Possible to prove mathematically that the football spot was bad?

[boneh3ad]

I always figured RFID sensors in the ball (tips, around the perimeter traced by the semiminor axis, etc.) plus sensors embedded in the field to track them would be able to do this pretty accurately. As mentioned before, syncing the position to the game clock would mean that instant replay would allow the precise location of the ball to be known at the point of any frame of video.

Of course, placing the sensors could be tricky, but it would be a fun thing to try.

Obligatory edit to support my own almae matres:
Go Illini!
Gig 'Em!
(But mostly Go Illini!)

 

[Chestermiller]

A tracker inside the ball won't help work out where the carrier's knee touched the ground.

And please change the title of the thread - American Football has no axioms and so it is not possible to prove anything about it mathematically.

Thanks for the unsolicited advice on terminology. Here's an axiom for you: it's a bad idea to try to make a PF moderator appear ignorant.

 

[boneh3ad]

And please change the title of the thread - American Football has no axioms and so it is not possible to prove anything about it mathematically.

I bet you're popular at parties.

 

[pbuk]

I bet you're popular at parties.

I wouldn't make that comment at a party. Or in the "General Discussion" forum here - but this is a Physics forum where technical questions are asked and technical answers given.

 

[boneh3ad]

I wouldn't make that comment at a party. Or in the "General Discussion" forum here - but this is a Physics forum where technical questions are asked and technical answers given.

Would you accept "use mathematics to reduce the uncertainty of a football spot to triviality" as an alternative proposal?

Give me a break. You can have an intelligent discussion without being a pedant.

 

[Greg Bernhardt]

Ok folks the tangent is over. Let's bring this back on topic or posts will start being removed.

 

[gmax137]

RFIDs, sensors, etc. etc. are no good here, because Chet wants to investigate the position of the ball at a certain time in a game that already took place in 2016. He is not interested in any other game. So the question is, do we have available data from the 2016 game to definitively prove the particular "spot" in question was bad?

 

[boneh3ad]

RFIDs, sensors, etc. etc. are no good here, because Chet wants to investigate the position of the ball at a certain time in a game that already took place in 2016. He is not interested in any other game. So the question is, do we have available data from the 2016 game to definitively prove the particular "spot" in question was bad?

I suppose I interpreted this as requesting tech that could prevent these kinds of issues in the future, but you may have a point.

I would imagine doing this a posteriori would be effectively impossible since the ball is often obscured from view. I suppose you may be able to look at how a runner is carrying a ball and run some numbers on how balls typically behave and create a confidence interval on where it is located, but even that seems iffy to me. It would be based on a lot of assumptions.

 

[Dr. Courtney]

Video analysis is straightforward with a well marked playing surface.

For me, the challenge would be more in the error analysis (what are the uncertainties in position for any given "spot") and in the definition of "bad spot."

Sports fans tend to focus on numbers without consideration of uncertainty. If the spot should have been at the 40 yard line, a foot short of the first down line to gain, most fans would argue it should not have been a first down, regardless of the uncertainty in the spot from the video analysis. Given sufficiently high frame rate, video resolution, and an unobstructed view of the player and ball as he hit the ground, it should usually be possible to reconstruct a video spot to much better than a foot. With original video of high profile NCAA games, video quality is rarely the issue - the challenge in accuracy is with an unobstructed view.

Defining a "bad spot" also requires attention. I would think a good approach would be to reconstruct all the spots from a large number of games in the Big 10 to get an accurate distribution of official spot accuracy. Then one defines a "bad spot" something like outside 2 standard deviations from the mean. or worse than 95% of spots in the Big 10. If the accuracy of a given spot is only worse than 50-75% of spots in the given conference, it may be unfortunate, but it really was not that bad.

 

[eachus]

There are two contributions to spotting the football. Let's assume we have a football with an IoT tag that can be placed accurately to within a fraction of an inch, and the officials not only have the necessary tools to place the ball, but the needed training with the tools so they can do it quickly.

Now we have just a few problems remaining. Where was the ball when the ball carrier stepped out of bounds? No way to get that right without instrumenting the sidelines. Might as well do the same for the out-of-bounds at each end of the field. That leads to the problem of documenting the position of the ball when a runner's knee touches. This is more of a timing issue, and we could put a device on the official's whistle. Might not be perfect but about as good as you can get. Much better than the current situation where the delay in hearing the whistle depends on the distance from the listener--electronic or otherwise.

Now, what about fouls away from the ball? When the penalty is measured from the point of infraction, the referee's judgement is back in the game. One fix would be to define the penalty from the position of the ball at the time of infraction--or the line of scrimmage when pass interference occurs as the ball is being thrown.

Have I covered all cases? Probably not. The NFL is smart enough that if they introduced such technology, they would limit its use to specific situations to begin with. Until then, the NFL and NCAA will have to rely on the integrity of the officials. Notice that the Michigan--Ohio State call was one of the few cases where the team could appeal to the league, and perhaps get the call, and game result changed. This is much more common, if still rare, in major league baseball, where the game can be replayed from the point of the error. Michigan not appealing--whether successful or not--leads me to guess that the official call was right, or if wrong not provably so from all the video evidence.

 

[boneh3ad]

Video analysis is straightforward with a well marked playing surface.

For me, the challenge would be more in the error analysis (what are the uncertainties in position for any given "spot") and in the definition of "bad spot."

Sports fans tend to focus on numbers without consideration of uncertainty. If the spot should have been at the 40 yard line, a foot short of the first down line to gain, most fans would argue it should not have been a first down, regardless of the uncertainty in the spot from the video analysis. Given sufficiently high frame rate, video resolution, and an unobstructed view of the player and ball as he hit the ground, it should usually be possible to reconstruct a video spot to much better than a foot. With original video of high profile NCAA games, video quality is rarely the issue - the challenge in accuracy is with an unobstructed view.

Defining a "bad spot" also requires attention. I would think a good approach would be to reconstruct all the spots from a large number of games in the Big 10 to get an accurate distribution of official spot accuracy. Then one defines a "bad spot" something like outside 2 standard deviations from the mean. or worse than 95% of spots in the Big 10. If the accuracy of a given spot is only worse than 50-75% of spots in the given conference, it may be unfortunate, but it really was not that bad.

I would like to see said data then broken down by referee to show us which ones are actually blind and which ones are just accused of such by fans.

 

[256bits]

There is also the forward progress rule, which is interpretive by the referee as to stoppage of play and placement of the ball.
No much a camera or any sensing unit can do here.

 

[Nik_2213]

Given my earlier comment about 'goal mouth technology', I was primed to notice a recent feature on the BBC TV's 'Click' IT program. This reported on commercial 'full 3-D' match recording, using dozens (40+) of 4K video cameras around the field plus 'cloud computing' to provide real-time, immersive, fly-through and play-back. In effect, it turned the event into VR, to be viewed / reviewed from any position, at any angle, at any speed.

I... I know just enough about algorithms & computational overheads of 3D modelling, CAD etc to be seriously impressed...

Okay, a surfeit of challenges would seriously slow most games, but a strict team limit should work. Also, a team may earn much kudos for challenging a marginal call that went in their favour...

 

[member 656954]

Obviously, plays towards the end of the game have a heavy influence on the outcome of a game.

Interestingly, 'Jeopardy!' contestant James Holzhauer adopts the exact opposite of this in his game play, and I head him on a NPR podcast say that coaches should go hard early because the risk is lower and the payoff higher. Early success places the team in a better position later in the game, yet, it seems that many games are played exactly opposite to this, leading to that heavy influence.

 

[Comeback City]

Interestingly, 'Jeopardy!' contestant James Holzhauer adopts the exact opposite of this in his game play, and I head him on a NPR podcast say that coaches should go hard early because the risk is lower and the payoff higher. Early success places the team in a better position later in the game, yet, it seems that many games are played exactly opposite to this, leading to that heavy influence.

That's what I was talking about in my first post (#3)... the entirety of the game is important, obviously.