• Tony Bonk

Why is Buffalo 53% Win Probability?

As you go through the week and prepare your picks, fantasy teams, and wagers. You will likely find something that stands out, something that completely disagrees with your take that you must dive in further.

That happened to me this week, when looking through and their NFL Elo ranking and game projections. It had Buffalo with a 53% win probability over the LA Chargers. I think Buffalo is a bad team, so over 50% chance surprised me and prompted the question:


Why does have Buffalo as 53% win probability?

Answer (538) uses a power rank model to compare teams, called Elo ratings, which I think works great over long periods, but doesn't include significant player / coaching changes or key match-ups. In our Buffalo example, it won't account for going from Tyrod Taylor to Nathan Peterman or Josh Allen, a switch that takes their most important position from average to unproven (or worse).

For Elo ratings, each team has a rating and that increases or decreases based on game outcome and the Elo rating of their opponent, just like if you created your own power rankings and moved teams up and down based on each week's performance. The more games (more data points), then the more confident you are with trusting your rating, just like in Elo.

To account for offseason changes (player, coaching, schemes, etc.), 538 does adjust every team to be closer to the mean during the offseason (moving everyone one-third of the way closer to the mean). That is helpful and great way to bake in a regression to the mean into each team but does not account for major changes.

Elo will need a few weeks to adjust for teams that have undergone major offseason changes, so be aware of injury and key matchups. For example, does a heavy favorite have a poor secondary and run game (cue the Lions; although not how many times they will be a favorite the rest of the year), making it unlikely to hold off a good passing attack or be able to run out the clock? Or regarding players, was there a major injury, such as is Rodgers or Kiezer playing for GB (that one is obvious...)?

Players and matchups are quiet hard to account for in a mathematical model based on head-to-head results but the Elo system gives a solid baseline, especially over the long-term. Early in the season, it is most suitable for teams that have remanned relatively unchanged in key areas, such as the Patriots for example (head coach, QB, and scheme all the same; only a change at D-coordinator), and less suitable for teams with a major change, such as Arizona (new coaching staff, new QB, new defensive scheme). For teams like Arizona, their Elo will likely take a few weeks to settle into a more trustworthy state.

So, back to Buffalo's 53%.

We only have one game in the books, so we don't want to overreact to one poor performance. Regardless, due to the QB change, poor O-line play, and lack of any obvious match-up advantage, I think it should be lower, more like 30-35%. That aligns with a +6.5 line (as of writing: LAC @ Buf +7.5). Buffalo does have a couple things going for them, Bosa is expected not to play, it is Buffalo's home opener, and most teams don't get blown out two weeks in a row. I expect Buffalo to keep it close in the first half but LAC to pull ahead in the second, with their offensive talent being too much for Buffalo and a rookie QB in Allen to make a few mistakes that add up.

For More Information

Check out, it is fun reviewing all their projections and ratings, especially now that you have insight into how they are creating. They have a good page discussing how their projections work and it is worth a read:

Elo ratings were originally developed for rating chess players, by Arpad Elo, a Hungarian-American physics professor. Back in the 1930s, Professor Elo taught physics at Marquette University in Milwaukee and was a chess master. Fun fact, at that time, Milwaukee was one of the nation's leading chess cities and Professor Elo won the Wisconsin state championship eight times.


©2020 by

This site does not endorse, recommend, or support illegal betting or gambling under any circumstances. All information on this site is for entertainment purposes only and is not intended to be used in any direct or indirect violation of local, state, federal or international law(s).

Please be smart, abide by the laws of your area. Don't do illegal things.