Scrapping the system: Another week, another prediction method
Brian Beaudry
Thursday June 23, 2011
After only one week, the four-tier system is no more, thanks in part to Utah’s huge win (and a couple of huge upsets by the system’s standard) to regain its ranking in points per possession exchange (PPPE) by leaping an entire 0.229 PPPE in the standings.
That left fifth-place Chicago about as close to 11th-place Dallas (0.146) as they were to fourth-place Utah (0.144), Utah further than that from third-place Orlando, Orlando more than three times that from Arizona, and Arizona roughly three times that from Jacksonville. That’s too many tiers, without even getting into the teams in the negative rankings.
Standings after Week 15:
1 | JAX | 1.386 | 7 | GEO | 0.17 | 13 | PHI | -0.489 |
2 | ARI | 0.959 | 8 | CLE | 0.153 | 14 | MIL | -0.516 |
3 | ORL | 0.515 | 9 | SPO | 0.125 | 15 | TB | -0.535 |
4 | UTA | 0.329 | 10 | TUL | 0.072 | 16 | IOW | -0.697 |
5 | CHI | 0.185 | 11 | DAL | 0.039 | 17 | KC | -0.784 |
6 | SJS | 0.175 | 12 | PIT | -0.317 | 18 | NO | -1.102 |
Since those tiers didn’t work out, I tried messing around with standard deviations, finding that Tampa Bay’s win over Orlando had dragged Orlando back into the 1 standard deviation range from the mean, and brought Tampa Bay themselves within that range this week.
With a standard deviation of 0.621, the only teams not included in the pack in the middle are Jacksonville and Arizona at the top (Tier 1 still exists!) and Iowa, Kansas City, and New Orleans at the bottom, despite Kansas City’s big win over Cleveland this weekend, which vaulted them up 0.162 PPPE.
There isn’t really a whole lot more to say about this week’s performances, aside from being able to add San Jose’s performance against Arizona to the list of teams that lost games despite scoring over 5 points per possession (PPP). However, seven teams won this weekend after scoring 5+ PPP, leaving the percentage of winners at over 85 percent once that number is achieved.
Since this season has been largely about exploring what I might be able to do with the numbers, all these numbers aren’t based on the most efficient means of sorting data – I have no real way of going back through and seeing what numbers have changed week to week throughout the season, short of deleting a whole lot of information and undoing the damage. However, after three weeks of posting some of the data, I should be able to throw in a quick graph next week to show how the teams are bouncing around in the PPPE rankings. The week after that, I can start building some offensive/defensive splits to show which unit is responsible for the fluctuations.
I’m currently working on figuring out what I want to include in analysis starting in Week 1 for the 2012 season. If you’d like to know something about events, possessions, and see what their relationship, if any, is to wins and losses, please email me by using the link at the top of the article.
The new non-tier system makes this week’s picks far more difficult after a 6-3 (2-1 in upset picks) Week 15, so I’ll adopt the standard deviation system:
Tier 1: (>2 StdDev above everyone else): Jacksonville
Tier 2: (1-2 StdDev): Arizona
Tier 3: (0-1 StdDev): Orlando, Utah, Chicago, San Jose, Georgia, Cleveland, Spokane, Tulsa, Dallas
Tier 4: (0-(-1) StdDev): Pittsburgh, Philadelphia, Milwaukee, Tampa Bay
Tier 5: ((-1)-(-2) StdDev): Iowa, Kansas City, New Orleans
As with last week, let’s say that two levels above another team makes you untouchable to them. Numbers in parentheses are the difference between the home team and away team in PPPE; a positive number means they’re better, a negative number means they’re worse.)
Tampa Bay at Jacksonville (+1.921 PPPE): Tampa Bay, I like you guys and you’re coming on strong at the right time of the season to potentially make a run at that last seed in the American Conference…but c’mon, it’s the Sharks. They are murderous this year. I can’t pick against them.
Kansas City at Utah (+1.113 PPPE): Utah’s two tiers above Kansas City, is at home, and the offense seems to have found itself again after looking lost without Aaron Boone for the past few weeks. Utah takes it.
San Jose at Cleveland (-0.022 PPPE): San Jose was this week’s best loser, gaining 0.105 PPPE after hanging in tough against a highly-ranked team on both sides of the ball in Arizona. They can’t keep losing forever, can they? This is their fifth game in the last six on the road after a cushy home schedule to start the season off…I’ll go with Cleveland here. I just can’t back a 1-5 road team on the third leg of a three-game road trip against a team with a similar ranking.
Arizona at Philadelphia(-1.448 PPPE): Arizona is just a far, far better team than Philly. Arizona.
Milwaukee at Pittsburgh (+0.199 PPPE): Pittsburgh scores surprisingly low in these rankings, though I guess it’s less of a surprise after Utah beat them senseless last week. Milwaukee, however, is bad enough that getting beaten by 15 by Jacksonville raised their score (because Jacksonville is so good on both sides of the ball). I’ll take Pittsburgh and be happy they’re at home for the game. Offer void if Bernard Morris does not play for the Power.
Georgia at Orlando (+.345 PPPE): Orlando is better, at home, and pissed off after losing to their rivals last week on some interesting plays and calls down the stretch in Tampa Bay. I’ll take the Predators.
Spokane at New Orleans (-1.227 PPPE): New Orleans continues to be awful (nearly an automatic loss pick in the same way Jacksonville is an automatic win pick), and Spokane should have enough, even with the potential loss of half of their offense to the UFL.
Dallas at Tulsa (+.033 PPPE): Yeah, Dallas has a better record, but don’t call this an upset, because Tulsa is the better team by an admittedly pretty insignificant margin. And they’re at home. You have no idea how much I want Dallas to win this just to complete their pattern for the season (all their road games have been back-to-backs: L, L; W,W; L,L; W, (W)). That would be cool. But I’m still going to take Tulsa.
Iowa at Chicago (+.882 PPPE): Chicago loves splitting games against opponents (1-1 vs Tulsa and Kansas City this year; 1-1 vs. Milwaukee (regular season), Iowa, and Dallas last year). They won the first matchup against the Barnstormers on the road, just as they did last year. I need an upset pick, and I’ll give Iowa a chance to redeem themselves after failing me last week against Dallas.
Random Fact of the Week: If you’re going to make a road trip, you might as well hope you have them in back-to-back games. While AFL teams are only 47-72 (.395) on the road this season, they’re 16-15 (.516) in the second game of back-to-back road games. 14 of the league’s 18 teams have won the second half of a back-to-back. Chicago earns special unique points for winning both such games after losing the first of the two games, and winning both such games by the exact same score, 50-49.
Don’t stay out on the road too long, though; teams that have had three consecutive road games are 1-2 in those games, and Philadelphia, who had to play its first four games on the road, lost the league’s only fourth game of a back-to-back-to-back-to-back road stretch.
Brian Beaudry has been both an intern and a Director of Communications within the AFL and occasionally provides analysis of the league and the Portland Thunder as an Oregon resident. He maintains a blog, Wrong Way Sports, and you can reach him with research or graphic design suggestions on Twitter at @BeaudryPDX.