Player ratings

How it works

The ratings at a glance

• Effectiveness - how good he is when he's on the field: how much each play he's involved in helps his team score, compared with an average NFL player at his position.
• Usage - how much his team uses him: how many plays he's part of in a typical game.
• Overall - the two together: how many points per game he adds, compared with the kind of backup a team could put in his place. This is the main ranking.
• If he's out - how many points per game his team would lose if he missed a game, based on who would actually replace him.
• Confidence - how much to trust the Effectiveness rating yet. Early in a season it's Low for most players, because a few games isn't enough to judge anyone.
• This season so far - what he's actually done this year alone, with no caution applied. The most up-to-date picture, but a few games can make it jump around.

Words you'll see

Play

One snap of the ball, from the moment it's snapped until it's whistled dead. A team runs about 60 plays in a game.

Dropback

A play where the quarterback drops back to pass. It counts whether he throws, gets sacked (tackled before he can throw), or runs with the ball instead ("scrambles").

Carry

A running play where a player takes the ball and runs with it.

Target

A pass thrown to a player, whether he catches it or not.

Snap

A play a player is on the field for. Linemen and defenders are measured per snap, because they're on the field for plays whether or not the ball comes their way.

Expected points (EPA)

The main way these ratings measure value. From years of NFL games we know, for any situation (which down it is, how many yards to go, where the ball is on the field), how many points a team scores on average from there. Every play changes the situation, so every play changes that number. EPA (expected points added) is that change. Example: on 3rd down with 8 yards to go, a 10-yard catch keeps the drive alive and might be worth +1.5 points; an incompletion means the team probably has to punt, so it might be worth -1.5. Adding up a player's EPA over his plays shows how much he actually helps his team score, not just how many yards he gets.

League average

What an average NFL player at the same position does on the same kind of play. Effectiveness is measured against it: above zero is better than average, below zero is worse.

Replacement level / backup level

How well the backups at a position actually play - the kind of player a team could turn to if its starter got hurt. Overall measures how much better than that a player is.

Charted stats

Some stats - pressures on the quarterback, missed tackles, dropped passes, yards before and after contact - aren't in the official box score. They come from Pro Football Reference, a stats site whose staff watch every play and log these details.

Depth chart

Each team's list of who plays each position, in order: the starter, then the first backup, and so on.

The 0-99 scale

Every rating is on a 0-99 scale. The marks on it come from real players in past NFL seasons (): a 99 means as good as the very best player at his position was at the same point in one of those seasons, a 90 means about as good as the 5th-best, and so on down. The tables below show exactly what each mark means.

A player is always compared with past players at the same point of their season: after 2 games, with players after 2 games; after 10, with players after 10. That way a player who keeps playing at the same level keeps the same rating all season, instead of creeping up just because more games have been played. And because the marks come from past seasons, this isn't a ranking against this year's players - in a strong year, lots of players can rate high. The tables below show the scale at this point in the season.

Effectiveness: how good is he on each play?

Effectiveness is how much a player helps his team score on each play he's part of, measured in expected points (see "Words you'll see" above) and compared with an average player at his position doing the same kind of play (). Each kind of play is compared separately - for example, a quarterback's passes with other quarterbacks' passes and his runs with other quarterbacks' runs - so a player can't look good just because he gets easier kinds of plays.

Confidence: how much to trust Effectiveness yet

Judging a player from a few games is like judging a basketball shooter from a few shots: one long touchdown or one bad interception can make anyone look great or terrible. So until a player has been part of enough plays, his Effectiveness rating assumes he's about as good as a typical player with his job - a typical starting quarterback if he starts, a typical backup if he's a backup - and moves toward his own numbers with every play he's part of.

• Low - he hasn't been part of many plays yet, so his rating is still mostly the typical player's numbers.
• Medium - his own plays are starting to move his rating.
• High - he's been part of enough plays that his rating is mostly his own numbers.

How many plays that takes is different for each position. We worked it out from past seasons, by checking how many plays it took before a player's own numbers did a better job than the typical player's of predicting how he'd do for the rest of the season:

Some positions take more than a season. A running back's value per carry depends a lot on his blockers and the defense in front of him, and the numbers we have for cornerbacks and safeties miss most of what they do. So nearly every player at those positions stays at Low or Medium. That doesn't mean they're bad players - it means the numbers can't reliably tell them apart yet.

This season so far

Under each Effectiveness rating is what he's done this season alone, with none of the caution above: his own value per play in this year's games, on the same kind of 0-99 scale (compared with what past players' numbers looked like after the same number of games). It's the most up-to-date picture of how he's playing right now. The number of plays it's based on is shown next to it - the fewer the plays, the more it can jump around.

Usage: how much does his team use him?

Usage is how many plays he's part of in a typical game (). Next to it you'll see his role (for example "Lead back" or "Primary target") and his trend: whether his last two games show him being used more ("Rising role") or less ("Shrinking role") than earlier in the season - a backup taking over, or a starter losing his job. Games he left early because he got hurt don't count against him.

Overall: how many points does he add?

Overall answers: how many points per game does this player add, compared with a backup? It's how much better he is on each play than a backup-level player, multiplied by how many plays he's part of. A very good player who barely plays adds little; a good player who's part of every play adds a lot. It's the main ranking.

"How many plays" is measured as his share of his team's plays - for a quarterback, what share of his team's passing plays he's part of - applied to an average team. That way a quarterback on a team that runs the ball a lot isn't marked down just because his coach calls fewer passes, while a backup or a part-time player still counts for less than a full-time starter.

If he's out: how much would his team miss him?

We look up who is next in line on his team's depth chart (skipping anyone who's also injured) and compare the two: how much better he is per play than that player, times how many plays he's part of in a game. The result is the points per game his team loses. "No meaningful loss" means the player behind him has been about as good. For a backup, the comparison is with the player behind him, not the starter in front of him. This is the number TOPICAL's weekly game predictions use for injuries.

Against good, average and bad opponents

Open a player to see how he's done this season against good, average and bad opponents - defenses for offensive players, offenses for defenders. Opponents are sorted by how well they've played this season (the best 10, the middle 12, the worst 10), and each player is shown next to what an average player at his position did against the same opponents. This is extra context only: it doesn't change any rating.

How each number is worked out

Every stat mixes four things: this season (every game so far, with his last two games counting a little more or less), a small part of last season (helpful in the first few weeks, then it fades out), and a starting guess - a projection of his stats where one exists, otherwise what a typical player with his job does. How much each part counts wasn't chosen by hand: we replayed the seasons week by week and picked the mix that best predicted what each player actually did for the rest of the season. Open any player to see his exact mix.