Division odds, week 3

Here are the odds of winning the division after week 1, per Football Outsiders (click to enlarge):

The Saints started 0-3 for the second consecutive year and the third time under Sean Payton. According to Football Outsiders, the Saints have about a 10% chance of winning the division. Unfortunately, that might overstate the case…per FiveThirtyEight, only about 2% of teams that start the season 0–3 make the playoffs, and 0–3 teams finish with an average of about 4.9 wins. I don’t see any reason to believe the Saints will do much better.

The Falcons, on the other hand, ascend the Mountain of Probability thanks to their in-division win on Monday. Sigh.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

NFL-wide playoff odds, week 2

Both Football Outsiders and 538 produce playoff odds every week. They results are broadly similar, but sometimes it’s interesting to compare them. Here they both are in one plot (click to enlarge):

The farther a team is from the diagonal line, the more the two models disagree. The models still disagree over Denver and have a pretty significant disagreement about Tennessee, as well: Football Outsiders has them at about 36% likely to make the playoffs vs. about 11% likely in the 538 model. This stuff will sort itself out over time, I suppose.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

Division odds, week 2

Here are the odds of winning the division after week 1, per Football Outsiders (click to enlarge):

The Saints started 0-2 for the third consecutive year and the fifth time under Sean Payton (well, one was under Aaron Kromer). They’ve yet to make the playoffs after starting 0–2. Needless to say, the game against Atlanta is important.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

Division odds, week 1

Here are the odds of winning the division after week 1, per Football Outsiders (click to enlarge):

This isn’t a great start, but I don’t think anyone reasonable expected the Saints to compete for the division. That said, losing the first game at home to an Oakland team that was highly beatable is a bit frustrating and significantly hinders the Saints’ chance of making the playoffs at all. The Saints’ last week 1 victory was in 2013…it’s hard when you’re starting in the hole each year.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

NFL-wide playoff odds, week 1

Both Football Outsiders and 538 produce playoff odds every week. They results are broadly similar, but sometimes it’s interesting to compare them. Here they both are in one plot (click to enlarge):

The farther a team is from the diagonal line, the more the two models disagree. The strongest disagreement right now is over Denver, where 538 has them about 74% likely to make the playoffs and Football Outsiders has them about 37% likely. Although Football Outsiders’ methodology isn’t 100% clear, I think this might be the difference: 538’s ELO system doesn’t know that Peyton Manning has retired, whereas Football Outsiders’ DVOA takes this into account. The same is likely the case for Carolina and Josh Norman. Another likely factor: Football Outsiders’ model bakes in some regression for both the Carolina and Denver defenses, whereas ELO doesn’t attempt to make those sorts of predictions.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

Bring out your dead

In light of the news that the Saints have cut CJ Spiller, here’s a graph of NFL teams’ dead space as a percent of total cap, 2011–2016. Data from spotrac.

The other outliers in terms of dead space (Oakland & Jacksonville in 2013, Indianapolis in 2012) were part of salary cap purges and were sharp peaks before big dropoffs. The Saints should have a significant drop-off in dead space next year, though 2018 might be bad if Brees leaves.

This is how things get ugly.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

10 years of Brees to Colston

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

Week 1 in review: Net yardage, Glicko, etc.

It’s not shocking that the Saints lost: they were playing a better team on the road. But it would have been a nice win to have, especially because the NFC South promises to be a competitive, if not stellar, division.

Here are the Glicko rankings after week 1 of the NFL season. I seeded the 2015 rankings with mean-regressed 2014 rankings. In other words, these rankings take last year’s ranking into account, at least a little bit. Don’t forget the rankings are dumb in that they only take a few things into account: wins, losses, consistency, and opponent strength. So if Aaron Rodgers (or, more pertinently, Jordy Nelson) got hurt last week, the Packers’ Glicko rating would not change until they started losing games.

In the chart, teams with green dots are above league average, teams with grey dots are around league average, and teams with red dots are below league average, at least according to the Glicko ratings. Also note that there’s a bug in the code I used to generate this chart, so the Bengals should be green and the Jets should be grey. Sigh.

Week 1 2015 Glicko ratings

Glicko has the Saints right at league-average, in the same ballpark as the Falcons and Panthers and notably better than the Bucs. Sounds about right.

Other graphs

I’ve been playing around with a few other graphs in an attempt to analyze last week’s game. They’re all variations on the theme of how to visualize possessions and field position within a game. I’m not sure how useful they are, but here goes.

This first one I’m calling a Net Yardage chart. It shows the cumulative net yardage for the Saints: a positive net yardage means the Saints have gained more yards than their opponents, a negative net yardage means they’ve been out-gained. These graphs count punts, punt returns, and penalties, but don’t count kickoffs and kickoff returns. Check it out:

Week 1 Net Yardage Chart

This chart shows that Arizona was fairly dominant in the first half: at no point did the Saints have more total yards than the Cardinals. That flipped quickly in the second half when Brees connected with Snead for a 63-yard gain. In all, the second half was much more even in terms of net yardage. The difference: the Cardinals scored touchdowns, the Saints scored field goals.

Here’s another chart I’ve been working on. I’m calling it a Ball Travel chart. It graphs the field position of the ball over time. Each individual line follows the ball as it moves across the field towards the Arizona end zone (marked 100 on the graph) or the New Orleans end zone (marked 0 on the graph). Once a team scores, the chart starts over. This might be more clever than it is informative, but I kind of like it. You’ll probably want to click to enlarge it:

Week 1 Ball Travel Chart

Finally, here’s one I’m calling a Drive Chart. It’s essentially the same as the Ball Travel Chart, but it displays each drive individually, color-coded by team. The dashed grey lines represent punts. Again, maybe more clever than useful, but I’m not sure.

Week 1 Drive Chart

I’m still working on automating these, so I probably won’t produce them every week. But it’s fun to screw around.

Exclusive Week 2 preview!

The Saints are, like, way better than the Bucs. They’d better win.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

Updated Glicko chart

Here’s a Glicko chart for the 2014 NFL season. Tech details and optimizations still to come, but for this version I calculated 2013 Glicko ratings and used them as a starting point for the 2014 season. I also included error bars, which represent a 95% confidence interval for the rating. Essentially, if the bar from one team overlaps with the point from another team, then their ratings are statistically indistinguishable from each other. The dotted gray line is league-average. Check it out:

2014 NFL Glicko ratings

What I like about this is that it makes it easy to lump the league into three different categories: teams that are clearly above average (everyone from the Seahawks->Packers), teams that are about average (Chargers->Rams), and teams that are clearly below average (Vikings->Bucs). More in-depth analysis might help you distinguish the teams on a more fine-grain scale, but this rough cut is both a good start and a good reminder that we don’t know much after a 16-game season.

Once I get the final version refined (hopefully before the season starts), I’ll roll out a post explaining it in detail. Until then, consider this a work-in-progress.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.

Glicko Ratings for NFL Teams

I’ve been reading a bit about different rating systems (such as the ELO Ratings used by FiveThirtyEight) and came across an intriguing rating system developed by BU professor Mark Glickman. They’re called the Glicko Ratings. One nice thing about Glicko ratings is that they allow you to calculate deviations, which is a rough measure of uncertainty in the rating.

So I’m going to try to work with Glicko ratings a bit this year, updated weekly in the newsletter and then the site. More details to come (including some technical info), but for now, here is a first cut at the weekly Glicko ratings for the NFC South in 2014. Click to enlarge.

initial glicko ratings

There are many things wrong with this initial analysis. For example, they assume that all the teams are equally good going into the season, which isn’t exactly true. FiveThirtyEight solves this problem by using prior-season ratings (accounting for some regression toward the mean) as an input for week one. I’ll do that eventually, which should make the ratings a little better. There are other optimizations to make, too, in time.

That said, this first cut looks pretty good to me.

---

Like what you see here? Please consider supporting me on Patreon. Your support helps keep me blogging plus, you know, kids are expensive.