Football Outsiders content published by ESPN

ESPN Upset Watch: 49ers-Rams

Photo: USA Today Sports Images

I'd like to thank Vegas for making the No. 1 team in DVOA three-point underdogs this week, making our Upset Watch choice fairly easy. The Rams have more rest and the 49ers have more injuries, but it's still easy to see a path to a San Francisco victory. Cover Watch looks at the return of Sam Darnold against the Cowboys.

View Full Article


5 comments, Last at 15 Oct 2019, 9:04am

1 Diametrically opposed to…

Diametrically opposed to what PFF is saying, they love the Rams. So far, however, whenever I’m noticing a difference in what analytic website says, DVOA usually does better than most other metrics

3 Their ELO model gives more…

Their ELO model gives more credence to past performance (even in previous years) so it hasn't believed in the 49ers' start as much as believing in the Rams history. However, this year there is a bunch of subjective (and objective via FO DVOA) data that can discredit this approach, like Garappollo being injured last year, the Rams o-line being much more inefficient this year, the 49ers pass rush being vastly improved, and Goff looking like he's a bit lost out there.

PFF also does a bad job of considering how teams match up vs one another, and they tend to overrate QB play when making single game predictions. The aforementioned 49ers pass rush vs the Rams o-line is one I think is likely to prove decisive in this game. The Rams offense was so deadly last year because the o-line was otherworldly, but even by PFF's numbers, their o-line this year has been the worst in the league so far (behind the Lolphins!!). With that hindrance, Goff looks pedestrian and McVay's scheme doesn't work as fluidly. On the flip side, the 49ers have both their tackles out, but the Rams edge rush isn't as potent so this advantage isn't as drastic.

Another thing that people are using as justification to bet the Rams is that McVay has had extra time to prepare vs the 49ers short week. However, while I do think this is a slight advantage, it probably isn't that big of a deal vs Kyle Shanahan compared to your regular replacement-level coach, and the 49ers could have prepped some for this game during the bye week. Additionally, it's a division game so the teams are more familiar with each other, making the gameplanning time aspect less important.

That being said, I do like listening to the PFF Forecast podcast when it comes to betting. It's free so you don't need to pay anything to get access to some their data and most of the time their picks are quite good. Usually FO and PFF are mostly in alignment, but this Rams-49ers game they are at the opposite pole, which I find interesting and hopefully I explained why that is and why I tend to side with the FO data here.

4 PFF uses an ELO model? I…

PFF uses an ELO model? I didn't realize that. I knew 538 did, and I enjoy looking at their model as a fun novelty. It's generally accurate a lot of the time, moreso than "common wisdom", but I would never ever take it seriously.

ELO was never meant to be used in this way. It works well for rating chess players though, despite not being meant for that either.

5 yeah im guessing they chose…

yeah im guessing they chose to use an ELO
model because all their data is based on individual player performance, which is limited in evaluating coaching, scheme, and what the player is asked to do and therefore pretty noisy. Because of that noise generated, they tried to use an ELO model to balance out how much fluctuation there is when your data is based on single player performances. Whereas FO uses an efficiency model with opponent adjustments (PFF also doesn’t use this) to pinpoint what a team is doing well and what it isn’t, which makes their single-season data more reliable, IMO.

There is some value with PFF and ELO simplifying things so you don’t make drastic mistakes, like you said, but it can only go so far. Ironically, I find a lot of betting markets tend to mirror ELO ratings very closely, so clearly Vegas just not being stupid and overreacting is enough of an edge to beat the public. Regardless, all of those tools require contextual and situational interpretation (the “eye” test) as well. I find FO the most useful for betting purposes, though.