Methodology and Stat Glossary
How percentiles are calculated, who qualifies, where the data comes from, and what every stat on CFB Savant means.
CFB Savant is modeled on Baseball Savant’s approach: a leaderboard is only useful if you can see what each number means, how it was calculated, and how it compares with everyone else. This page is the site’s glossary and methodology in one place. It explains the percentile system, the qualification rules, the data source, and every metric on the site — what it measures, how it is calculated, how to read its percentile, and an example.
On this page
How percentiles work
A percentile answers one question: within this season, against this qualifying pool, how unusual was this number? A 99th-percentile passer rating means only about 1% of qualified quarterbacks in that season were better. A 50th percentile means middle of the pack. A 10th percentile means most qualifiers were better.
The formula is the same everywhere on the site. For a given metric, count how many qualifiers in the pool were worse than this player or team, divide by the pool size minus one, multiply by 100, and round to the nearest whole number:
Three rules matter for reading percentiles correctly.
- Percentiles are within-season only. A 2026 quarterback is compared with 2026 qualifiers, not with 2019 qualifiers. The Compare tool respects this: each player is measured against his own season’s pool, so cross-season comparisons compare percentiles, not raw totals from different eras.
- Percentiles use the full pool, even when you filter. Filtering the leaderboard to one conference does not re-rank players against only that conference. It shows conference players with their national percentiles intact, so you can see how a conference stacks up nationally.
- Some percentiles are inverted. For stats where fewer is better — interceptions, interception percentage, points allowed, yards allowed, and opponent third-down percentage — fewer earns a higher percentile. A 90th-percentile interception percentage means excellent ball security, not many interceptions. The glossary entries below flag every inverted metric.
Percentile bars on player and team cards use a consistent 0-to-100 scale. The radar chart plots the same percentiles on multiple axes at once, so you can see a profile — efficient but low-volume, high-volume but turnover-prone, explosive but inconsistent — in one glance.
Who qualifies
Percentiles are only meaningful if the pool is restricted to players who actually played enough to be compared. CFB Savant applies a minimum workload rule per category. Players below the line are not ranked, even if their per-attempt averages look spectacular on a handful of throws or carries.
| Category | Qualification rule | Why |
|---|---|---|
| Quarterbacks | 14 pass attempts per team game | Ensures the quarterback was a real starter-level passer, not a backup who threw a few passes. |
| Rushing | 6.25 rush attempts per team game, quarterbacks excluded | Focuses the board on running backs and other non-quarterback rushers with a real workload. |
| Receiving | 1.875 receptions per team game | Includes regular contributors at any position while excluding one-catch outliers. |
| Team offense | All FBS teams | Every FBS team qualifies. No minimum beyond playing FBS games. |
| Team defense | All FBS teams | Every FBS team qualifies. No minimum beyond playing FBS games. |
Qualification uses team games played, not the player’s appearances, so early-season numbers stay comparable when teams have played different numbers of games. A quarterback on a team that has played 6 games needs 84 attempts to qualify (14 × 6). One on a team that has played 5 games needs 70.
Data source and refresh
Statistics are sourced from CollegeFootballData (collegefootballdata.com) via its season stats, team stats, and games feeds. The site covers the College Football Playoff era, 2014 to the present. Every season is regular season only, with no bowl games. Points for and against come from the games feed for completed regular-season games.
During the season, data is refreshed weekly. Each refresh refetches the current season, rebuilds the JSON files that power the leaderboards, regenerates the static leader pages, and redeploys the site. The Trending tab compares the new snapshot with the previous one to show who gained production since the last refresh.
In-progress seasons reflect only the games included in the current data snapshot. That means early-season percentiles move quickly: one big game can swing a percentile by 20 points in Week 2, while the same game barely moves it in Week 12. Treat early-season percentiles as a snapshot, not a final judgment.
Team colors and abbreviations come from the same source feed. Player positions come from the source feed. CFB Savant does not invent players, teams, or stats — if a player is not in the source data, he is not on the site.
What percentiles do not do
- They do not adjust for opponent strength. A quarterback who feasts on weak defenses and one who produces against top defenses are ranked on the same scale. Percentiles describe production, not schedule-adjusted quality.
- They do not adjust for pace, weather, or game script. A team that runs 90 plays a game will accumulate more yards than one that runs 60, even at the same efficiency. Per-play and per-game numbers help, but they do not fully control for context.
- They are descriptive, not predictive. A 99th-percentile rusher has been excellent so far. That does not guarantee he will stay there, or that his team will win its next game.
- They are not a single ranking. There is no overall “best player” score. The radar and the per-metric bars are the point: a player can be elite in one area and average in another, and both facts matter.
Quarterback metrics
Quarterback leaderboards rank qualified FBS quarterbacks. Rushing stats for quarterbacks are shown on the card for context, but quarterbacks are excluded from the main Rushing leaderboard. View the quarterback leaderboards.
Passer Rating Rating
A single-number summary of passing efficiency, using the NCAA passer rating formula. It rewards completions, yards, and touchdowns, and penalizes interceptions.
How it is calculated
Rating = (8.4 × Yards + 330 × Touchdowns + 100 × Completions − 200 × Interceptions) ÷ Attempts. All inputs are season totals from regular-season games.
How to read it on CFB Savant
Higher is better. Because the formula blends volume-free efficiency components, a quarterback can rank highly without leading in raw yards. Compare rating percentile with yards-per-attempt percentile to see whether efficiency and volume agree.
Example
A quarterback with 250 completions on 380 attempts for 3,400 yards, 30 touchdowns, and 5 interceptions has a rating near 178. That would sit near the top of most seasons, not because of one huge game, but because every component points the same way.
Yards Per Attempt YPA
Average yards gained on every pass attempt, including incompletions and interceptions. The cleanest single measure of downfield passing efficiency.
How it is calculated
YPA = Passing Yards ÷ Pass Attempts.
How to read it on CFB Savant
Higher is better. YPA punishes check-down-heavy passing that pads completion percentage without moving the ball. A high completion percentage with a low YPA usually means short throws.
Example
A quarterback averaging 10.5 YPA is pushing the ball downfield efficiently. One averaging 6.5 YPA is completing passes but not creating much yardage per throw.
Adjusted Yards Per Attempt AYA
Yards per attempt with a bonus for touchdowns and a penalty for interceptions. AYA answers the question YPA leaves open: were those yards actually valuable?
How it is calculated
AYA = (Passing Yards + 20 × Touchdowns − 45 × Interceptions) ÷ Pass Attempts. The 20 and 45 weights are the site’s fixed adjustment, applied identically to every quarterback in every season.
How to read it on CFB Savant
Higher is better. AYA will rank a quarterback with many touchdowns and few interceptions above a quarterback with the same YPA but more picks. When AYA percentile is well above YPA percentile, ball security and red-zone finishing are doing extra work.
Example
Two quarterbacks both average 9.0 YPA. One has 25 touchdowns and 3 interceptions, the other has 15 touchdowns and 12 interceptions. The first will have a much higher AYA, and should — his attempts produced more points and fewer giveaways.
Completion Percentage Comp%
Share of pass attempts completed. A measure of accuracy and throw selection, not of yardage.
How it is calculated
Completion % = 100 × Completions ÷ Attempts.
How to read it on CFB Savant
Higher is better, but read it with YPA. A high completion percentage on short throws is less valuable than a slightly lower one on deeper throws.
Example
A quarterback completing 70% of passes at 6.0 YPA and one completing 63% at 9.5 YPA are very different players. The second is creating far more yardage per attempt.
Yards Per Completion YPC
Average yards on passes that were actually caught. It isolates the depth of completed throws from accuracy.
How it is calculated
YPC = Passing Yards ÷ Completions.
How to read it on CFB Savant
Higher is better. A high YPC with a modest completion percentage suggests a downfield attack. A low YPC suggests a short, quick passing game.
Example
A quarterback with 3,000 yards on 200 completions averages 15.0 YPC. One with 3,000 yards on 280 completions averages 10.7 YPC. Same yardage, very different styles.
Passing Yards Per Game Yds/G
Passing yards averaged across the team’s games, not the quarterback’s appearances. This keeps per-game numbers comparable across players on teams that have played different numbers of games.
How it is calculated
Passing Yds/G = Passing Yards ÷ Team Games Played.
How to read it on CFB Savant
Higher is better. Per-game numbers are more comparable early in a season than raw totals, when some teams have played five games and others have played seven.
Example
A quarterback with 1,800 yards in 6 team games averages 300.0 Yds/G. One with 1,800 yards in 5 team games averages 360.0 Yds/G, a meaningfully higher pace.
Touchdown Percentage TD%
Share of pass attempts that ended in a passing touchdown.
How it is calculated
TD% = 100 × Passing Touchdowns ÷ Pass Attempts.
How to read it on CFB Savant
Higher is better. TD% rewards red-zone finishing and big-play passing. Read it with interception percentage: a high TD% with a high INT% is a boom-or-bust profile.
Example
A quarterback with 30 touchdowns on 350 attempts has a TD% of 8.6%. One with 15 touchdowns on 350 attempts is at 4.3%, half the scoring rate per throw.
Interceptions and Interception Percentage INT / INT%
Total interceptions and the share of attempts intercepted. The only quarterback metrics on CFB Savant where fewer is better.
How it is calculated
INT is the season total. INT% = 100 × Interceptions ÷ Pass Attempts.
How to read it on CFB Savant
For both, the percentile is inverted: fewer interceptions earns a higher percentile. A 90th-percentile INT% means the quarterback protected the ball better than about 90% of qualifiers, not that he threw more picks.
Example
A quarterback with 2 interceptions on 350 attempts (0.6%) will rank near the top in INT% percentile. One with 14 interceptions on 350 attempts (4.0%) will rank near the bottom, even if his yardage totals look similar.
Rushing metrics
Rushing leaderboards rank non-quarterback rushers who meet the workload rule. Quarterbacks are filtered out entirely, so the board compares running backs, receivers on jet sweeps, and other non-quarterback ball carriers with each other. View the rushing leaderboards.
Yards Per Carry YPC
Average yards on every rushing attempt. The core efficiency measure for non-quarterback rushers.
How it is calculated
YPC = Rushing Yards ÷ Rushing Attempts.
How to read it on CFB Savant
Higher is better. YPC separates efficient runners from high-volume runners. A back with 800 yards on 100 carries (8.0 YPC) was far more efficient than one with 1,000 yards on 220 carries (4.5 YPC).
Example
A back averaging 6.5 YPC is creating chunk yardage on a typical carry. One averaging 3.8 YPC is getting stopped near the line most of the time.
Rushing Yards Per Game Yds/G
Rushing yards averaged across the team’s games.
How it is calculated
Rushing Yds/G = Rushing Yards ÷ Team Games Played.
How to read it on CFB Savant
Higher is better. Per-game numbers stay comparable when teams have played different numbers of games.
Example
A back with 700 yards in 5 team games (140.0 Yds/G) is outpacing one with 800 yards in 7 team games (114.3 Yds/G), even though the second has more total yards.
Rushing Yards and Touchdowns Yds / TD
Season totals for rushing yards and rushing touchdowns. Volume measures that reward durability and workload.
How it is calculated
Totals are summed from regular-season games. Touchdowns are rushing touchdowns only.
How to read it on CFB Savant
Higher is better for both. Read totals with YPC: high totals with low YPC mean heavy workload; high totals with high YPC mean dominance.
Example
Two backs each have 900 rushing yards. One needed 130 carries, the other needed 210. The first was far more efficient; the second was leaned on more heavily.
Longest Run and Fumbles Long / FUM
Longest rushing gain of the season, and total fumbles lost or charged on rushing plays. Long shows big-play ability. Fumbles show ball-security cost.
How it is calculated
Long is the maximum single rushing gain in yards. Fumbles are the season total from the source feed.
How to read it on CFB Savant
For Long, higher is better. For Fumbles, fewer is better, and the percentile is inverted where shown. A back with a 70-yard long and zero fumbles has both explosion and security.
Example
A back with an 83-yard long run has at least one home-run play. A back with 5 fumbles has a recurring cost that offsets some of his yardage.
Receiving metrics
Receiving leaderboards rank pass catchers at any position who meet the reception workload rule. View the receiving leaderboards.
Yards Per Reception YPR
Average yards on every catch. It measures how much yardage a receiver creates per time he is actually targeted successfully.
How it is calculated
YPR = Receiving Yards ÷ Receptions.
How to read it on CFB Savant
Higher is better. Deep threats and yards-after-catch receivers rank high. Possession receivers who catch many short passes rank lower, even with big reception totals.
Example
A receiver with 900 yards on 35 catches (25.7 YPR) is a big-play receiver. One with 900 yards on 75 catches (12.0 YPR) is a high-volume possession receiver.
Receptions Per Game Rec/G
Receptions averaged across the team’s games. A workload and involvement measure.
How it is calculated
Receptions Per Game = Receptions ÷ Team Games Played.
How to read it on CFB Savant
Higher is better. A high Rec/G with a high YPR is a rare, dominant combination. A high Rec/G with a low YPR suggests a short-area target who is fed often.
Example
A receiver with 40 catches in 5 team games (8.0 Rec/G) is the focal point of his offense. One with 40 catches in 7 games (5.7 Rec/G) is involved, but less central.
Receiving Yards Per Game Yds/G
Receiving yards averaged across the team’s games.
How it is calculated
Receiving Yds/G = Receiving Yards ÷ Team Games Played.
How to read it on CFB Savant
Higher is better. Yds/G is the best single per-game summary of receiving production, because it blends volume and efficiency into one pace number.
Example
A receiver at 155 Yds/G is producing like a No. 1 option every week. One at 65 Yds/G is a solid contributor, not a centerpiece.
Team offense metrics
Team offense leaderboards rank all FBS teams on scoring and yardage. View the team offense leaderboards.
Points Per Game Pts/G
Average points scored per game, from completed regular-season games. The headline measure of team offense.
How it is calculated
Points Per Game = Total Points Scored ÷ Games Played. Points come from the games feed, not the stats feed.
How to read it on CFB Savant
Higher is better. Points per game is an outcome, not an efficiency measure. Read it with yards per play and third-down percentage to see whether the scoring is sustainable or driven by short fields and turnovers.
Example
A team at 49.6 Pts/G is scoring on most possessions. One at 21.0 Pts/G is struggling to finish drives, even if its yardage totals look respectable.
Total Yards Per Game and Yards Per Play Yds/G / Yds/Play
Total offense per game and per play. Yards per game measures volume; yards per play measures efficiency.
How it is calculated
Yards Per Game = Total Yards ÷ Games. Yards Per Play = Total Yards ÷ Total Plays (pass attempts + rush attempts + sacks where counted by source).
How to read it on CFB Savant
Higher is better for both. A team with high Yds/G but low Yds/Play is running many plays. A team with high Yds/Play is explosive on a per-snap basis.
Example
A team at 550 Yds/G and 8.2 Yds/Play is both productive and explosive. One at 400 Yds/G and 5.0 Yds/Play is grinding.
Pass and Rush Yards Per Game Pass Yds/G / Rush Yds/G
How a team’s yardage splits between passing and rushing, per game.
How it is calculated
Each is that category’s yards divided by games played.
How to read it on CFB Savant
Higher is better within each, but the split itself is descriptive, not a ranking. A balanced team and a pass-heavy team can both be excellent. The split tells you how they get there.
Example
A team at 334 Pass Yds/G and 220 Rush Yds/G is balanced with a passing lean. One at 120 Pass Yds/G and 280 Rush Yds/G is a run-first offense.
Third-Down Conversion Percentage 3rd %
Share of third-down attempts converted to a first down.
How it is calculated
Third-Down % = 100 × Third-Down Conversions ÷ Third-Down Attempts.
How to read it on CFB Savant
Higher is better. Third-down percentage measures drive extension. It correlates with scoring because it keeps possessions alive.
Example
A team converting 51% of third downs is sustaining drives. One converting 30% is punting often, even if it gains yards on first and second down.
Team defense metrics
Team defense leaderboards rank all FBS teams on points and yards allowed, plus havoc stats. Remember: for points allowed, yards allowed, and opponent third-down percentage, fewer is better and percentiles are inverted. View the team defense leaderboards.
Points Allowed Per Game PA/G
Average points allowed per game. The headline measure of team defense.
How it is calculated
Points Allowed Per Game = Total Points Allowed ÷ Games Played.
How to read it on CFB Savant
Lower is better, so the percentile is inverted: fewer points allowed earns a higher percentile. A 95th-percentile PA/G means the defense allowed fewer points than about 95% of FBS teams.
Example
A defense allowing 9.8 PA/G is elite. One allowing 38.0 PA/G is giving up points on most opponent possessions.
Yards Allowed Per Game YA/G
Average total yards allowed per game, split into pass and rush components on the team card.
How it is calculated
Yards Allowed Per Game = Total Yards Allowed ÷ Games Played. Pass and rush splits use the same denominator.
How to read it on CFB Savant
Lower is better, so percentiles are inverted. Read total yards allowed with points allowed: a defense that bends but does not break will rank better in points than in yards.
Example
A defense allowing 250 YA/G is smothering. One allowing 480 YA/G is giving up chunk yardage consistently.
Sacks, Tackles for Loss, and Takeaways Sacks/G / TFL/G / TO Forced/G
Per-game rates for sacks, tackles for loss, and takeaways forced. These are havoc measures — how often a defense creates negative or turnover plays.
How it is calculated
Each is the season total divided by games played.
How to read it on CFB Savant
Higher is better for all three. Havoc stats explain how a defense gets off the field: sacks kill drives, tackles for loss create long third downs, takeaways end possessions.
Example
A defense at 3.2 sacks per game and 1.2 takeaways per game is disruptive every week. One at 0.8 sacks and 0.4 takeaways is passive.
Opponent Third-Down Percentage Opp 3rd %
Share of opponent third-down attempts that the defense allowed to convert.
How it is calculated
Opponent Third-Down % = 100 × Opponent Conversions ÷ Opponent Attempts.
How to read it on CFB Savant
Lower is better, so the percentile is inverted. A low opponent third-down percentage means the defense gets off the field.
Example
A defense holding opponents to 33% on third down is winning money downs. One allowing 50% is letting drives continue.
Questions about a specific number? See the Contact page for how to report a stat that looks off.