[larkinforever]

[quote author=TRD link=topic=16412.msg186122#msg186122 date=1484248782]
[quote author=gocard14 link=topic=16412.msg186100#msg186100 date=1484224794]


A play calling AI would be really interesting to see,

I think this is a very good idea.

Watson > Shaw, any day of the week.  (And I don't mean DeShaun Watson).
[/quote]

I am sure somebody, probably several groups, have tried it. Hard to score though, not like chess where the play call determines the outcome. "Good enough" plays well executed >> "Optimal" plays poorly executed. And it is hard to know, and tell the computer, which players can reliably do what etc. Also, you have to get a team to actually use the AI result or it won't be known if the plays were any good. It would be a fun experiment though.
[/quote]

The AI computer might be able to sift through all of that instantaneously, though - each player's likelihood of executing well on that particular play, including a sensitivity analysis for each player's execution, each opponent player's execution of his assignment or even things like likelihood to go rogue, etc., etc. (much beyond an algorithm, however simple or complex, it could almost predict the future, perhaps!)... :D (Minority Report...but for football.)

[larkinforever]

[quote author=TRD link=topic=16412.msg186122#msg186122 date=1484248782]
[quote author=gocard14 link=topic=16412.msg186100#msg186100 date=1484224794]


A play calling AI would be really interesting to see,

I think this is a very good idea.

Watson > Shaw, any day of the week.  (And I don't mean DeShaun Watson).
[/quote]

I am sure somebody, probably several groups, have tried it. Hard to score though, not like chess where the play call determines the outcome. "Good enough" plays well executed >> "Optimal" plays poorly executed. And it is hard to know, and tell the computer, which players can reliably do what etc. Also, you have to get a team to actually use the AI result or it won't be known if the plays were any good. It would be a fun experiment though.
[/quote]

The AI computer might be able to sift through all of that instantaneously, though - each player's likelihood of executing well on that particular play, including a sensitivity analysis for each player's execution, each opponent player's execution of his assignment or even things like likelihood to go rogue, etc., etc. (much beyond an algorithm, however simple or complex, it could almost predict the future, perhaps!)... :D (Minority Report...but for football.)

[CTcard]

In what world is shifting to a shotgun on third and six inches in the Sun Bowl and then pitching it to the short side a good call, especially for a "power running" team?

Not exactly on the topic of your post, but I did want to throw in a word about this one because it was a fascinating play in many aspects.

The pre-snap shift - which included going to the shotgun and unbalancing the line by two - completely threw off UNC. It created a situation where Stanford had a massive manpower advantage on the left side, which by definition means Stanford had a big manpower deficit on the right side.

ANY play to the left would have scored a TD at that point, outside of somebody tripping over their shoelaces. Then, strangely, we ran the play to the right.

It seems as if Stanford expected UNC to react more properly to the shift, with the default play trying to then "catch" them going back to the weak side. However, it would seem impossible to design such a play/shift without some option to go to the strong side if the defense reacted anything like the manner in which it did. I would have to think Burns missed that.

Anyway, I found it fascinating because a setup that looked to have produced a fantastic situation for an easy TD ended up a significant loss with really no hope of success - presumably because a check was missed at the line.

[CTcard]

In what world is shifting to a shotgun on third and six inches in the Sun Bowl and then pitching it to the short side a good call, especially for a "power running" team?

Not exactly on the topic of your post, but I did want to throw in a word about this one because it was a fascinating play in many aspects.

The pre-snap shift - which included going to the shotgun and unbalancing the line by two - completely threw off UNC. It created a situation where Stanford had a massive manpower advantage on the left side, which by definition means Stanford had a big manpower deficit on the right side.

ANY play to the left would have scored a TD at that point, outside of somebody tripping over their shoelaces. Then, strangely, we ran the play to the right.

It seems as if Stanford expected UNC to react more properly to the shift, with the default play trying to then "catch" them going back to the weak side. However, it would seem impossible to design such a play/shift without some option to go to the strong side if the defense reacted anything like the manner in which it did. I would have to think Burns missed that.

Anyway, I found it fascinating because a setup that looked to have produced a fantastic situation for an easy TD ended up a significant loss with really no hope of success - presumably because a check was missed at the line.

[washingtonismoney]

Right. If anything the catastrophic shift play shows the weakness of blaming everything on play calling, as you could plausibly argue either Shaw should've called a timeout (he tried, allegedly) or the players should've been able to count. If you *really* want to blame the offensive coaching staff for that play, I'd argue the problem comes from the underlying coaching -- i.e. that the players are unable to think on their feet and adapt when the situation scotches Plan A. At any rate it's hard to say the problem is the playcall.

[washingtonismoney]

Right. If anything the catastrophic shift play shows the weakness of blaming everything on play calling, as you could plausibly argue either Shaw should've called a timeout (he tried, allegedly) or the players should've been able to count. If you *really* want to blame the offensive coaching staff for that play, I'd argue the problem comes from the underlying coaching -- i.e. that the players are unable to think on their feet and adapt when the situation scotches Plan A. At any rate it's hard to say the problem is the playcall.

[Goose]

In what world is shifting to a shotgun on third and six inches in the Sun Bowl and then pitching it to the short side a good call, especially for a "power running" team?

The literal answer to the question is "In a world where the defense reacts as they have been instructed". Usually, that is a pretty good strategy. Probably not a good idea expecting the other team to be idiots. As others have pointed out, it would have been nice if Burns had realized he was dealing with a mistake and just run left on his own. At worst, he should have called time out. It may be that they were running a simplified offense, since he was the backup,  and he had no option to audible, and maybe not enough time even if he did have an audible available.
FWIW, I also would argue that we were not really a "Power running" team this year. Our record near the goal line speaks for itself. When CMac was present, we could kind of run inside, but nearly always we ran outside the Guards and very often we ran outside the Tackles/TE. With Love, we probably ran inside even less. This year, we also were not moving good teams on the LOS reliably. So we may have thought we were a power team, but we often didn't act like it.

[Goose]

In what world is shifting to a shotgun on third and six inches in the Sun Bowl and then pitching it to the short side a good call, especially for a "power running" team?

The literal answer to the question is "In a world where the defense reacts as they have been instructed". Usually, that is a pretty good strategy. Probably not a good idea expecting the other team to be idiots. As others have pointed out, it would have been nice if Burns had realized he was dealing with a mistake and just run left on his own. At worst, he should have called time out. It may be that they were running a simplified offense, since he was the backup,  and he had no option to audible, and maybe not enough time even if he did have an audible available.
FWIW, I also would argue that we were not really a "Power running" team this year. Our record near the goal line speaks for itself. When CMac was present, we could kind of run inside, but nearly always we ran outside the Guards and very often we ran outside the Tackles/TE. With Love, we probably ran inside even less. This year, we also were not moving good teams on the LOS reliably. So we may have thought we were a power team, but we often didn't act like it.

[BostonCard]

Also, in a world where your first two running plays from the one failed to make it into the end zone.

Had Burns run the other way we would be debating whether it was a smart play all or we were just lucky that UNC shifted the wrong way.

BC

[BostonCard]

Also, in a world where your first two running plays from the one failed to make it into the end zone.

Had Burns run the other way we would be debating whether it was a smart play all or we were just lucky that UNC shifted the wrong way.

BC

[Goose]

Also, in a world where your first two running plays from the one failed to make it into the end zone.

Fully and heartily agree.

Had Burns run the other way we would be debating whether it was a smart play all or we were just lucky that UNC shifted the wrong way.

BC

Even better, had he run the other way and somehow not scored, we would be arguing about which tree to hang him from, and should Shaw be next for allowing such insubordinate behavior.

[Goose]

Also, in a world where your first two running plays from the one failed to make it into the end zone.

Fully and heartily agree.

Had Burns run the other way we would be debating whether it was a smart play all or we were just lucky that UNC shifted the wrong way.

BC

Even better, had he run the other way and somehow not scored, we would be arguing about which tree to hang him from, and should Shaw be next for allowing such insubordinate behavior.

[CTcard]

Also, in a world where your first two running plays from the one failed to make it into the end zone.

Actually, repeating the second down play (with the same results) would have scored a TD. [And I am not completely sure he didn't get in on that play - but he certainly didn't get in definitively.]

The first down play would have been a TD had Scarlett just blasted into the line instead of taking his time and trying to juke outside, giving the guy coming around from the backside time to catch him. The OL really did clear bodies out of the way on that one - a frustrating miss.

[CTcard]

Also, in a world where your first two running plays from the one failed to make it into the end zone.

Actually, repeating the second down play (with the same results) would have scored a TD. [And I am not completely sure he didn't get in on that play - but he certainly didn't get in definitively.]

The first down play would have been a TD had Scarlett just blasted into the line instead of taking his time and trying to juke outside, giving the guy coming around from the backside time to catch him. The OL really did clear bodies out of the way on that one - a frustrating miss.

[BostonCard]

Except there would have been no way of knowing how the original play was drawn up, and I don't think Shaw would have thrown his QB under the bus.  Even now, he only stated that he was trying to get a time out, not that Burns blew it by not recognizing the play (though the latter is inferred).

BC

[BostonCard]

Except there would have been no way of knowing how the original play was drawn up, and I don't think Shaw would have thrown his QB under the bus.  Even now, he only stated that he was trying to get a time out, not that Burns blew it by not recognizing the play (though the latter is inferred).

BC

[Goose]

Except there would have been no way of knowing how the original play was drawn up, and I don't think Shaw would have thrown his QB under the bus.  Even now, he only stated that he was trying to get a time out, not that Burns blew it by not recognizing the play (though the latter is inferred).

BC

Your right. We would just be where we are now, questioning "Why the **** did he call THAT play".

[Goose]

Except there would have been no way of knowing how the original play was drawn up, and I don't think Shaw would have thrown his QB under the bus.  Even now, he only stated that he was trying to get a time out, not that Burns blew it by not recognizing the play (though the latter is inferred).

BC

Your right. We would just be where we are now, questioning "Why the **** did he call THAT play".

[jacketree]

(still need to see that flick)

Check this one out.

https://www.youtube.com/watch?v=vMuD7I9clNc

Total AYFKM.  But it worked.

I could probably spend a day hunting clips down and post a few hundred.  Especially in the NFL.  They throw on 3rd and short ALL THE TIME.  I think Coach Shaw really calls games like an NFL coach.

[jacketree]

(still need to see that flick)

Check this one out.

https://www.youtube.com/watch?v=vMuD7I9clNc

Total AYFKM.  But it worked.

I could probably spend a day hunting clips down and post a few hundred.  Especially in the NFL.  They throw on 3rd and short ALL THE TIME.  I think Coach Shaw really calls games like an NFL coach.

[larkinforever]

The risk-reward calculation just might pencil out on that play.

The weirder ones, I think, are where your team is looking at 3rd-and-1 or 4th-and-1 and your "unexpected"/"surprise"/"break tendencies" play call might reasonably be expected to yield at best only 4 or 5 yards or so, but means, of course, that you're moving away from your bread-and-butter and probably at a fairly high tradeoff in expected conversion percentage.

[larkinforever]

The risk-reward calculation just might pencil out on that play.

The weirder ones, I think, are where your team is looking at 3rd-and-1 or 4th-and-1 and your "unexpected"/"surprise"/"break tendencies" play call might reasonably be expected to yield at best only 4 or 5 yards or so, but means, of course, that you're moving away from your bread-and-butter and probably at a fairly high tradeoff in expected conversion percentage.

[TRD]

I think the EE/CS dept. should get on this AI play calling idea.  #MSPCGA  (make Stanford play calling great again)  ::)

[TRD]

I think the EE/CS dept. should get on this AI play calling idea.  #MSPCGA  (make Stanford play calling great again)  ::)

[Goose]

I think the EE/CS dept. should get on this AI play calling idea.  #MSPCGA  (make Stanford play calling great again)  ::)

I'm not so sure. The quarterback would have to look at a lot of adds and then get a play selected by a vote on social media. ;)

[Goose]

I think the EE/CS dept. should get on this AI play calling idea.  #MSPCGA  (make Stanford play calling great again)  ::)

I'm not so sure. The quarterback would have to look at a lot of adds and then get a play selected by a vote on social media. ;)

[washingtonismoney]

Someone made a point re: play-calling becoming unpredictable, but actually game-playing AI have quite unconventional playing styles. In chess you often see AI making very odd moves that make no (human) sense, at least the way humans have traditionally been educated. I know Go players say the same thing of the recently omnipotent Go-playing computers. So theoretically I could see a football play-calling AI coming up with very weird ideas.

That's the thing about AI -- whether in stock trading, medical advice, whatever -- it's not trained to discard possibilities because of implausibility, as humans are. . Of course that's both a strength and a weakness.

[washingtonismoney]

Someone made a point re: play-calling becoming unpredictable, but actually game-playing AI have quite unconventional playing styles. In chess you often see AI making very odd moves that make no (human) sense, at least the way humans have traditionally been educated. I know Go players say the same thing of the recently omnipotent Go-playing computers. So theoretically I could see a football play-calling AI coming up with very weird ideas.

That's the thing about AI -- whether in stock trading, medical advice, whatever -- it's not trained to discard possibilities because of implausibility, as humans are. . Of course that's both a strength and a weakness.

[Goose]

Someone made a point re: play-calling becoming unpredictable, but actually game-playing AI have quite unconventional playing styles. In chess you often see AI making very odd moves that make no (human) sense, at least the way humans have traditionally been educated. I know Go players say the same thing of the recently omnipotent Go-playing computers. So theoretically I could see a football play-calling AI coming up with very weird ideas.

That's the thing about AI -- whether in stock trading, medical advice, whatever -- it's not trained to discard possibilities because of implausibility, as humans are. . Of course that's both a strength and a weakness.

Also, many (most?) AI systems are learning-based, not rule based. So they would have to keep their jobs long enough to "learn" how to call plays. It is pretty easy to get a small group of human chess/go players to play against the AI program. Much harder to get 22 representative football players. Also, an AI program can play chess against another AI program and learn that way. Can't do that in football.

[Goose]

Someone made a point re: play-calling becoming unpredictable, but actually game-playing AI have quite unconventional playing styles. In chess you often see AI making very odd moves that make no (human) sense, at least the way humans have traditionally been educated. I know Go players say the same thing of the recently omnipotent Go-playing computers. So theoretically I could see a football play-calling AI coming up with very weird ideas.

That's the thing about AI -- whether in stock trading, medical advice, whatever -- it's not trained to discard possibilities because of implausibility, as humans are. . Of course that's both a strength and a weakness.

Also, many (most?) AI systems are learning-based, not rule based. So they would have to keep their jobs long enough to "learn" how to call plays. It is pretty easy to get a small group of human chess/go players to play against the AI program. Much harder to get 22 representative football players. Also, an AI program can play chess against another AI program and learn that way. Can't do that in football.