A receiver outranked two 400 yard passers, in the same game

On August 14, in one of the season's early Utah high school games, Spanish Fork's game against Snow Canyon turned into a shootout. Snow Canyon's Parker Olson, class of 2027, went 32 of 42 for 402 yards and 5 touchdowns through the air, adding 73 yards and 2 more scores on the ground. Spanish Fork's own quarterback, Hawkin Rasmussen, class of 2029, answered with 27 of 37 for 405 yards, 6 touchdowns and an interception. Two 400 yard passing performances, in one game, on one Friday night.

Neither of them posted the highest RPM score from that game. Nixon Warren, Spanish Fork's class of 2027, caught 10 passes for 267 yards and 5 touchdowns, and Recruit Intel's Performance Score model, RPM, graded his night a 100.00, the ceiling of the scale. Rasmussen graded out at 98.10. Olson graded out at 97.10. The receiver who caught more than half his own quarterback's touchdown passes outranked both quarterbacks, including the one throwing him the ball.

That follows directly from how RPM's components are built, and the rest of this piece is the receipts: what RPM actually measures, how a 267 yard receiving night beats two 400 yard passing nights on the same field, why a huge single game from a freshman does not automatically print a perfect score, and exactly where the model's schedule-strength adjustment stands today, five scoring windows into the season.

What RPM actually scores

RPM (Recruit Performance Metric) is built from three components read off a player's own box score: volume (yards per game, scaled to the position), efficiency (yards per touch or attempt), and touchdowns per game. Every player is first sorted into one of four position families by the shape of that box score itself, not by a roster label: quarterback, running back, receiver and tight end together, or defense. A quarterback line is scored as a quarterback; a heavy receiving line from the same athlete is scored as a receiver. Warren's roster tag at Spanish Fork is quarterback, but his August 14 box score is a pure receiving line, so that is the family his performance is graded against.

Each of the three components is converted into a percentile rank against every other player scored in the same position family, in the same scoring window. That is the mechanism behind the receiver beating the quarterbacks: a quarterback's 400 yard, multi touchdown night is compared against every other quarterback's night from that same window, and in the August 12 window, Olson's 402 yards ranked him in the 99th percentile of volume among the 105 quarterbacks scored that window, but his efficiency, yards per attempt across 42 throws, ranked only in the 82nd percentile, and that pulled his overall grade down. Rasmussen's efficiency ranked higher, in the 88th percentile, which is why he outgraded Olson even though the two quarterbacks threw for almost identical yardage, 405 to 402. Warren's receiving line, by contrast, ranked in the 100th percentile on efficiency and at 99.6 on both volume and touchdowns among the 240 players scored in the receiver and tight end family that window, because a 10 catch, 267 yard, 5 touchdown night is close to the ceiling of what a receiver's box score can show. The three components are then blended into one number (volume weighted heaviest, efficiency and touchdowns behind it, with the exact weighting tuned slightly differently for quarterbacks than for the other three families), and that blended number is itself converted to a percentile within the family one more time. That final percentile, 0 to 100, is RPM.

Warren's 100.00 stands alone atop the receiver and tight end family that window; Rasmussen's 98.10 sits third among the 105 quarterbacks scored. Neither number is close enough to the other to need a tie-break, but percentiles do saturate at the top, so when two players in the same family both print 100.00, Recruit Intel's boards break that specific tie on the underlying blended index before it gets converted to a percentile, the number beneath the printed grade. Warren's own index for this game works out to 56.215 by that formula, for context, even though nothing about his game required breaking a tie to get him there.

Why a huge freshman debut still does not print a perfect score

Florida's scoring window of August 21 tells the other half of the story. A freshman defender at First Academy, a private school in Leesburg, recorded 14 tackles and 3 interceptions against Lakeside Christian, a monster line for a class of 2030 player in his first month of varsity football. Measured against every defensive performance Recruit Intel scored that window (n=4,802, all states), his volume ranks at 97.9 and his efficiency at 99.9; measured against Florida's own defensive performances that same window, his line ranks third of 706.

But his final RPM comes in at 99.50, not a 100.00. RPM applies a credibility factor tied to games played, a real part of the published formula: a player with only one game on the board has that game's grade pulled toward the middle of the family, not toward zero, and it takes eight games before a performance counts at full strength with no pull applied at all. In this particular window that factor does not explain the gap, because all 4,802 defensive-family performances scored that week carry a single game played; the same pull applies to every row in the family, so it changes no one's order, including his. What actually keeps him off the ceiling is his touchdown and turnover production component, which sits at the 48th percentile of that same defensive family, below the median even though his volume and efficiency are both near the top; enough other defensive performances that window combined all three components more evenly to edge him out of the very top slot. The credibility factor is real and it will start mattering here once these same players have more than one game on the board to average against, but it is not what produced this particular 99.50.

The schedule-strength term, still filling in

RPM also carries a schedule-strength adjustment, a bonus or penalty tied to how strong the opponent was. It is part of the published formula, and it is designed to apply to three of the four position families: quarterback, running back, and receiver and tight end. It does not apply to the defensive family at all, by design, regardless of how far the season progresses. It was not doing anything on either of the games above.

Recruit Intel has five weekly RPM scoring windows on record this season: August 12 (1,095 scored performances), August 17 (11,406), August 21 (9,805), August 23 (22,558), and August 28 (19,063). The schedule-strength term read zero for every single scored performance in the first three of those windows. It reads nonzero in the last two: 7,544 of 22,558 scored performances in the August 23 window, 33.44 percent, and 6,505 of 19,063 in the August 28 window, 34.12 percent. Scoped to the three families the term actually applies to, that is 7,544 of 11,588, 65.10 percent, and 6,505 of 9,778, 66.53 percent, a more complete picture of where the adjustment stands than the all-family share alone suggests.

Warren's, Rasmussen's and Olson's grades above, and the Florida freshman's grade, all carry a zero schedule adjustment; none of them were pushed up or down by the strength of who they played. For the three quarterback, running back and receiver and tight end performances above, that is a genuine gap that is filling in window over window as opponent-rating data accrues. For the Florida freshman's defensive line, a zero schedule term is not a temporary gap at all, since the model does not currently score opponent strength for defensive performances under any window. Boards should be read with both limitations in mind: an incomplete adjustment for three families, and no adjustment at all, by design, for the fourth.

Sourcing

The two games above are worked examples, not a claim about how the model scores on an average night. They are receipts for two specific, verified performances, chosen because both happened to land near the top of their state's board in the season's first month. Every figure above is drawn live from Recruit Intel's canonical database as of September 4, 2026: the box score lines from player_game_stats, the RPM components from ri_performance_scores, and the underlying blended index re-derived from each row's own stored percentile values using the published formula weights. This analysis is registered as RI-RF401 in Recruit Intel's research bank, and every number is independently reproducible from that finding's stored queries or a fresh query against the same tables today.

Correction

Correction, September 4, 2026. This piece was published on September 2 with a five window table whose fourth entry read August 22 (19,302 scored performances), and with the statement that the schedule-strength term read zero across the first four windows. Recruit Intel re-ran the weekly scorer on September 2 and that fourth window was rewritten under a later boundary: it is now stored as August 23, with 22,558 scored performances, and its schedule-strength term is no longer zero, it carries 7,544 nonzero adjustments. The August 22 window and its 19,302 count no longer exist in the database and cannot be reproduced, so both are cut here and replaced above by the live August 23 values, re-derived on September 4 and registered in RI-RF401. Three further fixes in the same pass. The credibility sentence read 9,800 of 9,805 defensive-family performances, which are in fact the whole window's counts across all four position families rather than the defensive family's own, and now reads the defensive family's 4,802 of 4,802. Two percentile ranges written as a vague band rather than a figure now print the registered values they stood for: 99.6 for Warren's volume and touchdowns, 97.9 and 99.9 for the Florida freshman's volume and efficiency. And the scoped August 23 share printed 65.08 percent, which does not reproduce: 7,544 of 11,588 is 65.10 percent at this piece's two decimal convention, and the piece's other three percentages reproduce exactly at two decimals, so 65.08 was wrong against the numerator and denominator printed beside it. It now reads 65.10 percent. Every other figure in this piece reproduced exactly against the live database on September 4, 2026.