THE UPSIDE INDEX · A GUIDE TO THE NUMBERS

What do our scores mean?

Use this guide to understand what each score tells you, how to read it and what it cannot promise. Higher is not always better: an earlier draft pick has a lower number.

NFL Draft scores

UIR · Upside Index Rating · 0–100: Our rating of a college player’s potential to earn a lasting NFL starting role, predicted from his college performance and physical traits. Higher is better. A score of 80 does not mean an 80% chance of success. Positions are judged by the skills their jobs require.
Upside Draft Projection · our predicted draft order: Where our draft forecast places a player in the class. Number 10 is ahead of number 40. This is a ranking, not an exact pick prediction. It can differ from UIR: draft expectations are not the same as our view of a player’s NFL potential.
Draft Market Scanner · two position ranks: Compares our order with other draft boards using the same players at the same position. If we rank a receiver 5th and draft boards rank him 12th, we rank him 7 places higher. That is a reason to look closer, not proof he is a bargain. The biggest-gap view accounts for the size of each position group.

How we build and check prospect ratings

We consider college production, playing time, physical traits and the position a player plays. We compare past prospects with what they later achieved in the NFL. A player’s actual draft pick is a benchmark, never an input to his UIR.

Some historical ratings were rebuilt later rather than saved before those drafts. Some past classes were studied more than once. These are not independent tests of a new live prediction. Small position groups and missing age or recruiting information limit what we can conclude. We do not change a score simply to put a famous player higher.

Generational: elite at his own position

This label needs a UIR of at least 97, a place in the class’s top 32, top-7.5% marks in three key position skills, and at least 10 games and 400 plays in the qualifying college season. For an edge rusher, those skills include pass-rush win rate: how often he beats a blocker. There is no required number of players who must receive the label. The label is a strict prospect description, not a promise of NFL greatness.

Fantasy projections and rankings

Projected points: Our estimate of a player’s fantasy score. It considers his role, recent play, team, opponent and availability. For tight ends it also uses how often he ran a route on his team’s pass plays, and how often those routes drew a target, over his previous three games (PFF charting, published each week before the next kickoff); a registered test on 2022-2025 found this more accurate. When that week’s charting is not yet published, the tight end forecast is made without it. It is not a guarantee. PPR gives one point per catch, half PPR gives half a point, and standard gives no points for catches.
Floor and ceiling: A lower and higher possible score. They help show risk. They are not the worst and best scores a player could ever have.
Chance of playing: Our estimate that an injured player will take part. It is already included in the points we show. Do not multiply the projection by that chance again.
Season-long value: Looks at the remaining games, not just this week. An injured player can regain value as his expected return gets closer. A bye is a week without a game, not a poor performance.

How injuries affect the numbers

Players confirmed out for the current game receive zero projected points for it. Unclear or conflicting reports stay labeled rather than guessed. Future weeks consider the injury, practice progress and past return patterns. A likely return date is not proof a player has been cleared.

We apply one injury adjustment. We do not add another penalty to points that already account for the injury. A teammate may gain work when someone is hurt; that extra role must be supported by the team’s depth and recent usage.

Tests on final game-day reports cannot prove how well an estimate made on Tuesday would work. A tested injury change that made immediate-week point errors worse was rejected.

How the waiver wire ranks pickups

Each available player shows three separate numbers: his projected points, his value over a typical replacement at his position, and how much he would improve a typical team's starting lineup. That last one is the fairest way to compare positions. A quarterback who scores more than a receiver still only helps the teams whose own quarterback is worse, which in a one-quarterback league is few of them. In Superflex leagues a second quarterback starts every week, so the same quarterback is worth much more.

We tested this on 56 past waiver weeks from 2022 to 2025, scoring each list by how much its top five pickups actually improved a typical lineup. Ranking by raw points improved a one-quarterback lineup by 0.17 points a week per pickup; ranking by lineup improvement added 0.67. The free preview now uses lineup improvement; when only a few available players would start for a typical team, the rest of its list is labelled bench depth and ordered by replacement value. The full board keeps replacement value as its default order because the two did equally well in one-quarterback leagues; you can switch the order. Past availability was estimated from production, not real league rosters, and we are now checking this week by week against real roster percentages (3 weeks complete so far, too few to judge yet).

Projections vs. actuals

Upside+ lets you look back at the forecasts saved before each week’s first game. We compare them with the full week’s results, without rewriting the old forecasts.

Rank difference = our projected rank − the actual finish. A player ranked 12th who finishes 5th is 7 places better than expected: +7, green, Outperformed. If he finishes 20th, he is 8 places worse: −8, red, Underperformed. Point difference means actual points minus projected points.

Actual finishes include players outside our original list. Ties share a place. Searching and filtering do not change the original ranks. A dash means we lack a reliable score or match, not that the player scored zero. Results wait for the whole week to finish. Future forecasts appear only when published; all 18 week choices do not mean all 18 forecasts are ready.

These are week-opening forecasts. Later injury news may change the live rankings. Custom league bonuses are not included.

The review table shows our rank, the actual finish and how many places we missed by. Green means a player did better than expected; it does not mean our prediction was accurate. The average miss uses the distance in either direction. “Within range” starts at six places, and you can choose three or twelve.

Which players and weekly calls count?

The starting view includes a player if either rank is in QB1–18, RB1–36, WR1–48 or TE1–24. This keeps strong finishers we missed in view. All players is always available. Searches and call-type filters do not change the summary totals for your selected position, scoring and player group.

A big-week call means the saved point forecast was at least 25% and three points above the player’s previous three to five recorded games. A low-week warning uses the same limits below that average. A big-week call must also rank him as a starter, and a low-week warning must rank him outside that group. We use QB1–12, RB1–24, WR1–36 and TE1–12 as a simple 12-team guide. A call succeeds only if the actual finish stays on that side of the starter line and the score clears the same point limit. A disappointing player we ranked as a starter cannot count as a correct warning. Byes are skipped; missing results and players without enough earlier games are excluded, never counted as wins. These labels review the saved forecasts; they were not separately published predictions. Later official stat corrections may change the recent-game comparison.

Completed results refresh in the existing early Tuesday windows, with a morning retry. A separate step lets the review update even if another rankings calculation fails. A missing or unfinished feed cannot erase an already completed week or create invented scores.

My Team and Draft War Room

My Team fills your league’s starting slots using the players and projections available. A player can only fill one slot. Known byes, ruled-out players and injured-reserve players cannot become new starters. After kickoff, a player already in a confirmed starting slot stays there. Unknown scores leave the total incomplete.

Add/drop and trade comparisons show the change in your best projected lineup before and after a move. If you select several weeks, each counts equally. These are lineup estimates, not proof that a trade is fair for a full season or that another manager will accept.

Waiver bids are spending suggestions based on your remaining budget. They do not predict the winning price. Backup claims depend on what happens to earlier claims; rerun the plan if the outcome changes. The tool never submits a league transaction for you.

Draft War Room runs possible drafts using player quality, draft expectations and team needs. Each pick has one owner, and a picked player leaves the board. Your saved plan keeps the choices, picks and ratings used at the time, so reopening it can show the same draft. Trying a trade does not mean an NFL team would agree.

Matchup scores · 0–100

A higher score means a more favorable matchup for that position. A score of 50 is neutral. The opponent can be easier for running backs and harder for receivers.

The forecast view uses only matchup changes that passed our tests. Where a tested change did not help, the forecast stays neutral. Recent opponent results can still be shown separately; they do not prove how the next game will go.

Your overall playoff score averages the weeks you selected. A bye counts as zero because there is no game; it is never treated as zero-point performance in the training games. An unknown matchup leaves the overall score incomplete. These ratings help compare schedules. They are not extra fantasy points and are not added again to a projection that already includes the opponent.

ULS · Upside Legacy Score · career points

ULS measures the career a player has already built. Higher is better, and there is no 100-point cap. It rewards excellent seasons, playing time, standout years and verified honors. Players are judged relative to their position. It is not a forecast for a college prospect or a perfect answer to every greatest-of-all-time debate.

Prime Legacy focuses on a player’s best five completed seasons. It helps show a shorter but outstanding career. A player needs at least three completed seasons with at least half a usual season’s workload. Career pace compares the same number of years after NFL entry. It does not award points for seasons that have not happened.

What counts, and what is still missing?

Career points combine sustained quality, exceptional seasons, a best-three-season bonus and limited playoff credit. The strongest eligible annual honor shares credit with performance; overlapping awards are not added again and again. Super Bowl MVP can add at most 2.5 points in a season. Team titles alone add no points.

Early careers and some positions have missing playing-time or performance records. We show those gaps instead of treating them as zeros. Most performance records start in 2006; kicking and punting coverage starts later. Incomplete career windows cannot fairly settle a whole-career comparison.

The full score still needs independent career comparisons from knowledgeable reviewers. Predicting future playing time does not prove a player had a greater career. A tested larger peak bonus and a tested quarterback adjustment failed required checks, so the existing parts were kept. Career-score research has not established a universal improvement for prospect ratings. We have not approved it as a new UIR training target.

How well do the weekly projections work?

We built the weekly method using 2000-2016, chose its settings using 2017-2021, then tested it on 2022-2025. The comparison is a player’s average score from his previous five games.

The model has not met every target. Results vary by the question being tested. We do not claim a clear improvement when the measured difference could be chance. Use the rankings as help, not certainty.

See every weekly test result

“Too small to tell” means the difference could be chance. We do not call it a proven improvement. Some results count players with recorded statistics, so they do not describe every injury absence.

Swipe the table on a phone to compare the results.

What we checked	Our result	Previous five-game average	Change in the better direction	What the test can tell us
Average point errorHow far our prediction was from the actual score. Lower is better.	4.656	5.049	7.78%	Clear difference
Points left on the benchPoints missed compared with a perfect lineup picked after the games. Lower is better.	97.090	108.129	10.21%	Clear difference
Share of possible lineup pointsHow much of a perfect lineup’s score our choices earned. Higher is better.	77.64%	75.32%	3.08%	Clear difference
Best players near the topHow well we placed the biggest scorers near the top of the rankings. Higher is better.	0.735	0.710	3.52%	Clear difference
Whole ranking orderHow closely the full list matched the actual finish. Higher is better.	0.533	0.490	8.73%	Clear difference
Choosing between two playersHow often our higher-ranked player scored more. Higher is better.	58.23%	55.66%	4.62%	Clear difference
Close start/sit choicesThe same comparison when the players looked closely matched before the game. Higher is better.	52.26%	51.98%	0.54%	Too small to tell

Our checks before calling a model better

lower average point error by at least 5%: Passed
better ordering of top players by at least 3%: Passed
fewer points left on the bench by at least 5%: Passed
higher-ranked player wins at least 60% of comparisons: Not met
no Position Mae Regression over 1%: Passed

The separate early-season role-change study reused past seasons that had already been examined. It is not an independent test. Other small position studies can be too small to settle whether a change helps.

Called It and other games

Called It: your decision grade measures your read of the defense and your play plan. Football points measure how your simulated drive ended. They are different scores. The original game gives the setting; your alternate play outcomes are simulated, not NFL tracking footage.

Other games: the game’s own rules explain points, hints, give-up penalties and daily limits. A fantasy point projection, a UIR score and a game score measure different things. Do not compare them as if they use the same scale.

Past draft classes: official picks show what happened. Historical UIR scores show how today’s method rates those past college profiles; they are not necessarily forecasts saved before draft day.

What we will always tell you

What a score measures and which direction is better.
When information is missing, incomplete or out of date.
When a tested change failed or the evidence was too weak.
Whether a result is a prediction, a simulation or something that already happened.

For a real comparison, open the weekly rankings, Market Scanner or NFL Legacy. See Free and Upside+.