Why don't teams hire game theory experts? (1 Viewer)

[32 Counter Pass]

On that Seattle goal line play with the Butler INT it was second and 2. The first
play they did run M Lynch and he lost a yard when he got tackled by Hightower.

That was the 3rd time the Pats stopped M Lynch in short yardage so the pass play
after that wasn't a bad call. B Browner stoning his guy at the line and R Wilson throwing
the ball to where the WR was supposed to be allowed M Butler to make the INT.

Hightower's tackle was the game saver-he got one hand on Lynch's thigh and got him down.
No one talks about this because we didn't get to see the replay. Seattle quickly lined up for
the next play and then it was game over.
 

I think you are incorrect about play sequence.

1:06 4th - SEA 1st & 5, at NE 5
Marshawn Lynch rushed to the left for 4 yard gain

0:20 4th - SEA 2nd & 1, at NE 1
Malcolm Butler intercepted Russell Wilson for 3 yards. New England committed 1 yard penalty

Link

 

[zftcg]

Well, if we are going to educate coaches as you state, then we need to provide a complete information set.  That is, not only do the probabilities of success need to be quantified, but also the consequences of failure.  By going for it on 4th down and failing, there are real issues - you most probably lost points if you are in close and you most probably lost significant field position if out in the middle of the field.  

Turning the ball over on downs with no appreciable change of field position is a turnover every bit as much as a lost fumble.  Turnover ratio has one of the stronger correlations to winning/losing to the best of my knowledge.  Then there is the momentum shift in stopping the other team on 4th down, which I have no clue how one quantifies it, but I can assure you is very real.

Have these things been considered as well by the number crunchers, or only the rate of success vs failure?

Yes, the calculations are rolled up to determine whether a given decision increases or decreases your overall chance of winning. Field position is a significant part of that.

And your assurances aside, momentum is not real.

 

[yabahos]

That was the 3rd time the Pats stopped M Lynch in short yardage so the pass play
after that wasn't a bad call. B Browner stoning his guy at the line and R Wilson throwing
the ball to where the WR was supposed to be allowed M Butler to make the INT.
 

From what I could find in the box score, Lynch had 3 attempts in the game on 3rd & short, was stopped twice and had one TD on 3rd & 2. 

I think the bigger issue was the timeout situation.   Seattle took one at 1:06 & comes out with the run by Lynch.  They had one more left & they didn’t snap the ball until 20 seconds remaining.  Basically running the clock all the way down.

This is the unforgivable part of the play.  Seattle didn’t plan for what happens if they don’t get into the end zone on the Lynch play.   If nothing else, you call two plays coming off the time out and tell everyone to get set and quick snap it with a fade to the back pylon.

At worst it would have been 3rd & 1 with 30-40 seconds & the clocked stopped.   They still would have a timeout and could have run the ball on 3rd down.  

 

[Bronco Billy]

Yes, the calculations are rolled up to determine whether a given decision increases or decreases your overall chance of winning. Field position is a significant part of that.

And your assurances aside, momentum is not real.

No offense intended, but without the data set those momentum articles are pretty meaningless.   I wouldn’t expect teams to score more stopping the other team on downs considering where it is likely they’d get the ball on the field and when it might occur in a game,  as opposed to a punt which is most likely an entirely different data set.

 

[Biabreakable]

Game Theory Cannot Rationalize Seattle’s Super Bowl Loss

A number of finance and economic writers have come out rationalizing Pete Carroll’s controversial play call in the Super Bowl which ended up being intercepted by the Patriots and leading to the loss.  The arguments all use Game Theory to argue that a pass play increased the odds of winning by increasing the randomness of the possible outcomes. In essence, you should try to fool the other coach by considering what his strategy will be. But this isn’t at all what Pete Carroll was trying to do and he knew he didn’t need to outguess the opposition.

I hadn't given it much thought before now. However it seems that game theory agrees with the decision to throw the ball when Seattle did which did not produce the desired result.

So defenders of game theory will say it was the right decision based on probability, they just didn't execute it well enough. Well I can't think of a more air tight perspective where game theory never has any accountability. Coaches do.

You don't let data make decisions for you. You make decisions using the best data that you have available to you. Probability never gets fired, coaches do.

 

[johnadams]

Here's a 12 year old article on Parag Marathe with the 49ers.  They were trying this out long ago and it caused riffs with Walsh and eventually Harbaugh.  Niners are very pretty tight lipped about his role, though I don't think he's ever been a game day guy, but as been involved in drafts and especially contracts. 

 

[Tool]

There was a guy who used to do this back in the 80's; think his name was Clarence Beaks. Not sure what he's up to now. Last I heard he was on his way to Africa.

 

[Zyphros]

Game Theory Cannot Rationalize Seattle’s Super Bowl Loss

I hadn't given it much thought before now. However it seems that game theory agrees with the decision to throw the ball when Seattle did which did not produce the desired result.

So defenders of game theory will say it was the right decision based on probability, they just didn't execute it well enough. Well I can't think of a more air tight perspective where game theory never has any accountability. Coaches do.

You don't let data make decisions for you. You make decisions using the best data that you have available to you. Probability never gets fired, coaches do.

It's an interesting read, which makes total sense.  From the article it appears that Carroll (and game theorists) wanted to give the illusion of a pass play to the defense, and not actually throw it.  Sounds good, so then why throw it?  Would that be on Russell Wilson instead of Carroll then?  Some QB's have higher intellect than others, and some have more freedom at the line to change play calls as well, so is Wilson the one that needs to be smarter with his choices or is it on Carroll?  If it is on Wilson then hiring x number of game theory guys wouldn't do a damn thing unless Wilson decides he needs to be smarter at the line.  This goes for all QB's not just Wilson of course but the example of that play is the easiest to point out.  

 

[Thunderlips]

This is a fun thread to read.  These types of decisions are critical to team success and thereby to a HC’s longevity.  That said, people actually expect a HC to put his job on the line by hiring and then delegating fully the responsibility of these critical decisions to math geeks, many of whom have little to no realistic football theory background?  Seriously?

And to the number crunchers here, how accountable do you intend to be for decisions that you are convinced that statistical analysis support yet still fail on the field due to all sorts of possible variables that you may not be able to quantify and are inherently uncontrollable?  How versed are you in Chaos Theory and how do you intend to account for it in your decision making process?

Clock management is one thing.  I do agree that coaches ought to have some kind of advisor who can put forth advice in managing time - but then is willing to say their piece very quickly and confidently and then shut up and allow the coach to make the decision in what is an incredibly short and stressful time period.  But play calling?  Maybe to educate the coach prior to the season and maybe even early in game weeks - provided there is sufficient background in multiple mathematical fields as well as strong expertise in the game of football itself.  But during the game and then expect the HC to defer?  That’s really just nuts.

Oh, and for all you guys who watched Moneyball and think football works the same way?  This ain’t baseball, which when you drill down can be whittled into a game of one on one between a pitcher and a hitter.  Football is much more immensely complex with multiple interactions between humans as individuals and in subgroups with diverse strengths and weaknesses every play.  The coaches who have tremendous success do everything in their scheming to reduce the number of variables in a play, and ultimately find isolated matchups that they can exploit.  But the actual play on the field and the interactions between all the variables usually means that these exploitable instances that actually occur happen somewhere between 4 and 7 times a game.  The rest of the game comes down to field position, maintaining control of the ball for extended periods by trying to optimize down & distance, reducing turnover opportunities for your offense while enhancing them for your defense, and trying to keep from being exposed to exploitable instances by the other team.  Very very complex stuff.

.

Football absolutely works the same way....at least for (what seems to be) the good teams.  You're seeing it more and more with the devaluation of positions such as LB and RB.  

 

[Tool]

Game Theory Cannot Rationalize Seattle’s Super Bowl Loss

I hadn't given it much thought before now. However it seems that game theory agrees with the decision to throw the ball when Seattle did which did not produce the desired result.

So defenders of game theory will say it was the right decision based on probability, they just didn't execute it well enough. Well I can't think of a more air tight perspective where game theory never has any accountability. Coaches do.

You don't let data make decisions for you. You make decisions using the best data that you have available to you. Probability never gets fired, coaches do.

A lot of questionable points in that article

1)Teams in a 2nd and 1 with 26 seconds left will score 94% of the time (via Pro Football Reference).

So is this stat score or score a TD. I highly doubt it's score a TD in which case stat is totally meaningless. Also, there are other very important variables, most importantly # of timeouts.

2) This had nothing to do with Game Theory or trying to outguess the other coach. Pete Carroll had the New England Patriots dead in the water. He knew that there was virtually no chance that the Pats could stop one of the best offensive lines in the NFL with the best running back in the NFL.

This is just flat out ridiculous

3) So what got into Pete Carroll’s mind? Well, Tom Brady did. Carroll’s team had just scored at the end of the second half in a 30 second drive that led to a touchdown. Carroll wanted to use all of the clock because he knew he’d only be up by 3 points. He didn’t want to leave the ball in Tom Brady’s hands for one more drive leading to a potential game tying field goal.

Wait so why was he trying a pass again? to use the clock? That doesn't make much sense to me.

 

[Thunderlips]

A lot of questionable points in that article

1)Teams in a 2nd and 1 with 26 seconds left will score 94% of the time (via Pro Football Reference).

So is this stat score or score a TD. I highly doubt it's score a TD in which case stat is totally meaningless. Also, there are other very important variables, most importantly # of timeouts.

2) This had nothing to do with Game Theory or trying to outguess the other coach. Pete Carroll had the New England Patriots dead in the water. He knew that there was virtually no chance that the Pats could stop one of the best offensive lines in the NFL with the best running back in the NFL.

This is just flat out ridiculous

3) So what got into Pete Carroll’s mind? Well, Tom Brady did. Carroll’s team had just scored at the end of the second half in a 30 second drive that led to a touchdown. Carroll wanted to use all of the clock because he knew he’d only be up by 3 points. He didn’t want to leave the ball in Tom Brady’s hands for one more drive leading to a potential game tying field goal.

Wait so why was he trying a pass again? to use the clock? That doesn't make much sense to me.

Yeah....he doesn't want to score with Lynch and give Brady time left.....but he has no problem scoring with Wilson to give Brady time left?

 

[Bossman]

BB has a football theory guy. His name is Ernie Adams and is a big part of the coaching staff.  He is under the radar by choice b/c EA doesn’t like the spot light.  

Ernie advises B.B. during games.  He sits in the booth and is the guy that is looking ahead at the possibilities.  

I know there have been a few articles about Ernie’s role on the Pats. 

Came here to post this.

People want to give BB all of the credit. Adams is in his headset during the game and has been for all of BB coaching career with NE as far as I know.

Tells Bill when to challenge a play and when not to, when to go for 2 points, etc, etc. BB doesn't make many mistakes because Adams is usually right.

 

[Biabreakable]

It's an interesting read, which makes total sense.  From the article it appears that Carroll (and game theorists) wanted to give the illusion of a pass play to the defense, and not actually throw it.  Sounds good, so then why throw it?  Would that be on Russell Wilson instead of Carroll then?  Some QB's have higher intellect than others, and some have more freedom at the line to change play calls as well, so is Wilson the one that needs to be smarter with his choices or is it on Carroll?  If it is on Wilson then hiring x number of game theory guys wouldn't do a damn thing unless Wilson decides he needs to be smarter at the line.  This goes for all QB's not just Wilson of course but the example of that play is the easiest to point out.  

Well as you point out it becomes Wilsons call as far as if he sticks with the play or audibles to something else.

Its the OC call to send in the play, which most will send in a play that has two other options based on the defensive look. So 3 plays.

The OC is making the call with guidance and permission from the HC. The HC may in communication with the OC say they want a certain play or play combination.

The game theory is something that coaches will be aware of during practice and game planning during the week for the upcoming game. This affects play calling and perhaps some level of their decision making about 4th downs and so on. If the defense they are facing is better against the run than the pass for example, this might be part of the plan about what to do on 3rd and short and also 4th downs. No defense is the same, so this would change each week.

The coaches may have a chart or something like that they use during game days that tells them the probability of decisions they are making as the game develops, but that doesn't mean they will always follow it. It is a tool like anything else, and only one of a lot of other things they are considering in their decision making process.

I think most of us agree that in this particular situation, the Patriots defense at the time being poor at stopping power runs and the Seahawks being good at converting power runs, that they likely should have tried to run it with Lynch on 2nd and 1 from the goal line instead of passing it.

Their decision to pass the ball likely was influenced by game theory, even though common sense may have suggested that they run it. So there are instances where following GT does not go well and in retrospect that decision seems like the wrong one. 

Carroll takes responsibility for the play after the game, as he should, it is ultimately his decision what they do.

However as an outside observer the coach is always going to get the credit or the blame for decisions they make during a game. If they are applying game theory or not. So the original premise of the post seems off to me, because coaches are using game theory to different degrees in their decision making. In this particular instance they were using GT and after the fact most, including Pete Carrol thought it was a bad decision at the time, informed by GT or not.

But the coach always gets the blame and game theory is never held accountable.

 

[Biabreakable]

A lot of questionable points in that article

1)Teams in a 2nd and 1 with 26 seconds left will score 94% of the time (via Pro Football Reference).

So is this stat score or score a TD. I highly doubt it's score a TD in which case stat is totally meaningless. Also, there are other very important variables, most importantly # of timeouts.

2) This had nothing to do with Game Theory or trying to outguess the other coach. Pete Carroll had the New England Patriots dead in the water. He knew that there was virtually no chance that the Pats could stop one of the best offensive lines in the NFL with the best running back in the NFL.

This is just flat out ridiculous

3) So what got into Pete Carroll’s mind? Well, Tom Brady did. Carroll’s team had just scored at the end of the second half in a 30 second drive that led to a touchdown. Carroll wanted to use all of the clock because he knew he’d only be up by 3 points. He didn’t want to leave the ball in Tom Brady’s hands for one more drive leading to a potential game tying field goal.

Wait so why was he trying a pass again? to use the clock? That doesn't make much sense to me.

1) I am not sure where he gets this information from PFR about scoring 94% of the time. I agree with you point that scoring a field goal would not have been enough, they needed the TD. Scoring a TD is likely a lower probability than 94% from that down and distance.

2) The author links an article in the first sentence of his article, first paragraph. The article published in the economist he is responding to is talking about the prisoners dilemma. This is just one of many 'game theory' articles that were published in response to the specific play being talked about here.

3) that is just speculation on the authors part. Time certainly was a factor in the decision though. They had decided to run the block down and only had one time out left. So if they ran on 2nd down and did not make it, they would need to throw the ball on 3rd down, to either score or stop the clock to try again on 4th down. The time preservation being a key factor it seems in Carrolls decision here. Now you could rewind this a bit as say they shouldnt have run off so much clock prior to the 2nd down, leaving them more flexibility to run on 2nd down, but they didn't. So it makes some logical sense I guess to throw on 2nd down instead of 3rd down to preserve the time out.

Of course the interception makes these time management decisions moot.

The point that I am trying to make however is that game theory was a part of the decision making process, but that no chart can cover every down, distance, time remaining, score and other variable being considered here. Coaches can make an informed decision, but the decision is still ultimately a human one.

The way people are describing game theory is as though it is infallible, while the coaches decision is always at fault. I don't think that is fair or true.

 

[Bossman]

This is one I notice many coaches get wrong often. Sometimes, some plays are worth a challenge, even if you're likely to lose it. There are some game-breaking plays late in the game that are worth risking the challenge. It's often that 1) we're close enough to two minutes that there's unlikely to be a more valuable play later worth challenging, and 2) if the current play stands as-is, it's unlikely that having the extra timeout will matter anyway. If those two conditions are true, then, a challenge is correct even if it's overturned 95%+ of the time. Further, a challenge when you're going to call a timeout anyway is worth it late in the game. You're stopping the clock either way, might as well roll the dice and see if you can overturn the last play. If not, hey, you get a longer timeout out of the deal anyway.

And, speaking of timeouts, there's nothing more pointless to me than some team burning one super early to save a delay of game or too many men in the huddle penalty. Who cares about 5 more yards in the first or third quarter? It's unlikely to matter as much as having that timeout in the endgame. Take the delay of game and move on. It's OK. You'll get another play off. 5 yards is not worth a timeout, pretty much ever. The conditions under which 5 yards matters are few--late in the game, on the edge of field goal range, score-dependent, clock-dependent, number of other timeouts, etc.  Kills me to see it when it happens.

I forget what game I was watching a couple weeks ago ... but the teams come out to start the quarter (2nd qtr?) and the offense takes a timeout.

WHAT?? .... You just had a 4 min break. You couldn't figure out what you were going to do? Amazing how often teams waste timeouts and how it burns them at the end of the half / game.

What frosted me the most was that we just got back to game action after a long commercial break .... then back to more commercials without even one play run. 

 

[jwb]

One problem with adopting any probability strategy is football is much more a "every game is critical" sport than others. With only 16 games, and immense pressure to win every one, it's much harder to accept long-term probability than an in-the-moment gut call. To use baseball as a contrast, it has failure built into the sport - you accept that roughly seven out of ten times, even a great hitter will fail. And all managers fully accept "you can't win em all". Try telling a football coach that. 

 

[johnadams]

I honestly can't imagine any situation where the first play of the second quarter is worth burning a timeout rather than taking 5 yards. There are literally zero ways that is the correct play.

One of my biggest gripes is watching teams do this on their side of the field early in halves.  In general, I would much rather the penalty over the timeout for a fairly simple reason.  You can get the yards back but can't get the timeout back. 

 

[zftcg]

Semantic nitpick: A lot of what's being discussed in this thread is simply "decision making". Game theory specifically applies to situations where your decision is shaped by your expectations as to what your opponent (or other stakeholders) will do.

Here's a good example (which was discussed a bunch in the "Post here when coaches do something stupid" thread at the time): Falcons-Saints TNF game, Saints down 3 late in the 4th, throw incomplete on 3rd and 1 from the 24. But there's an offensive holding penalty on the play. Quinn is faced with a choice: Accept the penalty and put the Saints in a 3rd and 11 from the 34 or decline it and make it 4th down.

Now, I have no idea what his thought process was in making his decision, but it should have been informed by what he thought Payton would do. I can't imagine that, if he had known Payton would go for it on 4th and 1, he would have declined the penatly. The thing is, he clearly should have expected Payton to go for it, because that was pretty obviously the right call for NO. (The alternative would have been to kick a long-ish FG that would only tie the game and give it back to Atlanta with plenty of time to mount a game-winning drive.) Based on that expectation, it made much more sense to accept the penalty.

So whether Quinn either miscalculated Payton's expected move or didn't consider it at all, he really could have benefitted from a smart game theorist at that moment. (Of course, after converting the 4th on a sneak, Brees bailed the Falcons out by throwing a dumb INT in the end zone.)

 

[TwinTurbo]

No coach is going to give up decision making power to some quant guy. But where it starts to get interesting is an AI assistant coach with a backend built by a team of quant guys. I think some coaches would secretly sign on for that. Hey Alexa, should I go for it here? It sounds like the Eagles sort of do this today and have some analytics tools that the coaches use in certain situations. 

 

[IHEARTFF]

Considering that most people, especially those who have had success in the past, are terribly irrational decision makers, all teams should be using advanced analytics. Most pro sports teams are using some form of data analysis to guide player evaluations, but in game decisions are less utilized. 

 

[GroveDiesel]

1) I am not sure where he gets this information from PFR about scoring 94% of the time. I agree with you point that scoring a field goal would not have been enough, they needed the TD. Scoring a TD is likely a lower probability than 94% from that down and distance.

2) The author links an article in the first sentence of his article, first paragraph. The article published in the economist he is responding to is talking about the prisoners dilemma. This is just one of many 'game theory' articles that were published in response to the specific play being talked about here.

3) that is just speculation on the authors part. Time certainly was a factor in the decision though. They had decided to run the block down and only had one time out left. So if they ran on 2nd down and did not make it, they would need to throw the ball on 3rd down, to either score or stop the clock to try again on 4th down. The time preservation being a key factor it seems in Carrolls decision here. Now you could rewind this a bit as say they shouldnt have run off so much clock prior to the 2nd down, leaving them more flexibility to run on 2nd down, but they didn't. So it makes some logical sense I guess to throw on 2nd down instead of 3rd down to preserve the time out.

Of course the interception makes these time management decisions moot.

The point that I am trying to make however is that game theory was a part of the decision making process, but that no chart can cover every down, distance, time remaining, score and other variable being considered here. Coaches can make an informed decision, but the decision is still ultimately a human one.





The way people are describing game theory is as though it is infallible, while the coaches decision is always at fault. I don't think that is fair or true.

This nails it. I do think that coaches do make some objectively bad decisions at times that are indefensible. But I also think that some stats guys grossly oversimplify things too. For instance, all of our current stats are based on what teams currently do in those situations. If teams "follow the stats" so to speak, that act itself may change expectations of the other team in those situations and how they behave and thus change the true potential outcome. 

For instance, if the statistics say that you increase your chances of winning by going for it on 4th and 1 at the 50 yard line and that you have a 70% chance to convert it if you run the ball versus a 30% chance to convert it if you pass, if everyone in the league is going to follow the stats, then everyone is going to try to run the ball in that situation every single time.

Only, by doing so, defenses now don't have to respect the pass and can stack the box against the run, which may drop the conversion rate enough that the very act of going for it no longer increases your chance of winning. Only, how much data do you need before you can know that to be true and react accordingly? And what if your team's offensive line has had multiple injuries or just isn't very good at blocking in short yardage. Or maybe the defense has 2 of the best run stoppers in the league out on the field. 

There is a ton of data and stats out there, but that doesn't make it meaningful or applicable. All 4th and 1s are not the same for a WHOLE lot of reasons: point in game, location on field, current score, personnel on field for both teams, weather, timeouts left, challenges left, etc. How many other 4th and 1 situations have really occurred that match up closely enough to all of those variables to truly be applicable? I'd wager not enough to be statistically relevant 

 

[Biabreakable]

I have a background in intelligence. So I am very familiar with what people are talking about in regards to game theory, remote sensing and data driven decision making.

If you really want to be serious about this, every team is different and every team they are playing against is different. To claim a one size fits all process to decision making just isn't how things really work, nor should it be.. 

 

[Leroy Hoard]

Came here to post this.

People want to give BB all of the credit. Adams is in his headset during the game and has been for all of BB coaching career with NE as far as I know.

Tells Bill when to challenge a play and when not to, when to go for 2 points, etc, etc. BB doesn't make many mistakes because Adams is usually right.

I give BB all of the credit for having this guy on his staff. Some of the biggest geniuses are those who realize that they themselves don't know everything.

 

[Bossman]

I give BB all of the credit for having this guy on his staff. Some of the biggest geniuses are those who realize that they themselves don't know everything.

From: Boston Globe 2015;

https://www.bostonglobe.com/sports/2015/01/29/mysterious-ernie-adams-patriots-man-behind-curtain/IrNCfgrysUphGpkcIjEaBL/story.html

When Adams worked with Belichick in Cleveland, Browns owner Art Modell famously said, “I’ll pay anyone here $10,000 if they can tell me what Ernie Adams does. I know he does something, and I know he works for me, and I know I pay him, but I’d love to know what it is.’’

No one on the Cleveland staff came forward with information for the reward.

............

Adams works up to 100 hours per week during the season. He studies film, devours statistics, reports on trends, and develops strategies on 2-point conversions, fourth-down attempts, and timeout preservations. He runs the vaunted Patriot “value chart,’’ helping Belichick on personnel decisions regarding free agents, trades, and the draft. He appears to be the voice inside Belichick’s head for 60 minutes every Sunday, but no one will say for sure.

........................

When the game starts, Adams is upstairs in the booth with the Patriots’ eye-in-the-sky team. He wears a headset. He has Belichick’s ear. When the Patriots have allowed NFL Films to record some of Belichick’s in-game dialogue, you hear a lot of, “What have we got, Ernie?’’

Down on the field, the dedicated Ernie Hotline is available for Patriots assistant coaches. Or in case there’s a scramble with Belichick’s headset. The Ernie Hotline is the backup. In case of emergency. Belichick’s lifeline.

Adams is the one who tells Belichick whether the Patriots should toss the red challenge flag after a questionable call. At halftime, Adams hustles downstairs with several other coaches and tells Belichick what needs to be done in the second half. The Patriots annually lead the league in effective second-half adjustments.

.................

Some teams DO have a game theory expert. 

I guess a better title for this topic should be "why don't ALL teams have a game theory expert?"

 

[Bossman]

From: Boston Globe 2015;

https://www.bostonglobe.com/sports/2015/01/29/mysterious-ernie-adams-patriots-man-behind-curtain/IrNCfgrysUphGpkcIjEaBL/story.html

When Adams worked with Belichick in Cleveland, Browns owner Art Modell famously said, “I’ll pay anyone here $10,000 if they can tell me what Ernie Adams does. I know he does something, and I know he works for me, and I know I pay him, but I’d love to know what it is.’’

No one on the Cleveland staff came forward with information for the reward.

............

Adams works up to 100 hours per week during the season. He studies film, devours statistics, reports on trends, and develops strategies on 2-point conversions, fourth-down attempts, and timeout preservations. He runs the vaunted Patriot “value chart,’’ helping Belichick on personnel decisions regarding free agents, trades, and the draft. He appears to be the voice inside Belichick’s head for 60 minutes every Sunday, but no one will say for sure.

........................

When the game starts, Adams is upstairs in the booth with the Patriots’ eye-in-the-sky team. He wears a headset. He has Belichick’s ear. When the Patriots have allowed NFL Films to record some of Belichick’s in-game dialogue, you hear a lot of, “What have we got, Ernie?’’

Down on the field, the dedicated Ernie Hotline is available for Patriots assistant coaches. Or in case there’s a scramble with Belichick’s headset. The Ernie Hotline is the backup. In case of emergency. Belichick’s lifeline.

Adams is the one who tells Belichick whether the Patriots should toss the red challenge flag after a questionable call. At halftime, Adams hustles downstairs with several other coaches and tells Belichick what needs to be done in the second half. The Patriots annually lead the league in effective second-half adjustments.

.................

Some teams DO have a game theory expert. 

I guess a better title for this topic should be "why don't ALL teams have a game theory expert?"

All you folks posting in here that want to poo-poo having this guy...

Of all of the games that have been screwed up by bad coaching decisions, imagine how many may have finished with a different result had the coach been advised to make correct in-game decisions.

What if the losing team had one or two more timeouts after not wasting them on a bad challenge ... or because they had to discuss whether or not to go for it on 4th down earlier in the game? 

What if the losing team had an extra timeout because the coach was advised to take the 5 yard penalty instead of burning a timeout on that fake "draw them offsides" 4th and 1 play. 

One or two timeouts could mean the difference in being able to get the ball back at the end of a game ... or being able to use the middle of the field on the final drive instead of having to find the sideline every play.

What if the losing team got the ball back with an extra 1:30 on the clock because of better clock management at the end of the half or end of the game?

How many times have we seen teems screw this up?

What if the coach of the losing team was advised on when to on-side kick and kick off deep when they numbers / stats said when any why to do so. (looking at you Steelers / Tomlin vs JAX last week)

Imagine how many coaches might not have lost their job if they're in game decision making was near perfect resulting in more wins.

...In a copy-cat league, why don't all teams have a guy? ... and if the Steelers do, he should be fired.

 

[yabahos]

All you folks posting in here that want to poo-poo having this guy...

...In a copy-cat league, why don't all teams have a guy? ... and if the Steelers do, he should be fired.

I don’t think there are a lot of people saying teams don’t need this type of guy.  I think people are saying it’s a lot harder to work game theory just because of the sheer amount of variables.  

The copycat aspect seems easier.  First you need a coach to accept that it works. Belichick was the first coach I can remember that would defer on a regular basis.  This was b/c Adams showed him in the long run it meant more possessions and the chance to get back to back possessions (end of the second/start of the third).   

I would bet that most coaches either don’t  have a background to understand the stats, don’t want to risk their job on the stats, or don’t put much weight behind the stats.   Moneyball in baseball took a while to catch on and now every team uses it.  

Belichick is in a unique situation where he is secure enough that if something doesn’t go his way he gets the benefit of the doubt with the fan base, ownership & the players.

Belichick surrounds himself with players and coaches that compliment his style.  He knows what he isn’t good at and finds someone he trusts to help out.

 

[IHEARTFF]

Just the fact that NE has a guy doing this, should be reason enough for all teams to copy. That is, if they are interested in winning, which, they don’t all appear to be. 

 

[zftcg]

ESPN.com had a great article this week about the Eagles' use of analytics. Turns out they're doing a lot of the stuff we've been talking about in this thread:

The situations in which the Eagles decide to strike might seem random, but in fact they are quite calculated. And they're often decided before the start of the game -- or even before the start of the season. The approach is driven by an analytics team so involved in the operation that two members of the department -- including coaching assistant/linebackers coach Ryan Paganetti, a Dartmouth grad with a degree in economics -- communicate with Pederson in-game.

The pair weighs in after just about every touchdown on whether to go for one or two; during the final two minutes of each half to discuss timeouts; and when the team gets into what is considered fourth-down territory, usually around midfield and beyond. Sometimes Pederson initiates the dialogue; other times, the men upstairs do.

"After [Pederson has] made the third-down call the phones can be silent for a few seconds, and one of the guys might chime in and say, 'Hey Coach, if this ends up short fourth-and-2' -- I'm using fake terminology -- 'it's green, go for it. The charts say go for it,'" explained offensive coordinator Frank Reich. "Or, 'Hey, if it's anything less than fourth-and-3, we're good. Other than that, it's your call, Coach.' Or, 'Anything more than fourth-and-10, no.'

"The analogy I think of is kind of like a stoplight. There's green, there's yellow and there's red, and then there's shades of green, there's shades of yellow and then there's shades of red. So some of them are, 'Hey, it's green. Yellow, proceed with caution' -- and that's how it operates."

 

[zftcg]

I have a background in intelligence. So I am very familiar with what people are talking about in regards to game theory, remote sensing and data driven decision making.

If you really want to be serious about this, every team is different and every team they are playing against is different. To claim a one size fits all process to decision making just isn't how things really work, nor should it be.. 

I agree, and anyone who understands quantitative decision making knows that you should talk in probabilities rather than absolutes. (Contrary to what many of his critics have said, Nate Silver never predicted that Hillary would beat Trump. He said she had a 65-percent chance of winning. The fact that she lost doesn't mean he was wrong; it could just mean we ended up in one of the 35% scenarios.)

That being said, one thing I've noticed is that too frequently, those who argue, "I know what the numbers say, but in this situation it didn't make sense" are using it to justify their conservative, by-the-book decision. (Not saying that's what you're doing here; I'm thinking more of coaches.) I can recall a couple counter-examples -- Del Rio going for two against the Saints in last year's opener, Payton going for it on 4th down against the Panthers in this year's wildcard -- but in general it only goes in one direction. And when you look at it in the aggregate, those "extenuating circumstances" become a lot less convincing.

Take my favorite hobby horse -- the decision on whether to go for two if you're down 14 late and score a TD. The numbers couldn't be clearer that you should. Now, have there been some specific scenarios where it might not have made sense? I'm sure you could find a couple. But when you look at all the times teams have been in that situation, and realize that, essentially, none of them have ever gone for two, it suggests something else is going on.

 

[Biabreakable]

I agree, and anyone who understands quantitative decision making knows that you should talk in probabilities rather than absolutes. (Contrary to what many of his critics have said, Nate Silver never predicted that Hillary would beat Trump. He said she had a 65-percent chance of winning. The fact that she lost doesn't mean he was wrong; it could just mean we ended up in one of the 35% scenarios.)

That being said, one thing I've noticed is that too frequently, those who argue, "I know what the numbers say, but in this situation it didn't make sense" are using it to justify their conservative, by-the-book decision. (Not saying that's what you're doing here; I'm thinking more of coaches.) I can recall a couple counter-examples -- Del Rio going for two against the Saints in last year's opener, Payton going for it on 4th down against the Panthers in this year's wildcard -- but in general it only goes in one direction. And when you look at it in the aggregate, those "extenuating circumstances" become a lot less convincing.

Take my favorite hobby horse -- the decision on whether to go for two if you're down 14 late and score a TD. The numbers couldn't be clearer that you should. Now, have there been some specific scenarios where it might not have made sense? I'm sure you could find a couple. But when you look at all the times teams have been in that situation, and realize that, essentially, none of them have ever gone for two, it suggests something else is going on.

Yes probability as represented as a percentage does always include the chance that the result will fall on the lower frequency outcome. Any given Sunday as they say, or to be more specific any given play can lead to a wide spectrum of results. Even with the same team, same personnel, running the same play, against the same defense, in the same game will lead to different results, for as variety of reasons too many to really list. Time of the game, down and distance would be variables that affect the defensive play call. And even if it is the same play call vs the same defense, the results will be different just because of execution.

The article you cite makes sense. 60% > 50%

I do think the math is over simplified here but sure I can see the argument for teams going for 2 point conversions more often than they do. It is still somewhat dependent on the situation. I think and how good the opposing defense might be against specific 2 point plays the team wants to run however.

The article does contradict itself somewhat as well. It assumes that the team will be able to score two TD with the time remaining and that the opposing team does not answer with any points of their own. The points scored in overtime contradicts this, and the score by the opposing team could have occurred during regulation, but is just precluded from the possibility by the assumption. Realistically the score by opposing team in overtime is the same thing, just happening at another time of the game. So this makes the comparison unfair in that sense and in my opinion not a rigorous investigation into the different probability factors at play. The data becomes cherry picked by the argument. 

The percentages themselves are from 32 different teams, different years of those teams all against different opponents. So it is hard for me to take any win percentage too seriously, when there are so many different variables that go into the calculation of that number.

This is a bit different than the go for it on 4th down arguments and not as clear cut.

Maybe there are some teams that don't have someone in house doing this analysis for them, but I doubt it is many teams. Some teams should perhaps make different decisions than they do and they could utilize the information provided by their analysts more than they do. I still think that teams are very aware of these things and they do consider the data when making their decisions.

One thing that hasn't been pointed out (because the articles are almost all pro data perspectives and propositions) is that you can make a decision to go for two points and then fail to execute, which can lower the teams win probability as a result. It makes more sense going for 2 points when the team is behind in the score than it might if the team is currently ahead in the score. It still comes down to execution, and I do not think these generalized percentages are good examples of the differences of each team at the given time, or the opponent they are up against.

 

[Biabreakable]

Let use the recent TD by Diggs.

If this play happens at any other part of the game than it did, still a great play but much less drama in the game than having it happen when it does. However because of when it happens, you likely have a different defensive play call and it changes the situational awareness of the players at that point in time. To me it kind of looked like he was trying to tackle Diggs from an angle that would keep him in bounds. Diggs is able to use this leverage against him, and the high throw causing Diggs to be elevated at the time the safety is trying to tackle him.

I am guessing that the coverage call and execution by the safety perhaps different because of the situation. So they would be in a different defensive play call and coverage, and the safety would be more focused on stopping the TD than trying to make a big play to end the game.  If there was more time, the safety perhaps takes a different angle or isn't as aggressive with the tackle attempt. If the ball were on target and Diggs doesn't have to leap for it. Perhaps he makes the tackle. He basically misses because of that leap being at the right time and because the ball is thrown too high.

You can't predict these things.

 

[Zyphros]

Let use the recent TD by Diggs.

If this play happens at any other part of the game than it did, still a great play but much less drama in the game than having it happen when it does. However because of when it happens, you likely have a different defensive play call and it changes the situational awareness of the players at that point in time. To me it kind of looked like he was trying to tackle Diggs from an angle that would keep him in bounds. Diggs is able to use this leverage against him, and the high throw causing Diggs to be elevated at the time the safety is trying to tackle him.

I am guessing that the coverage call and execution by the safety perhaps different because of the situation. So they would be in a different defensive play call and coverage, and the safety would be more focused on stopping the TD than trying to make a big play to end the game.  If there was more time, the safety perhaps takes a different angle or isn't as aggressive with the tackle attempt. If the ball were on target and Diggs doesn't have to leap for it. Perhaps he makes the tackle. He basically misses because of that leap being at the right time and because the ball is thrown too high.

You can't predict these things.

That's less about the decision of the defense to play in that situation than it was pure luck.  The situation called was, I'm assuming, the right play call from both teams.  Keenum makes a chuck to a WR on the boundary so he can possibly get out of bounds and stop the clock, and the Saints in a prevent type of defense trying to force them to stay in bounds.  Sure the play calls would be different with 8:00 left in the 3rd quarter, hell maybe Keenum even throws the ball away in that situation.  Either way the decision of each play seemed to be the right one, by both sides, the execution was the error by the defense.  That part doesn't play into the decision because coaches are going to assume player X does their job right?  You don't plan for failure, it's an easy way to lose.  Situations dictate those calls, but that's a bad example.  

Perhaps it's all the variables in the situations that are causing the coaches to misplace their thoughts in the process of making the call.  Like the Seahawks running the ball example I gave a while back, there was an article posted I think by you Bia, that showed the call was the right one (the illusion of pass, but in the end run it), maybe Carroll didn't express that concern to Russell the right way and Russell took what he thought was the best route.  We don't know and can only speculate.  At least that seems plausible.  So it could be the calls (at least in some cases, I highly doubt in all of them) are the right ones and then the decision ends up branching out into decisions by 2 or more people which cause the confusion or the "incorrect" play being executed (or botched).  

 

[tjnc09]

And, to his credit, Belicheck seems to be one of the only coaches who knows a lick about probability and game theory.

Patriots organization is well known for utilizing analytics.  In contrast, the Browns hire "football" people and fire executives who utilize analytics.

 

[pecorino]

Patriots organization is well known for utilizing analytics.  In contrast, the Browns hire "football" people and fire executives who utilize analytics.

If only there was a way to measure or quantify which organization is doing it right...,

 

[zftcg]

Patriots organization is well known for utilizing analytics.  In contrast, the Browns hire "football" people and fire executives who utilize analytics.

Again, it's not who you hire, it's whether the HC listens to those people. If the nerds are holed up in their cubicles and the coach resents them for being foisted on him, that's just going to make things worse.

Remember, the key to Moneyball was that Beane was a "baseball guy" who also had buy-in from the owner and was empowered to tell Howe what to do.

 

[Steeler]

ESPN.com had a great article this week about the Eagles' use of analytics. Turns out they're doing a lot of the stuff we've been talking about in this thread:

Sounds terrible!  Why would a coach who has been in football all hos life want this additional information??

 

[zftcg]

Has any reporter ever done a deep-dive story on Andy Reid? It's kind of fascinating. Here's a guy who is by all accounts one of the best coaches in the game. He hires great coordinators and has an impressive coaching tree. His players love him, and he has consistently developed solid QBs, including guys who were terrible once they went elsewhere. He has an absolutely brilliant offensive mind. He wins consistently wherever he goes (.604 regular season winning percentage, made playoffs in 13 of 19 seasons)

But he's a terrible in-game manager, and it has cost him, particularly in the playoffs, where he's 10-13 (1-4 with KC). He consistently botches challenges, time-outs, and pacing. He'll have stud RBs like Charles and Hunt, then fail to give them the ball in games where KC is leading.

And the thing is, everyone knows all of this. Ask any Philly or KC fan about Reid, and it will be the first thing they tell you. And yet I've never heard a convincing explanation of how this happens. Is Reid aware of this image? Does he realize it's what's standing between him and coaching immortality? What has he done to try to correct it?

Feels like the final season of The Wire, where understaffed newspapers are chasing fake stories about serial killers while missing the stories that actually affect people on the street (Marlow taking over the hookup, Omar's death).

 

[zftcg]

Sounds terrible!  Why would a coach who has been in football all hos life want this additional information??

https://www.youtube.com/watch?v=7a_sx3ozoXI

 

[tjnc09]

ESPN.com had a great article this week about the Eagles' use of analytics. Turns out they're doing a lot of the stuff we've been talking about in this thread:

Great article.  Love this type of stuff.  

 

[tjnc09]

Again, it's not who you hire, it's whether the HC listens to those people. If the nerds are holed up in their cubicles and the coach resents them for being foisted on him, that's just going to make things worse.

Remember, the key to Moneyball was that Beane was a "baseball guy" who also had buy-in from the owner and was empowered to tell Howe what to do.

http://blog.masslive.com/patriots/2018/01/sean_harrington_patriots_engin_1.html

sounds like NE surrounds their coaching staff with intelligent people with football experience as well.

 

[TwinTurbo]

Has any reporter ever done a deep-dive story on Andy Reid? It's kind of fascinating. Here's a guy who is by all accounts one of the best coaches in the game. He hires great coordinators and has an impressive coaching tree. His players love him, and he has consistently developed solid QBs, including guys who were terrible once they went elsewhere. He has an absolutely brilliant offensive mind. He wins consistently wherever he goes (.604 regular season winning percentage, made playoffs in 13 of 19 seasons)

But he's a terrible in-game manager, and it has cost him, particularly in the playoffs, where he's 10-13 (1-4 with KC). He consistently botches challenges, time-outs, and pacing. He'll have stud RBs like Charles and Hunt, then fail to give them the ball in games where KC is leading.

And the thing is, everyone knows all of this. Ask any Philly or KC fan about Reid, and it will be the first thing they tell you. And yet I've never heard a convincing explanation of how this happens. Is Reid aware of this image? Does he realize it's what's standing between him and coaching immortality? What has he done to try to correct it?

Feels like the final season of The Wire, where understaffed newspapers are chasing fake stories about serial killers while missing the stories that actually affect people on the street (Marlow taking over the hookup, Omar's death).

I think Reid is very similar to Marty Schottenheimer. Another great regular season head coach with an impressive coaching tree, but who consistently botched it in the playoffs. For both guys, it really boils down to playcalling. They both made extremely questionable playcalling decisions in the games they lost. For Schottenheimer, it was always him being far too conservative. For Reid, it was always him abandoning the run about 3 quarters to early. They both made the same exact mistakes over and over again and never changed.

I will say Reid got a bit unlucky this time around with the head shot to Kelce and that almost happened again yesterday to the Pats with Gronk. But it just highlights the contrast in coaching styles, as BB didn't fold up and made great second half adjustments like he always does, whereas Reid appeared to make no adjustments at all. I believe BB probably has well crafted playcalling plans already in place for if Gronk gets hurt, as there is a reasonable chance of that happening based on his injury history. I don't think Reid even considers those types of possibilities and only has one plan.

 

[zftcg]

I think Reid is very similar to Marty Schottenheimer. Another great regular season head coach with an impressive coaching tree, but who consistently botched it in the playoffs. For both guys, it really boils down to playcalling. They both made extremely questionable playcalling decisions in the games they lost. For Schottenheimer, it was always him being far too conservative. For Reid, it was always him abandoning the run about 3 quarters to early. They both made the same exact mistakes over and over again and never changed.

I will say Reid got a bit unlucky this time around with the head shot to Kelce and that almost happened again yesterday to the Pats with Gronk. But it just highlights the contrast in coaching styles, as BB didn't fold up and made great second half adjustments like he always does, whereas Reid appeared to make no adjustments at all. I believe BB probably has well crafted playcalling plans already in place for if Gronk gets hurt, as there is a reasonable chance of that happening based on his injury history. I don't think Reid even considers those types of possibilities and only has one plan.

But my question is, why is this discussion seemingly confined to message boards? Every football fan knows this about Reid, yet it never gets discussed. I'd love to see a post-game press conference where someone asks Reid, "You always seem to blow these types of games. What adjustments are you making so that this doesn't keep happening?"

 

[Riversco]

I think Reid is very similar to Marty Schottenheimer. Another great regular season head coach with an impressive coaching tree, but who consistently botched it in the playoffs.

Marty Schottenheimer only had a couple seasons where he had a franchise QB.
85-88: Bernie Kosar
89-91: Steve DeBerg
92: David Kreig
93: Joe Montana, went 11-5, won 2 playoff games, lost AFC championship game to Jim Kelly and the Bills.
94: Joe Montana was done
95-96: Steve Bono
97: Elvis Grbac
98: Rich Gannnon
01: Tony Banks
02-03: young Drew Brees who wasn't very good
04: Drew Brees emerges as a stud.  Lose first playoff game at home to the Jets.  In that game, Chad Pennington completes a miracle hail mary TD pass to Santana Moss in the end zone when he was double covered, Nate Kaeding misses a short FG and they lose.
05: Miss the playoffs in a bizarre season.  They blow out the Patriots in Foxboro, they blow out the Colts, but have all of these close losses by 4,3,3, and 2 points and are actually eliminated going into the last week of the season even though they are 9-6.  Then Drew Brees has his throwing shoulder destroyed and leaves San Diego.
06: rookie Phillip Rivers who was ok but not a franchise QB yet. They go 14-2 and lose to Tom Brady and the Patriots.

Marty had Montana in 93 & Brees in 04 and 05. Outside of that it was a struggle and you should give Marty a lot of credit for winning as much as he did with what he had.  Andy Reid never had a franchise QB in his life.  He had Donovan McNabb who was a good QB but not a guy anyone feared.  Now he's got Alex Smith.

 

[Short Corner]

Wow.  Okay.  Have it your own way and spin it up any way you prefer to reinforce yourself.  Have a great day.

 

[Short Corner]

Well, if we are going to educate coaches as you state, then we need to provide a complete information set.  That is, not only do the probabilities of success need to be quantified, but also the consequences of failure.  By going for it on 4th down and failing, there are real issues - you most probably lost points if you are in close and you most probably lost significant field position if out in the middle of the field.  

Turning the ball over on downs with no appreciable change of field position is a turnover every bit as much as a lost fumble.  Turnover ratio has one of the stronger correlations to winning/losing to the best of my knowledge.  Then there is the momentum shift in stopping the other team on 4th down, which I have no clue how one quantifies it, but I can assure you is very real.

Have these things been considered as well by the number crunchers, or only the rate of success vs failure?

This is already accounted for in the EV of the decision.  

 

[Short Corner]

I honestly can't imagine any situation where the first play of the second quarter is worth burning a timeout rather than taking 5 yards. There are literally zero ways that is the correct play.

I would guess that 90% of the population has no idea what this word means.

 

[zftcg]

I would guess that 90% of the population has no idea what this word means.

I disagree. Words don't have fixed meanings. They evolve to reflect common usage. "Literally" is widely understood to be used as a way to emphasize something.

Literally everyone who read @Walking Boot's post knew what he meant. 

 

[Bronco Billy]

This is already accounted for in the EV of the decision.  

I disagree based upon the numbers I saw.  IMO there has to be some segregation of the nature of the event - ie “going for it” on 4th down when it’s 4th and 12 with 35 seconds left in the game and the team is down by two scores, or “going for it” by kneeling on 4th down near the end a half or a game.  The odds of the opposing team scoring on the successive drive if the other team goes out on downs are nonexisent, yet I’m guessing those types of plays make up a significant portion of 4th down non-punts.

Perhaps you’ve seen where the data has been broken out to eliminate those types of plays, or have some knowledge that they are not part of the data set.  I didn’t see it.

 

[zftcg]

I disagree based upon the numbers I saw.  IMO there has to be some segregation of the nature of the event - ie “going for it” on 4th down when it’s 4th and 12 with 35 seconds left in the game and the team is down by two scores, or “going for it” by kneeling on 4th down near the end a half or a game.  The odds of the opposing team scoring on the successive drive if the other team goes out on downs are nonexisent, yet I’m guessing those types of plays make up a significant portion of 4th down non-punts.

Perhaps you’ve seen where the data has been broken out to eliminate those types of plays, or have some knowledge that they are not part of the data set.  I didn’t see it.

I think you misunderstand how EV calculations work. It's not that X% of teams convert on 4th down, therefore they should go for it. It's based on what your win percentage is in a given game situation (4th and 1, 7 minutes left in the 4th quarter, down 3 on opponent's 45, etc.) and how it changes. If you go for it and convert so it's now 1st and 10 on the 44, your win probability will tick up slightly. If you go for it and get stuffed, it will go down. If you punt, it could go up or down depending on the result of the play. Based on the potential changes in win probability, then, the numbers will recommend you should go for it if your chances of converting are higher than a certain percentage.

As for the scenarios you describe, if there are 35 seconds left and you're down by two scores, your win probability is effectively zero, so whatever you do won't have an impact. Also, why would any football team ever kneel on 4th down? The point of kneeling is to run the clock, but it automatically stops on a change of possession.

 

[Short Corner]

I disagree based upon the numbers I saw.  IMO there has to be some segregation of the nature of the event - ie “going for it” on 4th down when it’s 4th and 12 with 35 seconds left in the game and the team is down by two scores, or “going for it” by kneeling on 4th down near the end a half or a game.  The odds of the opposing team scoring on the successive drive if the other team goes out on downs are nonexisent, yet I’m guessing those types of plays make up a significant portion of 4th down non-punts.

Perhaps you’ve seen where the data has been broken out to eliminate those types of plays, or have some knowledge that they are not part of the data set.  I didn’t see it.

If you aren't factoring in score, field position and clock (among other things) in your EV projection you are doing it wrong.