When we talk about 'analytics' in the nfl.....

[Tacosman]

we really have to define what that even means with respect to the nfl.  

Because the definition simply is: information resulting from the systematic analysis of data or statistics.

Obviously everyone does this and has always done this, with the essence being what should we be concentrating on and finding a perfect fit between the system used and the strengths of our team with that data.  

Lets start with what it means(or really meant) with the baseball revolution that took place about 15-20 years ago: when people talk about moneyball, they arent simply talking about analytics.  For decades upon decades people in baseball were pouring over stats and data and getting information and making decisions about it.  The problem was they were focusing on the wrong data and the wrong statistics.  You can analyze it to death, but if you're doing it the wrong way it's going to be f'd up.  So this was the foundation of moneyball- moneyball wasn't "let's focus on the analytics".  moneyball was "we are focusing on the wrong stats now".  Thats an oversimplified version of course(lots of other things are part of it...positional value, scouting trends, defense independent pitching stats, etc), but you get the gist.  

What 'the process' meant in the nba(again, due to the nature of the nba) was that acquiring elite/transcendant players was essential to become a team with a title shot and the best chance to acquire those players is through very high(possibly #1) lottery picks.  And that there is no value in being somewhere between the 7th and 26th or so best team in the league....better to be terrible than mediocre or slightly below average because it gives you a slightly better chance to be great later and the chance of being great is all that matters.  

So again, 'moneyball' in baseball and 'the process' in the nba are both kind of lumped together sometimes, but they are radically different because of the very different nature of the nba and major league baseball.  If Lebron and Steph Curry happen to be on the same team that team is going to win/compete for an nba title.  If Mike Trout and Matt Scherzer are on the same team it's certainly awesome to have two superstars like that, but we don't know about their chances to win the world series without a lot more info.  Again very different leagues.

Now the nfl....what is the 'moneyball' and 'the process' equivalent in the nfl?  What does that look like?  I don't think anyone has figured that out yet....maybe there isn't one?  Meaning maybe there isn't a defining philosophy on how to most efficiently build championship teams.  

Interestingly, another component of nba analytics over the last several years is much more data on what is a 'high value' shot now than before.  Hence all the talk about how value open corner 3's off a catch are.  It's all about trying to find ways to garner the most ppp(points per possession) as possible,  and devising matchups and an offense around that.  We're not there yet in football and the nfl. 

But obviously the goal in any sports league today is to identify where you gain the most value at certain positions above replacement level for a certain cost.  Where can I 'get by' with a C+/B- guy and it won't kill me so I can pour my resources into finding an A guy where there is more value difference?  It seems like the patriots do this better than anyone....understanding that certain positions are fungible and others not so fungible.  

 

[Unsympathetic]

sports illustrated has done multiple feature-length articles. And this SI link has 10 sub-links each to their own article on areas of NFL analytics.

ESPN has done a feature report.

Here's a compilation of top analytics articles.

 

Here's the link to the comparison between Browns rebuild and Astros/76ers.

[Tour2ma]

Nice summary... stimulating thread...

My take on the baseball bit... In the beginning with Bill James in the bean factory, it wasn't the wrong data, it was the only data. And the stats weren't the wrong ones, they just weren't mathematically related to anything, i.e., wins or even runs scored. OBP, which emerged as an offensive key, wasn't new. It's import vs. BA or Slugging Percentage was simply unrecognized. Bill's regression analyses brought it to the fore along with the counter-productivity of bunting, steals, etc.

Add that in the beginning, when Bill completely deemphasized defense, it was not for lack of import, but for lack of meaningful data. Now, 20 years later, with new data gathering techniques, it's back as part of the equation. Just one indication of the evolving process that is Baseball Analytics.

 

One thing I strongly disagree with is your take on the goal of analytics. In baseball it was and in football still is not to " 'get by' with a C+/B- guy", but unearthing guys viewed as, and paid as, B's and C's, or D's, who can be A's if used properly. Hard to do in baseball now as nearly everyone is at least at the original Bill James' level, but still wide open in the NFL... and will be for quite a while. Just due to the nature of whole vs. sum of the parts of the two sports where baseball is at one extreme and football the other.

I'm hard pressed to think of a "team sport" further out on either extreme of the "lends-itself-to-analytics" continuum, but there may be one. Certainly basketball is not... While LBJ and Steph would be a great start to a team, it's be the next seven players rostered that would determine their Championship potential. And given what those two are paid, you'd better have a good analytics department to ID that next seven.

[BrownPile]

Nice thread.  You could open several large cans of worms with this topic.

The rise of use of Data Analytics in all forms of business is natural, given the evolutionary point we are at with the use of automation.  Like everything else, it will be overused, as well.  It's a very American thing to do... if some is good, more must be better and a lot more must be a lot better.  We are seeing this come to fruition now with Analytics.  In the sports world, we've arrived at the point where everything must be quantified to a micro level - the 'a lot more data must be a lot better'.

However, accumulating data, establishing norms and ranking existing players still leaves you at the point where a human must make decisions based on those results, following a set of criterion, in order to be achieve an established goal.

With Moneyball, the emphasis was really on the 'Money' part.  I want to score runs and scoring requires getting on base, but I also have far less money than other teams.  Having procured my data, I also assign a value to each ranked player based on pay.  So, if I have two players with an OBP of .350, I should be able to achieve the same results using the player making less money.

In evaluating that 'process', I can see that I could also remove the human from the evaluation and program a machine with the criterion and using the collected data run every conceivable scenario.  I'll further refine the 'process' to emphasize recent data (exponential averages) and account for variance (standard deviations) and also start accounting for player health and age.

Now, I'm at Predictive Analytics.

in your NBA scenario, i can guess that Predictive Analytics would tell you that populating a team full of currently ranked 'A' players (Lebron, Curry and others) would actually cause all of their future statistics to decline (pick a stat here... all 15 of them won't get as many per game, thus their production drops, thus their exponential averages decline, etc.)  Plus, it's most likely that I can't afford a team fully populated with current 'A' players.  No, my goal would more likely be to identify the players whose exponential averages show them moving on the trend lines from 'C to B' or 'B to A' players.  I can run all the scenarios to see which combination of players provides me with a team that can win most games and make the playoffs - possibly win a championship - as well as being one that I can afford.

I can also take the norms established and apply those to future players.  A player of this height, weight, speed, etc. falls within the established norms of current players and should be able to produce at an adjusted level, since the 'A' player in college isn't competing against the current professional ranks.

This discussion could go on at length.

One thing I would note, here.  As the use of programmed logic proliferates, the humans are removed from all decision-making points.  So, why would I need to employ and pay anyone millions to merely follow those outputs produced by the 'process'.

 

[Unsympathetic]

2 hours ago, BrownPile said:

As the use of programmed logic proliferates, the humans are removed from all decision-making points

"Tech" does not, cannot, and never will assign value on its own.. na ga happen.  Every value used by "analytics" .. was assigned by a person or consensus-based group.  Humans define the usefulness of traits -- each team in fact uses a different formulation.

Remember all the assertion about dictation from voice to text? Yeah, that will be 100% accurate precisely never.

AI won't be taking over driving either -- it's not difficult to drive between the lines when it's sunny and 90 degrees and the edges of the road&lanes are painted correctly.  Humans remain entirely required for the boundary conditions when driving.  This includes both location and road/lane boundaries despite the money poured into it by Waymo and Google et al because both of the error built into civilian GPS and the error of perceiving edges of the road if the road isn't painted or there's construction or if there's weather etc.

[LondonBrown]

I have an analytics based 2017 draft guide, I think it’s quite informative regards how teams are looking at things “analytically”. It’s definitely bigger than 500mb though so I’ll try and upload it offline 

[LondonBrown]

Well it’s nowhere near 500mb so that’s my technology knowledge exposed. Should be attached. 

2017 nfl draft guide giveaway copy.pdf

[Dutch Oven]

4 minutes ago, LondonBrown said:

Well it’s nowhere near 500mb so that’s my technology knowledge exposed. Should be attached. 

2017 nfl draft guide giveaway copy.pdf

Did you increase your export ability on your flux capacitor before attaching?

[Tacosman]

On 11/18/2017 at 7:56 AM, Tour2ma said:

Nice summary... stimulating thread...

My take on the baseball bit... In the beginning with Bill James in the bean factory, it wasn't the wrong data, it was the only data. And the stats weren't the wrong ones, they just weren't mathematically related to anything, i.e., wins or even runs scored. OBP, which emerged as an offensive key, wasn't new. It's import vs. BA or Slugging Percentage was simply unrecognized. Bill's regression analyses brought it to the fore along with the counter-productivity of bunting, steals, etc.

Add that in the beginning, when Bill completely deemphasized defense, it was not for lack of import, but for lack of meaningful data. Now, 20 years later, with new data gathering techniques, it's back as part of the equation. Just one indication of the evolving process that is Baseball Analytics.

 

One thing I strongly disagree with is your take on the goal of analytics. In baseball it was and in football still is not to " 'get by' with a C+/B- guy", but unearthing guys viewed as, and paid as, B's and C's, or D's, who can be A's if used properly. Hard to do in baseball now as nearly everyone is at least at the original Bill James' level, but still wide open in the NFL... and will be for quite a while. Just due to the nature of whole vs. sum of the parts of the two sports where baseball is at one extreme and football the other.

I'm hard pressed to think of a "team sport" further out on either extreme of the "lends-itself-to-analytics" continuum, but there may be one. Certainly basketball is not... While LBJ and Steph would be a great start to a team, it's be the next seven players rostered that would determine their Championship potential. And given what those two are paid, you'd better have a good analytics department to ID that next seven.

but the question is WHAT is the analytical key to unearthing these A guys in the nfl paid as C guys?  We knew what it was in baseball.  Do we know what it is in football?  Does anyone?  Certain organizations obviously seem better at it than others....

Also in basketball team with LBJ and Steph would have championship potential right there.  That is the whole fundamental nature of the nba right now- you win with great 12+ win over average players like LBJ and steph.....get two like that on any one team and it's all downhill from there.

[Tacosman]

In the nba scenario, the 'process' part of the equation is the understanding that unlike in the nfl or mlb you don't win titles by accumulating lots of 'good' players....rather you win titles by getting a true alpha player who changes everything....more than one if possible.  Basketball is using a ton of analytics in terms of trying to find out offensive strategies(hence the rise of the corner three), but the team building aspect is separate from this

[BrownPile]

9 hours ago, Unsympathetic said:

"Tech" does not, cannot, and never will assign value on its own.. na ga happen.  Every value used by "analytics" .. was assigned by a person or consensus-based group.  Humans define the usefulness of traits -- each team in fact uses a different formulation.

Remember all the assertion about dictation from voice to text? Yeah, that will be 100% accurate precisely never.

AI won't be taking over driving either -- it's not difficult to drive between the lines when it's sunny and 90 degrees and the edges of the road&lanes are painted correctly.  Humans remain entirely required for the boundary conditions when driving.  This includes both location and road/lane boundaries despite the money poured into it by Waymo and Google et al because both of the error built into civilian GPS and the error of perceiving edges of the road if the road isn't painted or there's construction or if there's weather etc.

I will have to disagree with the notion that this won't ever happen.  Some of it is still being developed right now, while some is actually already there.  Nothing is 100% accurate.  The car still replaced the horse, though, even though it isn't 100% safe.

However, we are also talking about different things here, too.  True AI would be cognitive thought.  Pretty far off from where we are now.

But Analytics and logic - let's call it Reasoning - has been around since the birth of the computer.  Some of you guys will remember the logic classes from high school or college... If (P and Q) and NOT R then Y.  Humans still define the thresholds, but your thermostat, CO detector, washing machine, coffee pot, car, refrigerator, PC, phone are all making decisions based on data.  To program a machine to account for a new variable of future data and make it relative to wins/losses and/or any other known data... thereby, programming a machine to assign the thresholds (or assign values - as you noted) is already being done.

One of my bigger points, though, in referring to how Analytics are being misused in sports today, is that people are focusing on the minutiae, as well as creating data sets from things that currently can't be quantified - the whole notion that 'a lot more data must be a lot better'.  The Defensive Runs Saved in Baseball is an example, where one of the components is 'range' of the defender.  No objective way to measure it is being employed, yet people are selling it as a valid statistic.  In the NFL, does it really matter that someone's wingspan is 1 cm longer than the next guy?

Further, I hoped to point out that I may not need a Sashi to evaluate data, when it is entirely possible to program the evaluation of this mass of data on a machine.  It's always been the case that human analysts/statisticians can manipulate the interpretation or presentation of data to reinforce whatever notion they wish - even to justify their own existence to the organization.

[BrownPile]

2 hours ago, Tacosman said:

but the question is WHAT is the analytical key to unearthing these A guys in the nfl paid as C guys?  We knew what it was in baseball.  Do we know what it is in football?  Does anyone?  Certain organizations obviously seem better at it than others....

Also in basketball team with LBJ and Steph would have championship potential right there.  That is the whole fundamental nature of the nba right now- you win with great 12+ win over average players like LBJ and steph.....get two like that on any one team and it's all downhill from there.

I'm thinking that you're looking for a 'silver bullet' here.  ONE piece of data to measure.  That may not exist.

There isn't a single piece of data or metric in either MLB or the NBA, so I don't think it exists in the NFL, either.  One big reason this is harder in the NFL is that a player doesn't play on both sides of the ball - both offense and defense.

Your team wins in these sports by scoring points and keeping the other team from scoring points.  In Baseball and Basketball, each player can be evaluated on both abilities of scoring and defense.  Not so in the NFL.

[Tacosman]

1 minute ago, BrownPile said:

I'm thinking that you're looking for a 'silver bullet' here.  ONE piece of data to measure.  That may not exist.

There isn't a single piece of data in either MLB or the NBA, so I don't think it exists in the NFL, either.

of course not.  But Im not sure in the nfl we are on to the same key aspects of things as mlb and the nba is.  It's possible that different organizations which are thought to be analytics heavy are coming at things from very different ways for example.  

You can be an organization hell bent on being analytical, and then just doing it all wrong.  

 

[BrownPile]

17 minutes ago, Tacosman said:

of course not.  But Im not sure in the nfl we are on to the same key aspects of things as mlb and the nba is.  It's possible that different organizations which are thought to be analytics heavy are coming at things from very different ways for example.  

You can be an organization hell bent on being analytical, and then just doing it all wrong.  

 

Since most organization employ their own Data Analysts, then they are all coming at it in different ways.  They have to sell themselves to the owners as having a keener insight into the data than everyone else.  Moving this evaluation to a machine would alleviate that and negate the need for these Analysts.

I did edit my original answer to include a difference in the NFL than either MLB or the NBA.  Players in the NFL don't play both offense and defense, therefore their contribution to the team's ability to score or prevent scoring can't be evaluated the same.  Thus, too, the rise of the never ending debate over which is more important in the NFL... scoring or preventing the other team from scoring.

[flyingfooldoug]

How does anal ytics deal with human nature? Or the  "football field awareness" of a player that numbers just can't cut it?  It doesn't and that's why we have zero playmakers from this FO. 

Oooooh analytics! New term, old BS. Numerology is for losers. 

[Tour2ma]

12 hours ago, Tacosman said:

but the question is WHAT is the analytical key to unearthing these A guys in the nfl paid as C guys?  We knew what it was in baseball.  Do we know what it is in football?  Does anyone?  Certain organizations obviously seem better at it than others....

Also in basketball team with LBJ and Steph would have championship potential right there.  That is the whole fundamental nature of the nba right now- you win with great 12+ win over average players like LBJ and steph.....get two like that on any one team and it's all downhill from there.

No idea what the key is... and agree with above comment that "it" may be a combination of many factors. When you add that in football those factors change with every position... that there's no commonality across our game that exists as it does with batting in Baseball... and that is the complexity that is football.

But the safe bet is that DePo and Co., as are the analytic arms of some other organizations, are a lot closer to knowing than we are. Amy have even nailed a few of the "simpler" positions, e.g., OL. Positions where even the eyes of rank amateurs like myself can rack up a reasonably good track record.

12 hours ago, Tacosman said:

In the nba scenario, the 'process' part of the equation is the understanding that unlike in the nfl or mlb you don't win titles by accumulating lots of 'good' players....rather you win titles by getting a true alpha player who changes everything....more than one if possible.  Basketball is using a ton of analytics in terms of trying to find out offensive strategies(hence the rise of the corner three), but the team building aspect is separate from this

All true, but the NBA formula is a moving target. Not long ago a single "Jordan" or "Bird" plus role players added up to a sustainable winning formula.

Then it became a two-player formula, e.g., Shaq plus Colby... or Colby plus Shaq depending on which you talked to, Magic+Karrem. Nothing really new as two greats have combined throughout NBA history and won, e.g, Alcindor+The Big O, Wilt+West.

Then it was three... LBJ+Wade+Bosh or LBJ+Kyrie+Love...

Then the Warriors blew off the lid off of Curry+Green+Thompson by adding Durrant... and now everyone is scrambling... Cavs, Celtics, Thunder... in a five-man game, not our 25+ game.

[The Gipper]

15 hours ago, LondonBrown said:

Well it’s nowhere near 500mb so that’s my technology knowledge exposed. Should be attached. 

2017 nfl draft guide giveaway copy.pdf

How about synthesizing down to about 2 paragraphs for us.

[Browns149]

Analytics has not worked so far.  And no matter how many decent to good players you have, when you don’t draft a franchise Qb when he is sitting right in front of you, you should be fired

[flyingfooldoug]

The key to finding great football players seems fairly easy to me. It takes one to know one.

 Hire BK as GM. End the BS now!

[flyingfooldoug]

Numbers to crunch. Drennan held a online survey today.

37% think that Modell moving to Bloitmore was the lowest point in team history.

1% think we can go lower.

62% feel this is the lowest it can possibly get.

Consistent double digit loss seasons. 12 out of 13

Pettine and Hoyer gave us the best year we had in 13 years. 7 wins. It'll take these bozos 8 years to get 7 wins.

All the other Superbowl winning teams did it without relying exclusively on analprobing...er, ytics. You remember? GM hires a coach then provides him with a roster. They don't use the clusterscrew method that Jimmah seems so enamored with.

 

[wargograw]

60% think 3-13 was just awesome!

[flyingfooldoug]

2 minutes ago, wargograw said:

60% think 3-13 was just awesome!

Where and when?

[flyingfooldoug]

2 minutes ago, wargograw said:

60% think 3-13 was just awesome!

More bs and lies for the fan base. To say 60% of browns fans were happy with 3-13 is nothing but a bald faced lie

[Ghoolie]

Analytics is horse-schidt. It is something invented to shield executives when they make bad mistakes.

" The analytics showed that this was a good pick" so don't fire me.

Horse-schidt.............TOTAL USELESS HORSE SCHIDT

Proof? Nobody spends more on analytics than the Browns.

HORSE SCHIDT invented by people who don't know WTF they are talking about.

[DieHardBrownsFan]

17 minutes ago, wargograw said:

60% think 3-13 was just awesome!

And you think 0-10 is awesome.

 

[wargograw]

1 hour ago, flyingfooldoug said:

More bs and lies for the fan base. To say 60% of browns fans were happy with 3-13 is nothing but a bald faced lie

It's a joke. 

1 hour ago, DieHardBrownsFan said:

And you think 0-10 is awesome.

 

No, I think where we're headed is superior to winning 4 games a year.

[stillmotion]

2 minutes ago, wargograw said:

No, I think where we're headed is superior to winning 4 games a year.

winning 5-6 games a year in 2019?

[Tour2ma]

8 hours ago, Browns149 said:

Analytics has not worked so far.  And no matter how many decent to good players you have, when you don’t draft a franchise Qb when he is sitting right in front of you, you should be fired

Nevermind...

7 hours ago, flyingfooldoug said:

The key to finding great football players seems fairly easy to me. It takes one to know one.

 Hire BK as GM. End the BS now!

How many "great players" are currently GMs?

Ozzie... Elway... Lynch (maybe)...

How many are steering great teams currently?

[flyingfooldoug]

4 hours ago, Tour2ma said:

Nevermind...

How many "great players" are currently GMs?

Ozzie... Elway... Lynch (maybe)...

How many are steering great teams currently?

Looks like a pretty good list. Ya, zero playmakers and Superbowls amongst them, ....just like the Browns?..thank you for helping me make my point. I'd take any of them in a heartbeat

[Nero]

1 hour ago, flyingfooldoug said:

Looks like a pretty good list. Ya, zero playmakers and Superbowls amongst them, ....just like the Browns?..thank you for helping me make my point. I'd take any of them in a heartbeat

As a player, Elway has 2 rings, and Lynch 1. Ozzie has won it as an executive, and participated in 3 Pro Bowls as a player. I don't know if you were talking ironically about the Superbowls or not. 

We still don't know how good of a manager Lynch is. But if he's a good one, let's say we have 3 GOOD PLAYERS being good GMs in the current NFL. 

It is 3 out of 32. 9,375%. That's Tour's point. Are going to take the risk and take one for the Browns, when the odds are against us? Being a good player doesn't automatically make you a good executive.