About those "win probability" charts ...

[Dome]

5 minutes ago, iPwn said:

There you go. 

Thank ya. Wasn't questioning you I just never see that stuff laid out anywhere and thought it might be behind a paywall. Never heard of SportingCharts. 

[pwny]

13 minutes ago, domepatrol91 said:

Wasn't questioning you I just never see that stuff laid out anywhere and thought it might be behind a paywall.

Oh, no understood. The last line there was acknowledgment that’s I spent way too much time explaining my method for something with a ton of estimations and rounding.

[Jakuvious]

@iPwn is the real MVP of this thread. Excellent posts.

Additionally, someone else mentioned it, but confirmation bias is absolutely in play here as well. The thread suggests one incident, where they thought the comeback was more likely an it happened. But a thread like this isn't made every time there's a 2% win probability, and the comeback doesn't happen. Effectively, you had a thought, one piece of evidence in your favor followed that thought, and you then projected that to mean the probabilities must be wrong. But probability doesn't work that way. It's akin to someone winning the lottery, and concluding the odds can't actually be that low.

[Quark3]

On 11/20/2017 at 8:23 PM, Jlowe22 said:

The probabilities are probably based on NFL averages, they go up some when you have a team and a QB that excels at throwing the ball, and a sketchy defense.

Still, it's rare.  15 points, not 14.  That means drive the length of the field quickly, get an immediate defensive stop, drive the length of the field quickly, score a two point conversion.  Make a stop in overtime, drive the length of the field for a field goal.

There was also one more stop the Saints needed to make, since Saints scored so quickly Reskins still had enough time to drive for a field goal.  They probably would have gotten the try off, if not for Cousins intentional grounding. 

Which, btw, the NFL told the Skins the other day was the wrong call. It looked wrong to me in the stands. The QB took a quick step back and fired the ball to the sidelines, where no receiver was, because the receiver broke in on  a slant and the QB expected him to run a quick out. Obvious miscommunication not IG, but the refs called it, and it was devastating to Washington. 

[Quark3]

22 minutes ago, Jakuvious said:

@iPwn is the real MVP of this thread. Excellent posts.

Additionally, someone else mentioned it, but confirmation bias is absolutely in play here as well. The thread suggests one incident, where they thought the comeback was more likely an it happened. But a thread like this isn't made every time there's a 2% win probability, and the comeback doesn't happen. Effectively, you had a thought, one piece of evidence in your favor followed that thought, and you then projected that to mean the probabilities must be wrong. But probability doesn't work that way. It's akin to someone winning the lottery, and concluding the odds can't actually be that low.

Two things. First, I'm not that good. I'm known for being a biased wanker, i view games with my hopes and wishes not head, which is why i smartly don't gamble on football. And yet I thought the Saints had a good chance to win from that position. Two scores in 6 minutes? That's two, two-minute drives, and then another two minutes to force a three and out. Less than 50%, sure, but such that it would be nothing special if they did. Defenses in that situation tend to play prevent, and that opens things up for quick passes down the field.  It was very easy to visualize Brees going 7/9 for 80 yards in 2 or so minutes. To me, regardless of what the stats say, it's no surprise when that happens at all.

Seriously, if you were watching some other game, and you were taken to Saints highlights and the analyst mentioned that Brees did that twice in a quarter at home against a 4-5 opponent, would you be shocked? Would your first reaction be "wow, that's pretty amazing, probably 1 in 100 chance of that happening"? Not to me.

Second, the lottery analogy doesn't really work because even though in the Powerball, the odds that I win are infinitesimally small, the odds that someone will win any given drawing are pretty high. With football, the 99% against doesn't have to happen. In this case, if the model was accurate it shouldn't have. But it did. In short, I'm aware of confirmation bias and file-drawer problems and all that. Seems like these models include a lot of information that raises the probabilities above what they should be. 

It's no defense IMO to say "you're arguing against the math", because the math is only as good as the input data - particularly how much data and how far back does it go - and the structure of the model, and we don't know those. 

[Jlowe22]

12 minutes ago, Quark3 said:

Two things. First, I'm not that good. I'm known for being a biased wanker, i view games with my hopes and wishes not head, which is why i smartly don't gamble on football. And yet I thought the Saints had a good chance to win from that position. Two scores in 6 minutes? That's two, two-minute drives, and then another two minutes to force a three and out. Less than 50%, sure, but such that it would be nothing special if they did. Defenses in that situation tend to play prevent, and that opens things up for quick passes down the field.  It was very easy to visualize Brees going 7/9 for 80 yards in 2 or so minutes. To me, regardless of what the stats say, it's no surprise when that happens at all.

Seriously, if you were watching some other game, and you were taken to Saints highlights and the analyst mentioned that Brees did that twice in a quarter at home against a 4-5 opponent, would you be shocked? Would your first reaction be "wow, that's pretty amazing, probably 1 in 100 chance of that happening"? Not to me.

Second, the lottery analogy doesn't really work because even though in the Powerball, the odds that I win are infinitesimally small, the odds that someone will win any given drawing are pretty high. With football, the 99% against doesn't have to happen. In this case, if the model was accurate it shouldn't have. But it did. In short, I'm aware of confirmation bias and file-drawer problems and all that. Seems like these models include a lot of information that raises the probabilities above what they should be. 

It's no defense IMO to say "you're arguing against the math", because the math is only as good as the input data - particularly how much data and how far back does it go - and the structure of the model, and we don't know those. 

The probability of that occurring is not nearly as likely as you think.  In fact, Brees has never made a comeback from that far down with that little time left.

[Jakuvious]

4 minutes ago, Quark3 said:

Two things. First, I'm not that good. I'm known for being a biased wanker, i view games with my hopes and wishes not head, which is why i smartly don't gamble on football. And yet I thought the Saints had a good chance to win from that position. Two scores in 6 minutes? That's two, two-minute drives, and then another two minutes to force a three and out. Less than 50%, sure, but such that it would be nothing special if they did. Defenses in that situation tend to play prevent, and that opens things up for quick passes down the field.  It was very easy to visualize Brees going 7/9 for 80 yards in 2 or so minutes. To me, regardless of what the stats say, it's no surprise when that happens at all.

Seriously, if you were watching some other game, and you were taken to Saints highlights and the analyst mentioned that Brees did that twice in a quarter at home against a 4-5 opponent, would you be shocked? Would your first reaction be "wow, that's pretty amazing, probably 1 in 100 chance of that happening"? Not to me.

Second, the lottery analogy doesn't really work because even though in the Powerball, the odds that I win are infinitesimally small, the odds that someone will win any given drawing are pretty high. Except with football, the 99% against doesn't have to happen. In this case, it shouldn't have. But it did. In short, I'm aware of confirmation bias and file-drawer problems and all that. Seems like these models include a lot of information that raises the probabilities above what they should be. 

 

Not being surprising does not change the fact that it is improbable. As a Chiefs fan, it was never surprising to me when Jamaal Charles had a 60+ yard TD. The capability was there. The opportunity was there. But it doesn't change that he did that a handful of times on almost 2000 touches. It never surprised me or shocked me, but it happened less than 1% of the time he touched the ball. Something can be unsurprising and still improbable. You're conflating those two things a lot when they are not the same. One is a matter of subjective expectation. The other is pure statistically likelihood.

You also keep cutting short the situation. In this post, you say it's easy to visualize Brees going 80 yards in 2 minutes. Sure. But it's not that. It's 80 yards twice with at least one defensive stop (it ended up being two) and a completed PAT and a completed 2 point conversion, all in under 6 minutes, and winning the coin toss in over time and scoring or losing the coin toss AND getting another stop AND scoring. It ended up being 3 scoring drives and 3 defensive stands, with 2 of each needing to come within that 6 minute window. Even if you think all of those individual things are more likely to happen than not, the aggregate of them is incredibly improbable. That's just the reality of it.

Hell, you act like it's a throwaway for Brees to score 2 TDs in one quarter. Brees has 2 passing touchdowns in less than half of his career full games. Let alone a quarter. Let alone in a 6 minute timeframe.

[Dome]

1 hour ago, Quark3 said:

Obvious miscommunication not IG, but the refs called it, and it was devastating to Washington. 

It was a bad call for sure.

But allowing Brees to go 11/11 for 164 and 2 TDs and Ingram to go 51 yards on two carries in OT is what killed them. The bad call was like a flat tire on the ambulance.

Bad calls happen every single game and it really really sucks when they come in the final minutes of a close game, but the Skins just laid down for the final 4 minutes of this one. 

[pwny]

2 hours ago, Quark3 said:

With football, the 99% against doesn't have to happen. In this case, if the model was accurate it shouldn't have.

Yeah...that’s not how this works. If the 1% wasn’t possible, it would be 100%. Largely improbable is not the same as impossible.

To make a come back, it takes a number of events happening. Those events stack probability and become increasingly harder to complete.

This model ends up giving New Orleans 53% chance of scoring each TD. It only says they need to stop Washington once, saying nothing of the second stop they needed. It gives them a 90% chance of winning should the game go to OT. And still the probability could only get to 2.5%. 

[Dome]

2 hours ago, Quark3 said:

With football, the 99% against doesn't have to happen. In this case, if the model was accurate it shouldn't have. 

On 11/20/2017 at 9:34 PM, domepatrol91 said:

I think a big part of the issue is people see 99.7 or 99.3 (or whatever the Saints/Skins or 28-3 percentages were) and just go “oh that’s impossible to come back from” and assume the stat is broken when it actually happens because they didn’t take the stat for what it was.

 

2 hours ago, Quark3 said:

Seriously, if you were watching some other game, and you were taken to Saints highlights and the analyst mentioned that Brees did that twice in a quarter at home against a 4-5 opponent, would you be shocked? Would your first reaction be "wow, that's pretty amazing, probably 1 in 100 chance of that happening"? Not to me.

It wasn't just that Brees scored twice in one quarter though. It was twice in the last 6 minutes, with a defensive stop in between and then a 2pt conversion after the 2nd, and then after that 2 pt conversion another defensive stop to end the 4th, and then another defensive stop in OT, and then an offensive drive and FG in OT as well.

Calling it 2 TDs in one quarter is really selling it short and leaves out the context that explains why they were only given a 1% chance to win.

[sp6488]

On 11/20/2017 at 7:40 PM, scar988 said:

IT was definitely 98% against it. Because in 98% of those 31-16 games with 6 minutes left, the team up 31-16 wins.

TBF, that's 1/50 pulling through, which doesn't seem totally farfetched. 

[sp6488]

On 11/20/2017 at 9:23 PM, Jlowe22 said:

The probabilities are probably based on NFL averages, they go up some when you have a team and a QB that excels at throwing the ball, and a sketchy defense.

Still, it's rare.  15 points, not 14.  That means drive the length of the field quickly, get an immediate defensive stop, drive the length of the field quickly, score a two point conversion.  Make a stop in overtime, drive the length of the field for a field goal.

There was also one more stop the Saints needed to make, since Saints scored so quickly Reskins still had enough time to drive for a field goal.  They probably would have gotten the try off, if not for Cousins intentional grounding. 

There are 250 some odd NFL games played every year.  Something with 1/100 chance has a good chance of happening more than once over the course of a season.

Edit:  When discussing the comebacks, you have to narrow it down to games where a comeback was actually needed.

And that's the rub.  You have to do all of that (2 scores including the 2PC) just to get the game to 50-50.

[sp6488]

On 11/21/2017 at 9:04 AM, Malik said:

I wish statistics was a required math course to graduate high school.

Along with econ...

[ninjapirate]

On 11/21/2017 at 6:04 AM, Malik said:

I wish statistics was a required math course to graduate high school.

I took a high school and college level stats class. A lot of people didn't seem to get it even when taking those courses. 

[Quark3]

On 11/22/2017 at 3:49 PM, Jakuvious said:

Not being surprising does not change the fact that it is improbable. As a Chiefs fan, it was never surprising to me when Jamaal Charles had a 60+ yard TD. The capability was there. The opportunity was there. But it doesn't change that he did that a handful of times on almost 2000 touches. It never surprised me or shocked me, but it happened less than 1% of the time he touched the ball. Something can be unsurprising and still improbable. You're conflating those two things a lot when they are not the same. One is a matter of subjective expectation. The other is pure statistically likelihood.

You also keep cutting short the situation. In this post, you say it's easy to visualize Brees going 80 yards in 2 minutes. Sure. But it's not that. It's 80 yards twice with at least one defensive stop (it ended up being two) and a completed PAT and a completed 2 point conversion, all in under 6 minutes, and winning the coin toss in over time and scoring or losing the coin toss AND getting another stop AND scoring. It ended up being 3 scoring drives and 3 defensive stands, with 2 of each needing to come within that 6 minute window. Even if you think all of those individual things are more likely to happen than not, the aggregate of them is incredibly improbable. That's just the reality of it.

Hell, you act like it's a throwaway for Brees to score 2 TDs in one quarter. Brees has 2 passing touchdowns in less than half of his career full games. Let alone a quarter. Let alone in a 6 minute timeframe.

Well, if something really is extremely unlikely, it *should* be surprising to us. E.g., i know if i won the powerball  tomorrow, i'd be dumbfounded with surprise. On the other hand, if won a coin flip, i wouldn't be. If something isn't surprising, it could mean the event in question isn't that unlikely.

And in cutting the scenario short, i was just saving space. If you notice, in the original scenario, i sketched the full thing out, at least to OT (brees down the field, three and out, brees down the field again, with one of the two scores ending in a 2-pt conversion) to tie at 31 and in to OT. 

As it was, it wasn't difficult at all. If you watched the plays, the Saints did nothing special. On both drives, Brees went into the gun, had plenty of time, threw to wide open receivers well down the field who gained yards and got out of bounds easily. There weren't any thread the needle throws, no miracle catches, no clutch 4th and 10 conversions. And the runners gained big yards on run plays thanks to huge holes. There wasn't anything special they pulled off on either drive, and they didn't need an onside kick - a usual sign of a desperate situation - either. They tied the game with so much time to spare that if it wasn't for that blown grounding call, the skins probably kick a winning FG in regulation.

 On the defensive stop, they did stuff the skins on a critical 3rd and one, but that was because DC stupidly ran a slow-developing slide down the line run that gives stacked lineman and blitzers plenty of time to penetrate and blow up the play. I told my wife "just QB sneak it, a quick punch straight ahead, don't give the Saints time to penetrate the backfield. Nothing slow-developing". And the skins ran a slow-developing run (doh!). IOW's, there just wasn't anything 99 to 1 about it.

It's kind of like those college football models where they look at a 5-game stretch and say Alabama has a 33% chance to win out. That's because some data says they have about a 80% chance to win each game and if you do the math you guys like, it says an overall 33% chance. But you just eyeball the 5 games, and you know Alabama isn't losing any of them, and it would actually be a much bigger surprise if they did lose any one of them. The input values just aren't correct.

In this case, whatever model produced the 99.6% probability seemingly missed critical values and variables. It either drew on data that wasn't relevant (e.g., games from years past when there was no 2-pt conversion) or didn't factor in characteristics of the situation (Brees at home vs a bad team) to derive that %.

IMHO.