Analytics and Football. Convince me please

OrangeGravy · Oct 20, 2021

How are analytics more than just passively relevant to football? I genuinely want someone to convince me. 17 game seasons with a single elimination tournament? 11 men trying to execute a play while 11 others physically fight them to keep them from executing that play. Physical and talent mismatches all over the place and never in the same place from play to play or game to game. How can you possibly count on probabilities when there are that many uncontrollable variables and with such a small sample? I admit I haven't taken a deep dive into it, but just superficially, I can't see it being that relevant outside of lumping multiple seasons together. I'm sure it helps assist in some areas, but can you really run a franchise on it and expect to be successful?

Im a former college baseball player and it makes perfect sense in baseball even if it has ruined the fan experience. You have 1 single exchange of batter and pitcher that the entire game pivots off of with 162 games played. It's easy to boil down the numbers to what should happen when this specific batter faces this specific pitcher. You can even boil it down a specific pitch from said pitcher to said batter. It works in from just about any angle you want to use it for baseball.

Again this isn't meant to bash analytics. I'm genuinely curious what you stats guys think about it. Do people think it should dominate decision making or should it be 1 of many things to take into account?

LTH · Oct 20, 2021

OrangeGravy said:
How are analytics more than just passively relevant to football? I genuinely want someone to convince me. 17 game seasons with a single elimination tournament? 11 men trying to execute a play while 11 others physically fight them to keep them from executing that play. Physical and talent mismatches all over the place and never in the same place from play to play or game to game. How can you possibly count on probabilities when there are that many uncontrollable variables and with such a small sample? I admit I haven't taken a deep dive into it, but just superficially, I can't see it being that relevant outside of lumping multiple seasons together. I'm sure it helps assist in some areas, but can you really run a franchise on it and expect to be successful?

Im a former college baseball player and it makes perfect sense in baseball even if it has ruined the fan experience. You have 1 single exchange of batter and pitcher that the entire game pivots off of with 162 games played. It's easy to boil down the numbers to what should happen when this specific batter faces this specific pitcher. You can even boil it down a specific pitch from said pitcher to said batter. It works in from just about any angle you want to use it for baseball.

Again this isn't meant to bash analytics. I'm genuinely curious what you stats guys think about it. Do people think it should dominate decision making or should it be 1 of many things to take into

I by no means am an analytics expert but here is my take on it. I think stats are not the tell all. Half way through the season you can't judge a football team on what they did the first part of the season because they are a different team through player growth, through injuries, through personal changes, through scheme adjustments. You can use those stats to monitor progress and change but if you were to judge a team on who they were 5 games ago as to who they are now it would in most cases be a mischaracterization...

I think stats are a really good tool for growth but I don't think they are an accurate account in every circumstance .

LTH

flv · Oct 20, 2021

You should be looking for every legal way to improve your chances of winning. If analytics help you to do that then embrace them. If they don't then don't. For the most part analytics help you make decisions in very specific scenarios. For example the Eagles were down 14 points late in the 4th quarter to the Buccaneers. Other things being equal the analytics say going for a 2 point conversion rather than kicking the XP gives you about a 3% edge, IE it clearly and demonstrably increases your chances of winning from 50% to about 53%. You still have to score another TD and play defense, but that edge is there...if you choose to take it.

I don't see any specific analytics in regards to running a franchise, other than say, prioritising certain player positions high in the Draft. Obviously no-one should be taking a FB over a QB with a top 3 pick just because the FB is ever so slightly better. If there are such analytics i'd be open to them.

OrangeGravy · Oct 20, 2021

flv":he8f93yo said:
You should be looking for every legal way to improve your chances of winning. If analytics help you to do that then embrace them. If they don't then don't. For the most part analytics help you make decisions in very specific scenarios. For example the Eagles were down 14 points late in the 4th quarter to the Buccaneers. Other things being equal the analytics say going for a 2 point conversion rather than kicking the XP gives you about a 3% edge, IE it clearly and demonstrably increases your chances of winning from 50% to about 53%. You still have to score another TD and play defense, but that edge is there...if you choose to take it.

I don't see any specific analytics in regards to running a franchise, other than say, prioritising certain player positions high in the Draft. Obviously no-one should be taking a FB over a QB with a top 3 pick just because the FB is ever so slightly better. If there are such analytics i'd be open to them.

Going for the 2pt conversion gives you the edge or converting it? You have much better chance of winning if convert for 2 every time you score a TD. What are the odds of converting the 2pt play? How do those odds change if you factor in opponent and how they're performing that game on defense? This is where I struggle with it. All things being equal doesn't exist in one football game to another.

flv · Oct 20, 2021

OrangeGravy":1br9hrzn said:
Going for the 2pt conversion gives you the edge or converting it? You have much better chance of winning if convert for 2 every time you score a TD. What are the odds of converting the 2pt play? How do those odds change if you factor in opponent and how they're performing that game on defense? This is where I struggle with it. All things being equal doesn't exist in one football game to another.

I don't want to get into an argument of how much less likely you'd be to convert a 2nd 2 point conversion because obviously there's a drop-off between your best 'secret play' and the 2nd best, etc. The analytics say:

Part 1:

If the average chance of kicking the XP is 100% and each team has an equal chance of winning the game should it go to OT then the average conversion rate on the 2 point attempt only needs to be 38.2% to be a break-even proposition in terms of winning the game.

Kicking XPs will give you a 0% chance of winning in regulation, a 0% chance of losing in regulation and a 100% chance of going to OT.

The alternative strategy is to go for 2 points on the 1st attempt and if you make it kick the 2nd XP. If you fail the 1st attempt you attempt it again on the 2nd TD. This strategy will result in you winning by 1 point in regulation 38.2% of the time, you will lose by 2 points in regulation 38.2% of the time, and you will got to OT 23.6% of the time.

Teams are less than 100% on XPs and they are generally better than 38.2% on 2 point conversions. Thus they are better off going for 2 points in this specific situation. Essentially they are gambling away the chance to lose by 1 point for extra for additional chances to win. Over the long run the team would score fewer total points by going for 2 if they have a conversion rate of between 38.2% and 50% but they would have a higher winning %.

Part 2:

If the 2 point conversion average was 40% and the XP average was 100% going for 2 would win the game 0.4 X 1.0 X 100% = 40%. If the team missed the 1st 2 point conversion but made the 2nd to tie the game then a tied game going to OT would be 0.6 X 0.4 X 100% = 24%. Missing both conversions and thus losing the game by 2 points would be 0.6 X 0.6 X 100% = 36%. Having a 40% vs 36% advantage would clearly be better than a 50% vs 50% stalemate and certain OT. If the conversion rate was above 40% then it's even more advantageous. Obviously XPs are less than 100% so that also makes going for 2 more advantageous.

Again, this only applies to the unique scenario of needing 2 TDs late in the game. If it was the 3rd quarter you wouldn't do this because 1) in the long run kicking the XP yields a higher total points haul, and 2 other scores are likely in the game.

Fade · Oct 20, 2021

If the Seahawks used analytics instead of the random streams of consciousness dribbling out of the brain of Pete Carroll, they wouldn't have sold the farm for Jamal Adams.

Convinced yet?

Palmegranite · Oct 21, 2021

Or as the great comedian and sports gambling prognosticater, Norm MacDonald used to say,

"the big reason anything can happen in the NFL compared to other sports is because the ball ain't round. If a ball is round and going into a round thing, it's more likely to go according to form. But if the ball's not round, and then it gets loose, you don't know what's going to happen. The ball can go anywhere. Know what I mean?"

To sum up, a prolate spheroid shape makes for poor analytics.

OrangeGravy · Oct 21, 2021

Fade":1stxk01q said:
If the Seahawks used analytics instead of the random streams of consciousness dribbling out of the brain of Pete Carroll, they wouldn't have sold the farm for Jamal Adams.

Convinced yet?

No. It makes sense to evaluate players cause you're using their entire careers worth of data points. In game stuff is more convoluted. There should be something between always going with your gut or always going with the analytics 9n the field

Sgt. Largent · Oct 21, 2021

OrangeGravy":3lpo1x36 said:
Fade":3lpo1x36 said:

If the Seahawks used analytics instead of the random streams of consciousness dribbling out of the brain of Pete Carroll, they wouldn't have sold the farm for Jamal Adams.

Convinced yet?

Click to expand...

No. It makes sense to evaluate players cause you're using their entire careers worth of data points. In game stuff is more convoluted. There should be something between always going with your gut or always going with the analytics 9n the field

It's both. It's always been both.

It's a big part of what sets the great coaches apart from the pretenders. I hate Belichick, but the reason he's the greatest NFL coach of all time is because he removes emotion completely from the equation when making not only organizational decisions, but analytical in game decisions.

Pete's the exact opposite of that. He thinks he has some deeply emotional gutteral connection with the game of football, and that allows him to use "feel" instead of cold hard analytics when making in game and clock management decisions.

Also why he's one of the worst in game and clock managers in the league, and coaches like Belichick are the best.

So convince you? Of what? Analytics are a very important part of any sport, because it gives you the proper context and information to make all these decisions we're talking about.

AgentDib · Oct 21, 2021

I agree with your premise that public analytics in football are rudimentary. There are also other angles that you haven't mentioned yet, such as the reliance on teamwork and lack of meaningful NFL data collection. Momentum is also a hotly debated concept that has been disproved fairly soundly in baseball, somewhat disproved in basketball, but appears to exist and matter in football.

That being said, you've probably heard the term 'Don't let the perfect be the enemy of the good.' If we can derive anything useful from lousy analytics then we should try to do so. Any overconfidence about the state of NFL data science is solely on the part of the fans and broadcasters. Coaches understand the limitations very well but they employ people to crunch what numbers they can anyway because it could give them an advantage.

Also, remember that what you see publicly is free and therefore more likely to be garbage. The guys who show they are actually good at this stuff like Brian Burke get scooped up very quickly. What's left on twitter is largely a bunch of people who maybe took a statistics class or looked at a couple of wikipedia articles but clearly don't really understand what they are talking about or the limitations of the countless assumptions they are making but not even stating.

OrangeGravy":2meqrki9 said:
How do those odds change if you factor in opponent and how they're performing that game on defense?

You are right that these are important, along with a host of other variables which vary on gameday and significantly change the analysis. Every team has guys behind the scenes who are trying to come up with these but at some point the coaches' intuition is important.

Fade · Oct 21, 2021

OrangeGravy":19nyd5b1 said:
Fade":19nyd5b1 said:

If the Seahawks used analytics instead of the random streams of consciousness dribbling out of the brain of Pete Carroll, they wouldn't have sold the farm for Jamal Adams.

Convinced yet?

Click to expand...

No. It makes sense to evaluate players cause you're using their entire careers worth of data points. In game stuff is more convoluted. There should be something between always going with your gut or always going with the analytics 9n the field

You just agreed.

The point of my comment is that the Seahawks are completely ignoring the analytics, and relying on Pete's gut as you worded it.

The numbers would tell you don't do that trade, with sirens flashing. But Pete had a gut feeling, and he knew he could make it work, and fit him in to the scheme. Pete is quoted this season, "We're still trying to figure out how to use him." Yay, us.

If the Jamal Adams trade won't convince you why analytics are needed, nothing will.

LTH · Oct 21, 2021

OrangeGravy":28jxxoxj said:
flv":28jxxoxj said:

You should be looking for every legal way to improve your chances of winning. If analytics help you to do that then embrace them. If they don't then don't. For the most part analytics help you make decisions in very specific scenarios. For example the Eagles were down 14 points late in the 4th quarter to the Buccaneers. Other things being equal the analytics say going for a 2 point conversion rather than kicking the XP gives you about a 3% edge, IE it clearly and demonstrably increases your chances of winning from 50% to about 53%. You still have to score another TD and play defense, but that edge is there...if you choose to take it.

I don't see any specific analytics in regards to running a franchise, other than say, prioritising certain player positions high in the Draft. Obviously no-one should be taking a FB over a QB with a top 3 pick just because the FB is ever so slightly better. If there are such analytics i'd be open to them.

Click to expand...

Going for the 2pt conversion gives you the edge or converting it? You have much better chance of winning if convert for 2 every time you score a TD. What are the odds of converting the 2pt play? How do those odds change if you factor in opponent and how they're performing that game on defense? This is where I struggle with it. All things being equal doesn't exist in one football game to another.

I agree I think you can monitor that teams success every time they go for the 2 point conversion you can't just make a choice off of that one stat because there are to many others factors including the emotion of a game and momentum at that specific moment...

LTH

Fade · Oct 21, 2021

When it comes to analytics on the field. It has to be tapered back.

Y'see analytics don't account for the RG playing injured, and the opponent having Aaron Donald on the other side. It doesn't account for the context of how well or poor the offense is playing as a unit, or likewise for the opponent.

A generic go for it "Yes" or "No" on 4th down, based on where you on the field just because the math told you so is absurd.

That being said if the strength of your team is on offense, you probably should be more aggressive going for it on 4th downs. Rather than punting to put the weaker defensive unit back on the field. And the inverse if the team's strength is the defense. (Slide the scale accordingly.)

There is a lot to unpack when using analytics on the football field, and should definitely be passed through a football mind, before just looking at the math as some sort of bible to go by, ignoring the conditions on the field. It is a tool nothing more.

Analytics are much more effective imo, for GM's evaluating positional value, and accruing value. Particularly draft capitol, and not getting hosed in trades.

Jville · Oct 21, 2021

Fade":316uhyvl said:
If the Seahawks used analytics instead of the random streams of consciousness dribbling out of the brain of Pete Carroll, they wouldn't have sold the farm for Jamal Adams.

Convinced yet?

Opinions can often be shaped by hatred.

And, posts often reveal those controlled by an unhinged hatred for others.

Must be a sad self-imprisonment.

Fade · Oct 21, 2021

Jville":1yys85ns said:
Fade":1yys85ns said:

If the Seahawks used analytics instead of the random streams of consciousness dribbling out of the brain of Pete Carroll, they wouldn't have sold the farm for Jamal Adams.

Convinced yet?

Click to expand...

Opinions can often be shaped by hatred.

And, posts often reveal those controlled by an unhinged hatred for others.

Must be a sad self-imprisonment.

You're projecting. :shock:

Jville · Oct 21, 2021

OrangeGravy":shctemdi said:
How are analytics more than just passively relevant to football? I genuinely want someone to convince me. 17 game seasons with a single elimination tournament? 11 men trying to execute a play while 11 others physically fight them to keep them from executing that play. Physical and talent mismatches all over the place and never in the same place from play to play or game to game. How can you possibly count on probabilities when there are that many uncontrollable variables and with such a small sample? I admit I haven't taken a deep dive into it, but just superficially, I can't see it being that relevant outside of lumping multiple seasons together. I'm sure it helps assist in some areas, but can you really run a franchise on it and expect to be successful?

Im a former college baseball player and it makes perfect sense in baseball even if it has ruined the fan experience. You have 1 single exchange of batter and pitcher that the entire game pivots off of with 162 games played. It's easy to boil down the numbers to what should happen when this specific batter faces this specific pitcher. You can even boil it down a specific pitch from said pitcher to said batter. It works in from just about any angle you want to use it for baseball.

Again this isn't meant to bash analytics. I'm genuinely curious what you stats guys think about it. Do people think it should dominate decision making or should it be 1 of many things to take into account?

Returning to the topic, I think application matters ..... always. I also think fans are mostly unaware of where all this evolved from. With regards to football, I think most are more interested in fantasy, gambling and today's digital board games than the actual field contests that generate the data. When addiction to sites sprout ..... subscribers can typically become adamant believers in assertions depicting athletes and play callers and coaches and owners as ignorant.

For actuaries in the business, understanding analytics is the first step in building useful tools. Most of the rest of us don't invest the review time to assess the usefulness of publications representing themselves as "analytic sites". So for most of their subscribers, comprehension is just a crap shoot. A fuzzy and incomplete simulation that falls short of faithfully describing from what it is collecting data. Producing much confusion and noise. And, that noise is imported into open forums.

Unbeknownst to most subscribers, some sites are tweaking their algorithms from year to year ....... trying to narrow the gaps between their products and the actual field reality. But, sometime I think sports publishers should have just left the old ancient box scores alone.

OrangeGravy · Oct 21, 2021

Jville":1n6q4abv said:
OrangeGravy":1n6q4abv said:

How are analytics more than just passively relevant to football? I genuinely want someone to convince me. 17 game seasons with a single elimination tournament? 11 men trying to execute a play while 11 others physically fight them to keep them from executing that play. Physical and talent mismatches all over the place and never in the same place from play to play or game to game. How can you possibly count on probabilities when there are that many uncontrollable variables and with such a small sample? I admit I haven't taken a deep dive into it, but just superficially, I can't see it being that relevant outside of lumping multiple seasons together. I'm sure it helps assist in some areas, but can you really run a franchise on it and expect to be successful?

Im a former college baseball player and it makes perfect sense in baseball even if it has ruined the fan experience. You have 1 single exchange of batter and pitcher that the entire game pivots off of with 162 games played. It's easy to boil down the numbers to what should happen when this specific batter faces this specific pitcher. You can even boil it down a specific pitch from said pitcher to said batter. It works in from just about any angle you want to use it for baseball.

Again this isn't meant to bash analytics. I'm genuinely curious what you stats guys think about it. Do people think it should dominate decision making or should it be 1 of many things to take into account?

Click to expand...

Returning to the topic, I think application matters ..... always. I also think fans are mostly unaware of where all this evolved from. With regards to football, I think most are more interested in fantasy, gambling and today's digital board games than the actual field contests that generate the data. When addiction to sites sprout ..... subscribers can typically become adamant believers in assertions depicting athletes and play callers and coaches and owners as ignorant.

For actuaries in the business, understanding analytics is the first step in building useful tools. Most of the rest of us don't invest the review time to assess the usefulness of publications representing themselves as "analytic sites". So for most of their subscribers, comprehension is just a crap shoot. A fuzzy and incomplete simulation that falls short of faithfully describing from what it is collecting data. Producing much confusion and noise. And, that noise is imported into open forums.

Unbeknownst to most subscribers, some sites are tweaking their algorithms from year to year ....... trying to narrow the gaps between their products and the actual field reality. But, sometime I think sports publishers should have just left the old ancient box scores alone.

Fans using this stuff to assess players and coaches like they're experts certainly doesn't help and it's exhausting to say the least.

Tokadub · Oct 22, 2021

That's a good question.

I would say analytics are made to be broken.

Like you said football is far too complex to boil it down to any kind of statistical solution.

I do however think it's our best estimate of what will happen in future games because it's looking at the data of the ENTIRE HISTORY OF THE NFL (or whenever they started recording all these things, they probably started tracking different things at different times).

When I say analytics are made to be broken, it just means that if your team has that "IT" factor I would ignore analytics. The Seahawks have won in very mathematically unfavorable positions DOZENS OF TIMES WITH RUSSEL WILSON (and a lot with Marshawn Lynch).

BUT... sadly this season I don't think we have any kind of "IT" factor...

Jamal Adams seems like a total bust waste of even his salary let alone 2 1st round picks.

Our defense seems unable to adjust and use players like Ryan Neal so we don't give up the most yards in the league...

We didn't run the ball much at all in the 1st half vs the Stealers, Pete said he was "pissed"...

LIKE WHAT IS GOING ON??? This season is a total joke!

But there is one thing we 100% do not have yet this season is that "IT" factor where a team cannot be measured by analytics. Right now I think all the analytics are correct we are garbage so far.

RiverDog · Oct 22, 2021

IMO baseball is better suited for analytics than football. Every batter has a batting average, every pitcher has an ERA. What are the odds of a left handed hitter getting a hit off a left handed pitcher? What are the odds of Player X getting a hit in the face of a shift vs. straight away? Those odds can be amazingly predictive of the actual result.

But in football, statistics are more nebulous. What are the odds of converting on 4th and 2 from the 38 vs. the odds of kicking a 56 yard field goal? Does it make a difference if you run or pass? What is your team's odds of success? How do they match up against the defense? What are the consequences of failure and how do you calculate those odds? And what about the wind and other weather conditions? Are you in a dome? In Mile High Stadium?

Baseball has many more samples, ie hundreds of at bats for each individual players and batters faced for pitchers. How many 4th and 2 situations are there in the NFL? There's also fewer variables in baseball. Weather and altitude normally isn't as much of a factor in baseball like it is in football.

That doesn't mean that analytics are useless. It's all relevant information that a head coach should enter into his own personal equation when he has to make a decision. In some situations, such as decisions as to whether or not to go for 2 following a touchdown in the 4th quarter, analytics can be extremely useful.

Bottom line is that analytics is a tool that can be used but should never replace intuitive analysis.

TwistedHusky · Oct 22, 2021

Analytics is the use of data (and computers) to make decisions, usually with large amounts of data.

That is a broad topic. You cannot just say it works or does not work.

It depends on the data you use, how you choose to interpret it, the models you build.

Just as important is the realization that what you do not know is often more important than what you do.

What people often do is identify the data they have, cobble together models based on them, and then try to draw conclusions from that. It can work but it isn't going to be consistently effective.

The best approach is to figure out the factors that are important, the data you have on those factors, and then the data you are missing. You can then figure out the gaps.

AI is great for noticing things, or finding patterns in the data. So it can recognize patterns in the data, and even identify salient factors you might have missed.

But both AI and ML have 3 massive issues:

- it overly weights past vs present or future
- it does not deal well with change (especially sea change)
- It cannot evaluate data in context

So you are still reliant on human expertise to make interpretations based on the resources/tools available.

The other thing to be aware is that making decisions based on analytics changes the very success rates the analytics shows you would have. (For same reason that chasing KPIs can distort the expected result hitting KPIs normally provides)

There is no analytics works or does not work. It depends on how you apply them. But AI/ML and even computer vision are going to be huge tools to help teams build competitive advantages so you probably should expect to see them.

Analytics and Football. Convince me please

Well-known member

Well-known member

Member

Well-known member

Member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Well-known member

Member

Well-known member

Well-known member

Similar threads