NFL Week 6 NV: feels like I'm winning... nothing at all

The Something Awful Forums > Discussion > Sports Argument Stadium > The Football Funhouse > NFL Week 6 NV: feels like I'm winning... nothing at all

«‹›16 »

fsif: Jul 18, 2003

The problem with DVOA is that it's completely opaque and that even a decade in, they're still making things up as they go along.

quote:

Taking a big hit in the ratings and also in the rankings are the Detroit Lions and the Buffalo Bills. Detroit falls from 11th to 12th with 30% adjustments and then down to 17th with 100% adjustments. Buffalo drops from ninth to 15th with both sets of adjustments. I will admit that those rankings feel a bit more in line with how these teams have played this year, suggesting that maybe we do need to incorporate opponent adjustments even after only three weeks of the season.

# ? Oct 8, 2019 13:26

Adbot: ADBOT LOVES YOU

# ? Apr 29, 2024 19:43

Vengarr: Jun 17, 2010; Smashed before noon

fsif posted:

The problem with DVOA is that it's completely opaque and that even a decade in, they're still making things up as they go along.

That was from an article examining whether starting the season with opponent adjustments is a good idea, despite the small sample size. They don�t normally turn adjustments �on� until Week 4.

# ? Oct 8, 2019 13:33

Nosre: Apr 16, 2002

Comparing week 5 teams to other historical week 5 teams is one thing, but this is more wild:

# ? Oct 8, 2019 14:08

kalensc: Sep 10, 2003; Only Trust Your Respirator, kupo!
Art/Quote by: Rubby

I'm curious to see what a graph of "Team DVOA by week" would look like. Anyone know if FO provides a full-season data summary somewhere? Or would it need to be manually scraped from their weekly posts.

# ? Oct 8, 2019 14:14

Vengarr: Jun 17, 2010; Smashed before noon

kalensc posted:

I'm curious to see what a graph of "Team DVOA by week" would look like. Anyone know if FO provides a full-season data summary somewhere? Or would it need to be manually scraped from their weekly posts.

I know they include a DVOA-by-week graph with each chapter of the almanac. Dunno if it�s on the website.

# ? Oct 8, 2019 14:17

kalensc: Sep 10, 2003; Only Trust Your Respirator, kupo!
Art/Quote by: Rubby

Nosre posted:

Comparing week 5 teams to other historical week 5 teams is one thing, but this is more wild:

Well, purely going by math:

- N.E is undefeated with only 1 decent divisional rival, and they beat them on the road;

- Going by stats or the ol' eye test, N.E is certainly stronger than Buffalo overall;

- Thus N.E winning the division has an extremely high probability at this juncture;

- The expected wins versus their weak schedule gives them a very high bye probability;

- K.C's in-conference loss hurts their projections for a top seed (although their match vs N.E will settle that potential 2-way tiebreaker);

- So if you put that all together in the magic opaque number machine (very debatable methodology) you can reasonably arrive at:

P(N.E in playoffs) > 99.7%
P(N.E wins AFCE) =~ 95.0%
P(N.E earns bye) =~ 90.7%
P(N.E wins divisional playoff) =~ 78.7%
P(N.E wins conference title) =~ 57.7%
P(N.E wins title) =~ 37.3%

Note that conditional probabilities are important to consider for those playoff matches. That is, the probability that B occurs if we assume A occurs. This would give a better representation of what FO is currently projecting for an individual Sunday in January.

P(B|A) = P(A & B occur) / P(A occurs)

P(N.E wins DivPO | N.E has bye) =~ 0.787/0.907 =~ 0.868*

(* note that this specific calc isn't quite right, as we would need to cut out the (N.E wins DivPO without having a bye) slice, but I am about to go into a meeting and it's close enough° given how dominant their Bye% is)

((° note that I loathe that phrasing and will probably edit the true calc in later cause mathematical minds can't. let. poo poo. go.))

So, right now, IF N.E earns a bye and thus hosts a WC game winner such as Buffalo, Baltimore, Houston, Oakland, Indy, etc, THEN FO is projecting an N.E win approx 7 out of 8 times.

P(N.E wins ConfPO | (P(N.E wins DivPO | N.E has bye)) =~ 0.577/0.868 =~ 0.665

So, IF N.E earns a bye AND then advances to the AFC title match, THEN FO is projecting an N.E win approx 2 out of 3 times.

P(N.E wins SB | (P(N.E wins ConfPO | (P(N.E wins DivPO | N.E has bye))) =~ 0.373/0.665 =~ 0.561

So, if N.E earns a bye AND advances to AFC title game AND wins AFC title, THEN FO is projecting an N.E win approx 5 out of 9 times.

Those playoff projections aren't that unreasonable, and the SB game itself is close to a coin-flip. What gives them such a massive lead in the SB Win% column is the projection that they have a 78%~ shot at #1 Seed and 13% shot at #2 Seed (thus the 90.7% bye). For comparison, K.C has the second largest Bye% at 46.7%, and thus FO projections account for K.C playing in the Wild Card round quite often. That extra game, plus a win sending them on the road in the divisional round, significantly lowers their SB Win%.

TL;DR: Read the whole post, cause math is fun. Also, gently caress the Pats & gently caress the NFL Corporation in general. Go Players Go.

kalensc fucked around with this message at 14:57 on Oct 8, 2019

# ? Oct 8, 2019 14:52

Hot Diggity!: Apr 3, 2010; SKELITON_BRINGING_U_ON.GIF

Does DVOA account for the Patriots being cheaters and frauds

# ? Oct 8, 2019 15:09

Randaconda: Jul 3, 2014; by Jeffrey of YOSPOS

can you do it again in Steiner math

# ? Oct 8, 2019 15:09

Rooster Brooster: Mar 30, 2001; Maybe it doesn't really matter anymore.

DVOA includes two great ideas you don't get in "normal" stats - success rate and opponent adjustments.

Success rate being the number of times a play is successful, something like get 40% of needed yards on 1st down, 80% on 2nd down, 100% on 3rd or 4th down. That's a neat idea, since it weeds out plays such as when a draw on 3rd and 25 gets a running back 15 yards; it wasn't really successful, so the running back doesn't get much credit for it.

Opponent adjustments are also a good idea, it's the formal way of saying "Yeah, but they've only played Miami, Washingon, and the Jets so far, wait until they play a real team!"

The problem with DVOA is it takes these two basic, good ideas and mushes them into an opaque mega-Excel document. They also do this stuff with "greatest team of all time!" every year to fill content, and it almost always winds up being due to small sample sizes that vanishes by week 8 or 9.

I wish they'd just post success rate percentages and opponent-adjusted yards and left it at that; maybe they do have that info in their paid site I dunno. I have written too much about DVOA

# ? Oct 8, 2019 15:34

Hunt11: Jul 24, 2013; Grimey Drawer

Hot Diggity! posted:

Does DVOA account for the Patriots being cheaters and frauds

If the Pats are frauds then everyone else in the league is much worse then that.

# ? Oct 8, 2019 15:38

Dutchy: Jul 8, 2010

Amy Pole Her posted:

Oh so it’s San Fran who gets built up before their first round exit

Extremely fine with this

# ? Oct 8, 2019 15:44

wyoak: Feb 14, 2005; a glass case of emotion; Fallen Rib

Are the niners actually good? I haven�t been able to see them play yet

# ? Oct 8, 2019 15:47

adaz: Mar 7, 2009