59 minutes ago, AlNFL19 said:

I’ve always thought it would be interesting to see the viability of a machine learning model that tracks how quickly a quarterback progresses from read to read. Obviously there are a lot of things keeping that from happening but I’m sure somebody could flesh it out and find something there with some decent results. 

Initial thoughts, not really well organized or hashed out:

What we're trying to measure here is a capacity to process information quickly.  That may or may not show up consistently, and may or may not correlate with the average time spent on a read (I'd guess there's a loose correlation), but where you'd run into trouble are 3 situations

• A QB looks at a potential receiver solely as a decoy, knowing immediately the ball is going to the second read. This is the biggest issue since it decreases the read time (which could be interpreted as an inflated ability to process information).
• A QB stares down a potential WR, then at the last second pivots to another target. This is easy enough to filter (it doesn't take anyone 1.5 seconds to make a read, throw it out as an intentional stare down).
• Plays with only one read, or get out quick type bubble screens. Again, easy enough to filter.

The problem, and this is kind of universal with capacity or maximum driven parameters, is that if you have any poor data in that sample, it carries straight through to the predictive part with no buffer and your model sucks. That's why I'm more into the lab-driven approach. 

For all the **** it gets, 40 yard dashes are super useful (either themselves or by the 10 yard split) as a way to just say what the maximum acceleration a player can produce is. Not directly related to football, but very specifically answers that question, and is pretty reproduceable as long as you're timing it and not running on a trampoline track and whatnot. 

That's why I'd go more "combine style" here. Do some type of reaction time test instead of scouring game footage, and then see how that matches up with film breakdowns.

 

In all of this, you need to design your measurements and observations with the final target in mind. In sports evaluation, that's a market inefficiency - some player that the current system sucks butt at projecting. In this case, the market inefficiency you're probably trying to go after is assessing the guys who have the capacity to make multiple reads, but don't have to do it because they run a 1 read college offense. You're not really going to get a ton of info on someone from Stanford, but you may be able to differentiate between physically gifted Air Raid dudes. But some algorithm that picks when a guy makes a decision won't help you here, because it's all "get ball, throw ball". You're looking at capacity to do something the player hasn't yet done by design.

Edited August 1, 2019 by ramssuperbowl99

[ramssuperbowl99]

46 minutes ago, ET80 said:

This is actually a fantastic read. Thanks for sharing.

Admittedly, this is why I lose my mind when people say "stats aren't useful, football's got too many variables". 

They totally are useful, or maybe more accurately, they totally can be useful. You have to get creative with what you measure because football is complicated, but at the end of the day most complicated stuff is just a bunch of simple stuff stacked on top of each other. Break it down to the simple stuff, measure it, predict with it, learn from your predictions, and repeat.

Edited August 1, 2019 by ramssuperbowl99

[ET80]

6 minutes ago, ramssuperbowl99 said:

Admittedly, this is why I lose my mind when people say "stats aren't useful, football's got too many variables". 

They totally are useful, or maybe more accurately, they totally can be useful. You have to get creative with what you measure because football is complicated, but at the end of the day most complicated stuff is just a bunch of simple stuff stacked on top of each other. Break it down to the simple stuff, measure it, predict with it, learn from your predictions, and repeat.

I completely agree with this - but I've also seen people become TOO reliant on statistics and deep dive analytics. 

There's a middle ground, and those who can walk that ground will probably be successful in this sport. 

[Kiltman]

On 7/26/2019 at 7:17 PM, AlNFL19 said:

What are your opinions on statistics in the NFL in general? Do you like them and think they provide a lot of useful information? Do you think they're stupid? Are you somewhere in between? What's your view on regular old NFL statistics versus stuff like Football Outsiders?

I'm a big fan of statistics because I think they can paint a great picture with the proper context, and they are obviously objective. I think statistics getting a bad rep is a little unfair. However, obviously, you can't say everything with them, but I disagree with the people who think that "you have to watch the film and statistics are useless." The way I see it, the bottom line is they're results. If you can acknowledge the context behind those results, I think statistics can be very interesting and useful in the NFL.

I love them as a tool, but they are overused in day to day arguments.
Often given without full context.

You can find or bend stats to anything you need it for really, and people do that liberally. 

There are players that get propped up by their teammates and coaches, but actually aren't that good by themselves, just like there are really good players that cover mistakes of bad teammates and coaches at the cost of their own stats.

Outside of stats, when you watch guys, you can just see some of them have something more than their peers. Also that some others are limited in what they can do.
But the numbers don't always show it.

[ramssuperbowl99]

37 minutes ago, ET80 said:

I completely agree with this - but I've also seen people become TOO reliant on statistics and deep dive analytics. 

There's a middle ground, and those who can walk that ground will probably be successful in this sport. 

There are people who lazily interpret statistics without understanding what was measured, the goal of the measurement, the variability across a population, the impact of covariates, etc. Those people tend not to be statisticians.

But a far larger segment of people in sports conflate finding a measurement that will lead to a predictive statistic, and interpretation of a statistic itself. There is very little reason to stop trying to find an improved measurement. And provided your stat is made with a target in mind and the interpretation stays within that target, you're not likely to get in trouble there either.

Edited August 1, 2019 by ramssuperbowl99

[Superman(DH23)]

Stats are a tool, and just like any other tool, they are only as good as the person using them.  I think analytics has a place in football, but I dont think it can ever replace the eye test either.

[SkippyX]

On 7/27/2019 at 12:28 PM, KellChippy said:

From 1990-2009 teams that swept the regular season series went 11-5 in playoff matchups.

The 1999 Jags went 15-0 against the rest of the NFL including beating Jimmy Johnson and Dan Marino 62-7 in the divisional round.

They went and 0-3 against the Titans including beatings by 27 and 19 in week 16 and in the AFCCG

This is Mark Brunell having an emotional breakdown over it:

 

Oh wait, that was his reaction to the NFL lies about ball deflation.

[SkippyX]

Stats can be very useful. Here is an example.

National media clowns will refer to Donovan McNabb is the best QB in Eagles history.

Here are some of his rankings as an Eagle QB (minimum 100 attempts)

• Completion % - 10th (behind Vick)
• TD% 19th
• Interception % - 5th
• Y/A - 17th (way behind Vick)
• ANY/A - 5th (behind Vick)
• Sack % - 14th (barely ahead of Vick)
• QB Rating - 6th (behind Vick)

If you are wondering how many of those categories you will find Foles and Wentz ahead of him, it would be every single one.

Garcia also mopped the floor with McNabb but only played 6 games with McNabb's teammates.

McNabb's rating was .1 below Sanchez and .1 ahead of Bradford.

So of the 10 QBs to start 5 or more games for the Eagles since 1999, I guess he was clearly better than Kolb, Feeley, and Mike McMahon.

A medicore dude does not become great after 10 years of mediocrity.

Sure he gets points as the 3rd best runner as an Eagles QB, but Cunningham and Vick kill him there.

6.7, 6.6, and 5.7 yards per carry

 

Stats are also fun when the myopic compilation guys cry about Warner and Terrell Davis and you can point to Testaverde's compile stats vs Joe Montana 

TD's 5.6 yards per carry in the playoffs packs a pretty good punch as does Warner's 102.8 playoff passer rating.

 

The most classic destruction by statistics was Dr Z vs Art Monk.

Dr Z accused Monk of 'catching 500 eight-yard-hooks to get first downs which is nice but not worthy of the Hall of Fame.'

If you take away the 500 catches and 4000 yards you are left with 440 catches at 19.82 yards per catch which makes him a first ballot guy as a whole.

[theuntouchable]

1 hour ago, SkippyX said:

Stats can be very useful. Here is an example.

National media clowns will refer to Donovan McNabb is the best QB in Eagles history.

Here are some of his rankings as an Eagle QB (minimum 100 attempts)

• Completion % - 10th (behind Vick)
• TD% 19th
• Interception % - 5th
• Y/A - 17th (way behind Vick)
• ANY/A - 5th (behind Vick)
• Sack % - 14th (barely ahead of Vick)
• QB Rating - 6th (behind Vick)

If you are wondering how many of those categories you will find Foles and Wentz ahead of him, it would be every single one.

Garcia also mopped the floor with McNabb but only played 6 games with McNabb's teammates.

McNabb's rating was .1 below Sanchez and .1 ahead of Bradford.

So of the 10 QBs to start 5 or more games for the Eagles since 1999, I guess he was clearly better than Kolb, Feeley, and Mike McMahon.

A medicore dude does not become great after 10 years of mediocrity.

Sure he gets points as the 3rd best runner as an Eagles QB, but Cunningham and Vick kill him there.

6.7, 6.6, and 5.7 yards per carry

 

Stats are also fun when the myopic compilation guys cry about Warner and Terrell Davis and you can point to Testaverde's compile stats vs Joe Montana 

TD's 5.6 yards per carry in the playoffs packs a pretty good punch as does Warner's 102.8 playoff passer rating.

 

The most classic destruction by statistics was Dr Z vs Art Monk.

Dr Z accused Monk of 'catching 500 eight-yard-hooks to get first downs which is nice but not worthy of the Hall of Fame.'

If you take away the 500 catches and 4000 yards you are left with 440 catches at 19.82 yards per catch which makes him a first ballot guy as a whole.

Oh man mike McMahon was terrible lmao 

i loved watching when he was with the Lions but terrible QB. 

[Kiwibrown]

On 27/07/2019 at 11:24 AM, Elky said:

 

They rely on analytics in new England.