In-Game Win Probability


Name:

Comment:


Refresh comments Note-Comments disabled during offseason.


Brian said...
Ok. Figured it out. Culliver's position was listed as 'NB', which is non-standard. I'm not going to rerun the version here for fear of contaminating the main database, but the correct boxscore can be found at the Graphs|Previous Seasons|... link.


Brian said...
That's weird. I still don't understand why. He had 4 tackles and 3.3 +EPA. I'll keep digging.


James said...
Hey, where's Culliver in the recap? He had 0.13 +WPA on Ryan's interception!


James said...
Anon, you can see from the chart the huge WP drop for Atlanta on Ryan's final two incompletions. He had -0.43 WPA on those two plays alone, while a TD was worth 0.35. Those plays were a swing of 0.78 WPA!


Anonymous said...
Brian, I found it surprising that M.Ryan's WPA was so low. Thoughts on why? INT and fumble? Just curious. Thanks.


Brian said...
Still waiting for final gamebook from the league.


Snowbody said...
Ray - I thought about the intentional safety idea too, but it's way risky. Too much could go wrong -- you don't manage to get into the end zone/the defense forces a fumble/a short kick->fair catch kick for 3 points or a FG to win. Punting it away forces a TD to win.


Brian said...
PatP-A WP estimate is not an estimate one team would never lose the lead. One reason a WP may be as high as say 0.98 is because it includes the many eventualities where the lead is blown but regained. Hope that helps.


PatPatriot said...
Sry about that Bal error, distracted! Anyway, let's look at the big picture. Maybe the NFL's just crazy this year but this stat seems like an anomaly. Just in the last two weeks, 3 different teams with a 98% chance of winning either lost or lost the lead. Denver did it, and Atlanta did it twice. In addition, Seattle lost after gaining a 92% chance of winning. So, out of five games(NE-BALT on now), 4 teams lost or lost the lead after gaining these huge numbers. If these are the true numbers, what are the odds against this happening? The number is so high as to be unrealistic. Yet it happened. So the question is, were the numbers too high, or is the NFL just insane this year?


Jonathan said...
It's crazy to think that the Atlanta failed 4th down at the end of the game was (so far) the most high leverage situation of NEW ENGLAND'S season.


Ray said...
in the 4th quarter with 13 seconds left, the niners had the ball at their own 15 and it was 4th down. They punted but can an argument be made for running backwards and taking as much time off the clock as possible and just take a safety? They were up 28-24 and on the safety kickoff there would be like a few seconds left at best.


Brian said...
One of the big keys to SFs win was that they uncharacteristically limited their penalties to 3 for 19 yds.


QCIC said...
Pat- A) Hou-NE, not BAL-NE B) Absolutely the numbers chnage that much. Don't be daft. Look at the difference in the moneyline in Vegas for a -6 game vs a pickem game. That play was worth maybe 4 pts, possibly more. That is a huge swing. Now granted at that point HOU probably wasn't 2:1, but that is because the simulator is using the situation, not adjusting for the strength of the teams, though I would bet quite a bit of money the strength of the teams matters a lot less than you would think.
8:23 in the 4th qtr

PatPatriot said...
Really? I refer you the the opening kick in last week's NE Baltimore game. One good runback and the Ravens now are 2-1 favorites to win, with almost 60 minutes to play. At that point, would you have taken a wager on the Ravens at 1-2 odds? I think not. I think the numbers can change too much on just one play. Today, when the Niners FG hit the post, they dropped 20% points. Way too much for my, um, common sense. One obvious problem is that the data is for all teams, not that particular one. And, of course, is there actually enough data to make these kinds of predictions? I don't claim to know all the answers here, but is meant as positive feedback.


QCIC said...
Pat- Vague comments like "sometimes the numbers here defy common sense" is not a criticism anyone here takes seriously especially when the common sense is so badly mistaken. If you want to contribute something constructive mention a specific instance where you feel the model is mistaken.


adam h said...
"Sometimes the numbers here simply defy common sense." Usually the problem is with common sense.


PatPatriot said...
Just to be clear, I'm NOT referring to the end of the game when the data wasn't available. I mean when it's actually working.


PatPatriot said...
This a great idea and all your work making this happen is appreciated. However, I think your defensiveness is blinding you to seeing the validity of what The Strategy Expert is saying. Sometimes the numbers here simply defy common sense. That tells me there's a problem with the model. I understand there's no perfect model so it can always be improved upon. If you're going to bother making the huge effort to post this stuff, I would think you would want it to be as accurate as possible.


Brian said...
As I mentioned, the league's data feed is still broke. I'll have final numbers and graph when the gamebook is posted. Refunds to everyone!


Brian said...
Strat Expert, Nothing wrong with my system. NFL's data feed is broken. As for difference between gamblers and actual league averages, I'll leave that to the 'experts' like yourself. Pls do a search on the main page for 'win probability model accuracy' and that might help.


Yippy Skippy said...
I'm guessing The (self proclaimed) Strategy Expert still thinks his high-school injury is all that kept him from the NFL.


QCIC said...
And yet "Strategy Expert" I still find it extremely useful and vastly superior to the graphs you post for free.


Thank you Brian said...
The (self-proclaimed) Strategy Expert --- why not contribute rather than complain?


Flash said...
Don't like it? Don't use it.
8:23 in the 4th qtr

The Strategy Expert said...
This chart system is broken, also there are many inconsistencies regarding the odds adjustment calculations from many obvious points if you track this closely. It's a great idea but a very flawed system that is not properly calculating true probabilities. Clean it up guys please.
8:23 in the 4th qtr

Anthony said...
Why is the win% graph still at 8:23?
8:23 in the 4th qtr

Dan said...
Well as a start, is there a (in this case inverse) correlation between time on the play clock left and play efficiency. Although there is a ton of sample bias there, since winning teams are more likely to snap with the clock low.
8:23 in the 4th qtr

Like a boss said...
Go 9ers
8:23 in the 4th qtr

Brian said...
Good question Dan. Would be hard to untangle.
8:23 in the 4th qtr

J said...
See, the thing about passing so often is your Offensive lineman get tired of back peddling and putting all the weight on their Achilles. Run blocking is the opposite and defensive line wears out. It is very difficult to measure the effect one play has on future plays but you can see the 49ers run heavy gameplan has worn out the falcons front and the falcons o line is worn out. That is why falcons are finding it hard to maintain a lead. Just look at how much more significant the pass rush is
8:23 in the 4th qtr

Dan said...
Well I guess it's clear which they chose.
8:23 in the 4th qtr

Dan said...
Have you ever studied play speed vs play efficiency? I.e. should Atlanta try to squeeze two posessions out of this 3 minutes or is it more important to get the right play.
8:23 in the 4th qtr

Brian said...
Data feed is lagging. Sorry folks. Close to 50/50 now.
8:23 in the 4th qtr

Brian said...
Probably upheld but definitely worth the challenge at this point.
8:23 in the 4th qtr

James S said...
At least it's a marginal improvement over "can't win in the playoffs."
8:23 in the 4th qtr

Brian said...
New narrative for ATL : can't hold big 2nd half leads in big games
11:41 in the 4th qtr

Dan said...
The best part of the momentum narrative is how momentum can be "swung". So to even people who "believe" in it, it's totally meaningless!
13:22 in the 4th qtr

Ken R said...
THAT was a high leverage play.
13:24 in the 4th qtr

Brian said...
Dear God.
14:13 in the 4th qtr

Brian said...
What happened to all that momentum now, Troy?
1:37 in the 3rd qtr

Brian said...
Right Troy. ATL is moving the ball because of the momentum from the missed fg. ...because they've been stopped cold until now....
1:37 in the 3rd qtr

GTW said...
This is odd... The ball position on that 4th down looked to be between the 20 & 21 yard lines. EP Total to Go4it from the 20 was 1.79, BUT 1.93 from the 21. EP Total for FG Att were 1.8 and 1.73. Tough call, but the Go4it EP's were what I found interesting.
1:37 in the 3rd qtr

Brian said...
Told ya. /outcome bias
5:47 in the 3rd qtr

Brian said...
Go.
5:51 in the 3rd qtr

Jay said...
I can't say I've really paid much attention to drive times on other games this season, but the drives in this game all seem to be taking very little time off the clock.
8:13 in the 3rd qtr

Justin said...
this is one of the first times that i followed the live wp but is there typicaly this large of decrepencies between this wp and and the live in-game line? the falcons were +125 (.44) this past commercial break
8:49 in the 3rd qtr

Sunrise089 said...
Brian, Ok, that makes sense. I was watching the game plus my daughter and missed the spot on the return. Maybe I'm underestimating turnover chances. I wonder with Kaepernick's athleticism though if there wasn't something low risk and high upside they could run. .18wp seems to call for SOME level of aggressiveness.


Brian said...
Usually you'd be right, but SF had a terrible return to the 15. In yrs past they'd be typically starting at their 30. Toughr call.


Sunrise089 said...
Commentators - only down 10 and "bad things could happen" if you try to make something happen. Might as well just forfeit if you're that terrified of running plays. Not as bad as Denver but still odd to just give away 2 time outs.


Sunrise089 said...
Seemed like Atlanta took that time out way too early. SF could definitely try for a FG with that much time.
0:25 in the 2nd qtr

Justin said...
wow, and just like that Atlanta at +125!
1:55 in the 2nd qtr

Sunrise089 said...
I'm not a skilled enough football viewer to pick out much in terms of scheme, but the narrative was that last week the 49ers brought in option looks they had hidden in prior weeks. Is the focus on Davis this week the same thing, or do you think it's randomness?
1:55 in the 2nd qtr

Brian said...
V. Davis has some monster WPA right now.
1:55 in the 2nd qtr

Brian said...
Don't forget the vig.
2:22 in the 2nd qtr

Justin said...
last timeout, while this model had Atlanta with a .82wp, live betting had it around -180 (implied .64). Interesting, maybe an overeaction and too much thinking out Seattles comback last week?
6:17 in the 2nd qtr

Brian said...
Whitner was smart not to make a big hit and draw a flag for defenseless receiver.
6:32 in the 2nd qtr

Brian said...
WP knows ATL is already up a TD. I'd trust EP there though--much less uncertainty.
1:49 in the 1st qtr

Brian said...
Kaep looks shaky. The risk reward equation is changing for SF. They'll need some breaks.
3:08 in the 1st qtr

Ian S said...
Interesting. EP says go for it, WP says attempt the FG. It was always going to be a field goal attempt no matter what the numbers said.
3:12 in the 1st qtr

Jay said...
Stuck at work for the Falcons game AGAIN this week :( Looks like a rough start for the 9ers so far.
7:01 in the 1st qtr

Brian said...
How important is smart play from your safeties? Very.
11:24 in the 1st qtr





2013012001