JuiceCrewAllStar
Whole Milk Drinker
- Joined
- Feb 4, 2014
- Messages
- 37,425
- Reaction score
- 10,202
what are the inputs for post-game win expectancy?
Yeah I mean Wake is more likely overall to end the season 9-3 than 11-1. We have the BC game and then five games where the spread will be single digits (Wake probably favored in all five at this point).
Bill Connelly's version includes a lot of the same underlying "key predictive stats" that go into SP+: success rate, turnover margin/luck, explosive play rate, starting field position, some version of "finishing drive" rate (aka number of drives converted into points v. what you'd "expect" from where the team was). I'm sure there are a couple more I'm missing.what are the inputs for post-game win expectancy?
Yeah I mean that's just basic win expectancy. Converting a rough estimate of projected lines into implied odds:How dare you tell people this.
I’ll be really disappointed if we lose to State without Leary. I know it’s in Raleigh and they probably (?) will get a little better offensively over the next few weeks but we have way more firepower.State: 51% (this is probably a little low - I'd have it 55%-60% with Leary out)
they’ve been progressively plummeting since the season startedPlummeting seems like a strong word for losing on the road to a team that’s on their level and losing an overrated QB. I think State beats VT and it will still be a tough game.
You’ve got it. I did the same breakdown/analysis on the season-long thread yesterday and the takeaways were that a) we will likely be favored in all of our remaining games; and b) the odds of going 6-0 and not dropping one or more games is like 9-10%.Yeah I mean that's just basic win expectancy. Converting a rough estimate of projected lines into implied odds:
Boston College: 98% (rounded down from Boyd's bets which has >20 point spread at 100%)
Louisville: 62%
State: 51% (this is probably a little low - I'd have it 55%-60% with Leary out)
UNC: 73%
Cuse: 66%
Duke: 75%
So expected record based on these would be 9.25-2.75 (4.25 expected wins of 6 games)
Yep, but > 50% for the Clemson loss. We were also below 50% for Liberty which...yeah.Bill Connelly also had WF's postgame win expectancy at <50% after the win at FSU
This is interesting to me, because those stats from second half FSU game go on to affect the analysis of Wake for determining future game spreads and are obviously incorporated into statistical models meant to rank teams/project games. Considering that half is 1/12 of our season so far, it seems like that would be a significant thing.Yep, but > 50% for the Clemson loss. We were also below 50% for Liberty which...yeah.
The FSU result shows the issues that you have with some analytics in football specifically given smaller sample size in one game. That game plays out completely differently if the game was closer in the first half because Wake both stopped doing what was working so well in the first half (airing it out all over the place) to stretch the clock out while also gaining fewer yards per carry driving the sucess rate down.
Obviously you could write a whole book about this, but a couple thoughts.This is interesting to me, because those stats from second half FSU game go on to affect the analysis of Wake for determining future game spreads and are obviously incorporated into statistical models meant to rank teams/project games. Considering that half is 1/12 of our season so far, it seems like that would be a significant thing.
Taking into account the fact that Clawson goes super-conservative on offense when up 2-3 scores more often than most coaches, I wonder if that's a reason that analytics often have pessimistic views of Wake Forest even though we win a lot of those types of games.
they’ve been progressively plummeting since the season started
1. Yeah, I assumed there was a garbage time cutoff, but didn't know what it was. There's certainly more nuance to playcalling than a black and white cutoff like that, but I understand that it's only possible to capture so much of that nuance in a realistic to build model.Obviously you could write a whole book about this, but a couple thoughts.
1. One argument is that these types of decisions somewhat even out over the course of the season with teams. This is one reason there is an SP+ garbage time cutoff for included stats (over 34 points second quarter, over 28 points in the third quarter, over 22 points in the fourth), however where the difference in style/playcalling/efficiency is from a fast-paced balls to the wall attack that Wake normally runs and a milk the clock, grind the ball out on the ground this may be worth a point or over multiple games. Another aspect here is that even though Wake did go conservative, it doesn't change that this still decreased our success rate and FSU's offensive success rate improved over the course of the game - aka staying "on pace" on first, second, and third down to convert first downs.
2. I'd also argue that the better metrics out there have at least some baked in preseason projections where you almost have to use recruiting rankings to some degree as these are a pretty good proxy overall for talent/expectations. I think since Wake does have a pretty solid development over at least a year or two that these rankings disproportionately impact Wake's outlook. I don't really know what a solution is for this because I think you need some starting point. I also don't know if this is a feature or a bug.
3. I think that even non-metrics based analysis (eye test, just looking at final score, etc.) includes drawbacks similar to "took foot off the pedal" that shows up in metrics too.
This is interesting to me, because those stats from second half FSU game go on to affect the analysis of Wake for determining future game spreads and are obviously incorporated into statistical models meant to rank teams/project games. Considering that half is 1/12 of our season so far, it seems like that would be a significant thing.
Taking into account the fact that Clawson goes super-conservative on offense when up 2-3 scores more often than most coaches, I wonder if that's a reason that analytics often have pessimistic views of Wake Forest even though we win a lot of those types of games.
feels like the two losses to UNC the last couple years followed this pattern and were, obviously, not winsTaking into account the fact that Clawson goes super-conservative on offense when up 2-3 scores more often than most coaches, I wonder if that's a reason that analytics often have pessimistic views of Wake Forest even though we win a lot of those types of games.
Well I didn't say we win all of them, just that we win a lot of them. Although I understand fans hating the strategy, because our lead usually gets smaller, and almost never bigger, while we do it.feels like the two losses to UNC the last couple years followed this pattern and were, obviously, not wins