Analytics vs Game Flow

[sb41champs]

Which do you prefer in a coaching staff:

 

1.  The significant use of analytics in play calling

 

2.  Play calling based on a wide variety of factors - which vary from week to week

[Four2itus]

This is the place to ask. Coaching dialogue in here is as thick as morning's wood smoke in a still meadow ven. 

[chad72]

28 minutes ago, Four2itus said:

This is the place to ask. Coaching dialogue in here is as thick as morning's wood smoke in a still meadow ven. 

 

Can you dumb the analogy down for us simple folks, man?  

[Four2itus]

1 minute ago, chad72 said:

 

Can you dumb the analogy down for us simple folks, man?  

Hey, it's not easy delivering backhanded compliments and visual poetry simultaneously. 

 

(It means there are a lot of armchair coaches in here). 

[Smonroe]

1 hour ago, sb41champs said:

Which do you prefer in a coaching staff:

 

1.  The significant use of analytics in play calling

 

2.  Play calling based on a wide variety of factors - which vary from week to week

It’s both, but #2 is first.  Analytics are cold fact based odds.  They don’t take into account weather, personal matchups, injuries, etc.

[chad72]

2 minutes ago, Smonroe said:

It’s both, but #2 is first.  Analytics are cold fact based odds.  They don’t take into account weather, personal matchups, injuries, etc.

 

Agree. You want a HC that grooms the players to play different styles, pass 40 times or run 40 times, and works with the GM and the other coaches to find such players on O and D that can play such styles or go against such styles.

 

There will always be playmakers that transcend styles but then you have Tier 2 and Tier 3 players on O and D that help more with the versatility including special teams, IMO.

 

 

[stitches]

1 hour ago, Smonroe said:

It’s both, but #2 is first.  Analytics are cold fact based odds.  They don’t take into account weather, personal matchups, injuries, etc.

I mean... in reality you can build analytical models that take those type of things into account. The question is whether or not you are limiting the sample you are drawing from too much to the point of it being less and less reliable. For example... how useful is the information that Darius Leonard has allowed 0% completion in coverage on a Sunday 1pm games in December, when it snows and he's matched up against a TE in the slot out of 21 personnel when the Colts lead with more than 10 points, when he was playing with injured thumb... if this has happened only once - i.e. 0/1 completions? Is this meaningful? Is it predictive in any way? Is it useful for a coach? I'd argue it's not and at the same time it's extremely precise about the situation and conditions of the game and matchups, etc. 

 

That's the thing with analytics... you can have wide variety of different data. The key is knowing what to do with it and how to filter it into USEFUL and actionable information. I'm a huge believer in the power of analytics in sports. But I also think some information produced by analytics can be very noisy to the point of being irrelevant and not useful at all. 

 

I think a coach who does the big things by the book(i.e. analytically sound), but on the margins and in very specific and unique circumstances manages to draw on his own experience and "sixth sense" to make pattern breaking decisions would be a very good coach. In certain respects, I think Reich is already there. Sometimes I wish he would call a more ... in rhythm game, but I also realize that it's probably not always possible due to personnel limitations or execution miscues. I kind of feel like my opinion of Frank as a coach has actually gone up this year. 

 

[Superman]

Why can't you call plays on a wide variety of factors, including analytics?

[WoolMagnet]

6 hours ago, sb41champs said:

Which do you prefer in a coaching staff:

 

1.  The significant use of analytics in play calling

 

2.  Play calling based on a wide variety of factors - which vary from week to week

I believe analytics (in football) are better utilized in gameplanning for an opponent or analysis of your team rather than in-game decisions.  The knowledge of “what has happened in this type of situation” in the past  is helpful in making real-time decisions, but only one tool, along with weather, home or road, score, time, and dozens of other factors.

  I prefer to “feel” the game in any sport really.  Analytics merely give you a benchmark to start the discussion, imo. 

[WoolMagnet]

It also seems Frank is feeling more and leaning less on the #s.

I think its working.

[Superman]

11 hours ago, WoolMagnet said:

It also seems Frank is feeling more and leaning less on the #s.

I think its working.

 

The Colts are still one of the most analytically synced 4th down teams in the league. They go for it nearly 60% of the time the analytics say you should go for it, which is the 6th highest mark in the league.

 

https://twitter.com/benbbaldwin/status/1476208693616259072/photo/1

 

Two things I think are relevant here: 1) no one is making 4th down decisions based solely on what the numbers say.

 

2) Frank Reich seems to strike a healthy balance between leaning on the numbers and making decisions based on the totality of information available to him. For example, his 4th and 4 decision against the Cardinals was probably influenced by the fact that the kicker had already missed a long FG, and he didn't want to try another 50 yarder. If Badgley made the first kick, Reich probably sends him out again.

[2006Coltsbestever]

12 minutes ago, Superman said:

 

The Colts are still one of the most analytically synced 4th down teams in the league. They go for it nearly 60% of the time the analytics say you should go for it, which is the 6th highest mark in the league.

 

https://twitter.com/benbbaldwin/status/1476208693616259072/photo/1

 

Two things I think are relevant here: 1) no one is making 4th down decisions based solely on what the numbers say; 2) Frank Reich seems to strike a healthy balance between leaning on the numbers and making decisions based on the totality of information available to him. For example, his 4th and 4 decision against the Cardinals was probably influenced by the fact that the kicker had already missed a long FG, and he didn't want to try another 50 yarder.

All good points, people can hate on Frank all they want but I have been a Colts fan since 1984 and going by Indy coaches since 84, he is the 2nd best coach we have ever had. Dungy is #1, Frank is a tad better than Marchibroda IMO. Frank brings that gambling factor into it and players love going for it on 4th and short.

 

Franks record:

2018 = 10-6/1-1 playoffs

2019 = 7-9, year Andrew left, we had JB at QB

2020 = 11-5/0-1 playoffs

2021 = 9-6 so far

37-26 = Regular Season, has his playoff win as well in 2018 when we had Andrew Luck.

[DougDew]

17 hours ago, Four2itus said:

Hey, it's not easy delivering backhanded compliments and visual poetry simultaneously. 

 

(It means there are a lot of armchair coaches in here). 

Yeah, I used to find that reading comments from armchair GMs (like me) annoying, until the arm chair coaches started showing me what annoying really looks like.

[2006Coltsbestever]

18 hours ago, sb41champs said:

Which do you prefer in a coaching staff:

 

1.  The significant use of analytics in play calling

 

2.  Play calling based on a wide variety of factors - which vary from week to week

#2 but I will go for it if I have 4th and 2 from the opponents 34 for example. I don't trust our kickers kicking anything over 50 yards. 50 and in I feel ok about it. If I have 4th and inches, I would go for it as well as long as we have the ball at the 50 or better. QB sneak is 90% full proof for a few inches.

[DougDew]

15 hours ago, stitches said:

I mean... in reality you can build analytical models that take those type of things into account. The question is whether or not you are limiting the sample you are drawing from too much to the point of it being less and less reliable.

To get it right, the data universe that must be collected to get to that highly granular level of processing is overwhelming.  You would have to collect every bit of data of every game to account for weather, injuries, etc. and then feed them into a proper algorithm to process.

 

The idea (ideology?) of analytics is based upon the notion that the human brain is biased and nonobjective...which is a false narrative (designed to sell analytics).  The human brain instinctively weeds out biases because it inherently want to get things right.  Some are better at it than others, but bias is not a universal problem.  The brain is a pattern-recognition machine.  But that's another topic.

 

But objectivism and biases advantages aside, gathering the amount of data needed to feed those more granular algorithms would take as much effort, if not more, than just sitting down and watching every play of every game tape of your next opponent.  So you may as well just do the latter, since coaches do that anyway.  The analytics is used as a support structure to see if the coach is somehow being biased or unobjective about something he is watching.  But an overwhelming universe of data would be needed to do it right.

[BlackTiger]

Analytics are really interesting to me.  There are essentially two types of data

 

Data collected from games in the past and data simulated from computer models.  They both have pros and cons and there is nothing wrong with questioning them.  neither is going to be gospel when it comes to in game decisions 

 

Interesting to see the 60% comment above 

[lester]

23 minutes ago, DougDew said:

But objectivism and biases advantages aside, gathering the amount of data needed to feed those more granular algorithms would take as much effort, if not more, than just sitting down and watching every play of every game tape of your next opponent.  So you may as well just do the latter, since coaches do that anyway. 

 

The coaches and players watch the tape. The tier 2 staff builds the regression models.

 

And then image if the coaches and the players had some sort of hand-held machine that could present them with any data on a whim? Maybe even on the sideline during the game? That would be powerful.

 

[DougDew]

3 minutes ago, lester said:

some sort of hand-held machine that could present them with any data on a whim? Maybe even on the sideline during the game?

They already have one, except its skull-held and it involves God-given memory banks.  (or notes that you take when watching tape.)

[IrsaysArmy]

Frank Reich for prez 

[Four2itus]

There are analytics, there is situational decision making, and there is a third thing.....intuition. The last is not reliable at all, but it occasionally does play a part in NFL history, good or bad. 

 

I think that the depth, breadth, and validity of experience, give rise to the proper choice in the moment. Historically, it is my opinion that preparation plays a bigger part in game decision making than the actual in-game choices do.  That is what BB is so damned good at. 

 

Frank is playing 3D chess when it comes to start-game play outlining, and in game play selection. Some of the failures in many fans minds, have quite possibly been tests or even purposeful false info for late season success. I honestly think he is wicked smart. He very well may have pushed limits, created false tendencies, and stretched the boundaries of information available to the oppositions archives. Simply said, I think he has set the team up for a nice late season run. 

[WoolMagnet]

1 hour ago, Superman said:

 

The Colts are still one of the most analytically synced 4th down teams in the league. They go for it nearly 60% of the time the analytics say you should go for it, which is the 6th highest mark in the league.

 

https://twitter.com/benbbaldwin/status/1476208693616259072/photo/1

 

Two things I think are relevant here: 1) no one is making 4th down decisions based solely on what the numbers say.

 

2) Frank Reich seems to strike a healthy balance between leaning on the numbers and making decisions based on the totality of information available to him. For example, his 4th and 4 decision against the Cardinals was probably influenced by the fact that the kicker had already missed a long FG, and he didn't want to try another 50 yarder. If Badgley made the first kick, Reich probably sends him out again.

Whatever the reason, Frank is making “smoother” decisions.

And by that i mean our offensive flow of plays is better. You can see it.  If he was a music conductor, it would be less staccato.

  There were times earlier when a play would come out of nowhere, or an overly aggressive play that failed killed our-momentum or lost the game.

  Going on 4th or not is nothing new.  But Frank seems to be seeing (finally) a bigger picture that includes momentum, fan excitement and noise, and especially the energy and flow of the game at THAT SPECIFIC moment in time.  Not necessarily the average outcomes of past and different situations.

  I still think Frank stubbornly put too much stock (at times) in the analytics.  I used to hate when he would say “the numbers said go.”  I’d think, “no, the numbers say in more cases than not, this should work.  But in my opinion, these numbers are usually too broad to directly relate them to a specific play, by specific players, etc.   

  Before a round of golf, i sometimes putt on the practice green.  Typically on courses i dont know.  I use it as a way to gauge the possible outcomes on the 18 similar but not exact greens.  I use the info, along with other unique  indicators on each green.

  At times, it seemed Frank looked at a chart, an made a decision based on which side of 50% the “numbers”fell on.

  It has felt different, inna good way, recently.  Less “trick”-type plays too.

[Superman]

26 minutes ago, Four2itus said:

Frank is playing 3D chess when it comes to start-game play outlining, and in game play selection. Some of the failures in many fans minds, have quite possibly been tests or even purposeful false info for late season success. I honestly think he is wicked smart. He very well may have pushed limits, created false tendencies, and stretched the boundaries of information available to the oppositions archives. Simply said, I think he has set the team up for a nice late season run. 

 

You're talking about the guy who went for it on 4th and 4, in OT, in his own territory, to avoid a tie. The same guy who went for it on 4th and 4 against the Cardinals, even though his OL was basically swiss cheese. These are not false tendencies. Frank Reich believes in going for it on 4th down. He doesn't stick strictly to the chart, but he definitely is as close to the cutting edge as any NFL coach right now, and he has a four year history of this.

[WoolMagnet]

44 minutes ago, lester said:

And then image if the coaches and the players had some sort of hand-held machine that could present them with any data on a whim? Maybe even on the sideline during the game? That would be powerful.

 

Robot coach!

[Superman]

3 minutes ago, WoolMagnet said:

Whatever the reason, Frank is making “smoother” decisions.

And by that i mean our offensive flow of plays is better. You can see it.  If he was a music conductor, it would be less staccato.

  There were times earlier when a play would come out of nowhere, or an overly aggressive play that failed killed our-momentum or lost the game.

  Going on 4th or not is nothing new.  But Frank seems to be seeing (finally) a bigger picture that includes momentum, fan excitement and noise, and especially the energy and flow of the game at THAT SPECIFIC moment in time.  Not necessarily the average outcomes of past and different situations.

  I still think Frank stubbornly put too much stock (at times) in the analytics.  I used to hate when he would say “the numbers said go.”  I’d think, “no, the numbers say in more cases than not, this should work.  But in my opinion, these numbers are usually too broad to directly relate them to a specific play, by specific players, etc.   

  Before a round of golf, i sometimes putt on the practice green.  Typically on courses i dont know.  I use it as a way to gauge the possible outcomes on the 18 similar but not exact greens.  I use the info, along with other unique  indicators on each green.

  At times, it seemed Frank looked at a chart, an made a decision based on which side of 50% the “numbers”fell on.

  It has felt different, inna good way, recently.  Less “trick”-type plays too.

 

That seems to be your perception. I don't think there's any data that backs it up, though.

 

For example, the chart I posted above shows that Reich does NOT just look at a chart and make a decision based on what the chart says. He never has. If we judged strictly by the analytic "go or no" chart, every NFL coach is overly conservative on 4th down (and 2 pt conversions, btw). 

 

He has always included those other factors into his decision making. For example, he was less likely to go for it in 2019 on 4th and 2+ with JB than in 2018 with Luck, or 2020 with Wentz. He was very much into "go for it" mode last year, because he had some JB packages that worked pretty well. Some of those packages seemed gimmicky, but they were very successful, percentage wise.

 

I also think people put way too much stock into what Frank Reich says in press conferences. It's 90% coachspeak. He puts the blame on himself, will never throw his players under the bus, and will never undermine their confidence publicly. He's also probably not going to fully explain all of his motivations when he can give a buzzword answer and move on.

 

I do agree that some of his 4th down decisions this year have been "smoother." He hasn't done the thing where they line up and hard count on 4th down, then call timeout, then line up and go for it on 4th down. He did that often in the past, and it drives me crazy. 

 

But in general, I think the play calls have been coming in slowly, and we're slow to break huddle and line up, and then barely getting the play off. There was a play Saturday where this happened, then the Cardinals showed a late blitz, Wentz didn't have time to adjust the protection, and the pressure got home. So this can be tightened up also.

[CurBeatElite]

19 hours ago, Smonroe said:

It’s both, but #2 is first.  Analytics are cold fact based odds.  They don’t take into account weather, personal matchups, injuries, etc.

 

They can pretty easily access this data and I would be very shocked if NFL teams aren't including these things in their models. As a fantasy football manager, I can take these things into account for guys on 32 different teams (sports betting allows bets to be made in real time based on various scenarios for lay people like us posters). I can almost guarantee that Reich can get far more information specifically about the Colts and their opponent and conditions he is standing in on the sidelines than I can on my computer from 2000 miles away.

 

Accessing data in real-time is becoming easier almost on a daily basis in all sorts of fields.  As @stitches pointed out, the bigger issue is choosing what information to use and when to use it.. in the project management world we are in the Era of 'information overload', with access to more data about more stuff than ever before.. the hard part is not getting bogged down by this and narrowing the focus to the data that is actually going to lead to informed decision making.

 

2 hours ago, DougDew said:

To get it right, the data universe that must be collected to get to that highly granular level of processing is overwhelming.  You would have to collect every bit of data of every game to account for weather, injuries, etc. and then feed them into a proper algorithm to process.

 

The idea (ideology?) of analytics is based upon the notion that the human brain is biased and nonobjective...which is a false narrative (designed to sell analytics).  The human brain instinctively weeds out biases because it inherently want to get things right.  Some are better at it than others, but bias is not a universal problem.  The brain is a pattern-recognition machine.  But that's another topic.

 

But objectivism and biases advantages aside, gathering the amount of data needed to feed those more granular algorithms would take as much effort, if not more, than just sitting down and watching every play of every game tape of your next opponent.  So you may as well just do the latter, since coaches do that anyway.  The analytics is used as a support structure to see if the coach is somehow being biased or unobjective about something he is watching.  But an overwhelming universe of data would be needed to do it right.

 

I get what you are saying and appreciate this post. I agree that a goal of analytics is to remove human bias.

 

That said, there is a ton of data available and many ways to sync databases together (eg, a barometer and wind gauge at the stadium could easily feed into a relationship database with info about how many carries a rb has or if players are injured, etc.).

 

There are also ways to deal with data analysis in terms of binary (i.e., 0 or 1, yes or no) and fuzzy (i.e., gradient, like a 0-10 scale) logic.  

 

Getting all the data is never going to be possible, but there sure is a darn lot of it available very rapidly and it is likely easily accessible, especially to NFL teams. 

 

There are literally millions of ways to build algorithms with as many data inputs as the NFL can access.  Choosing the right way to deal with data and how to plug them into an algorithm and apply the results for a quick decision is the hard part. There are likely lots of data available which is meaningless and other data which varies in importance from time to time. 

 

There are also data which probably are so common sense that they don't need to be put into a model. For example, wind speed can be 0 mph to 100+mph.  On a 4th and 3 from the 35 yard line with 60 mph wind in your face, you can probably throw the computer on the ground and decide a 52-53 yard field goal is a bad choice. 

 

And the one thing not measurable is a player's mindset. For example, there have been a few games where we've probably passed too much. It may take a Quentin Nelson telling Reich 'hey, look man. I am going to pancake this DT and cause tectonic plates to move if you just call a power run or two off the left guard. '

 

Long story short, there are so many ways to deal with data which can inform analytics and decision making.  Dealing with data can be subjective in ways to aim at an objective answer (eg, when to use binary vs fuzzy logic).  In a game with so many moving parts and variables, the correct decision is rarely going to be 100% handed to someone in the role of the decision maker. I think Reich is doing a dang good job. 

[NewColtsFan]

21 hours ago, sb41champs said:

Which do you prefer in a coaching staff:

 

1.  The significant use of analytics in play calling

 

2.  Play calling based on a wide variety of factors - which vary from week to week

Both.   
 

It doesn’t have to be one or the other.   It’s both. 

[Four2itus]

2 hours ago, Superman said:

 

You're talking about the guy who went for it on 4th and 4, in OT, in his own territory, to avoid a tie. The same guy who went for it on 4th and 4 against the Cardinals, even though his OL was basically swiss cheese. These are not false tendencies. Frank Reich believes in going for it on 4th down. He doesn't stick strictly to the chart, but he definitely is as close to the cutting edge as any NFL coach right now, and he has a four year history of this.

I wasn't speaking about Frank going for it.....it was more about too many passes in a row thing. Sorry to not give a better explanation. 

[Superman]

25 minutes ago, Four2itus said:

I wasn't speaking about Frank going for it.....it was more about too many passes in a row thing. Sorry to not give a better explanation. 

 

Ah got you.

[Jason_]

On 12/29/2021 at 10:54 AM, DougDew said:

Yeah, I used to find that reading comments from armchair GMs (like me) annoying, until the arm chair coaches started showing me what annoying really looks like.

 

 

Aww, poor little guy

 

[CoachLite]

On 12/28/2021 at 6:52 PM, Superman said:

Why can't you call plays on a wide variety of factors, including analytics?

Please, don't let stupid people do analytics.