The Legend

95th-percentile exit velocity (mph) — how hard he hits it at his best

A hitter's 95th-percentile exit velocity, in mph — how hard he hits the ball when he really connects. We use the 95th percentile instead of the average because top-of-range contact is the skill and mishits dilute the average. It is the most stable metric in our engine (year-over-year r = 0.88), which means it travels forward in time better than any batting average. In our current MLB sample the lightest regulars sit near 98–100 and the loudest near 112.

The same hitter's EV95 in the previous season, for the men who qualified in both

His EV95 last year, printed beside his EV95 this year. It is the only column on the Baseline that says anything about prediction — a leaderboard of the loudest bats shows that the metric finds the men you would have named, which is a different claim entirely. Two seasons on one row is the claim: if the number travels, it is a skill; if it scatters, it was a mood. Both seasons must clear 100 balls in play to appear here, because a fragment of a season beside a whole one would answer the question dishonestly. Across every pair we have, the correlation is r = 0.88.

Change in EV95 against last season, in mph — a difference, not a speed

This season's EV95 minus last season's, in mph. Positive means he is hitting the ball harder than he did a year ago. Read it as a difference and never as a speed: +1.4 is a gain of a mph and a half, not a hitter, and it cannot be compared with the EV95 column beside it. Most men land near zero, and that is the point of the column — the noise in it is what a real change has to beat. A large hold is worth a look in either direction, but it is a season in progress, not a verdict.

The weekly trailing EV95 line — the top of his range over time, not his average, so it can fall while his contact improves

The sparkline on a player page: his EV95 recomputed each week over a trailing window of balls in play, so the season has a shape instead of one number. Read it as the TOP of his range and never as the middle, because that is where a line like this misleads. A falling line does not mean he is hitting the ball worse. His hardest contact can come down in the same weeks his typical contact goes up — they are different questions, and only one of them is on the chart. We have watched it happen through the best month of a hitter's professional life: the trailing line fell 2.7 mph while his mean exit velocity over the same windows reached a season high. So "higher EV95 is better" is true, and "a falling EV95 line means he got worse" is a different claim that is not. The boards rank on the full-season figure, never on the trailing one.

Whether his results have kept up with his contact, run behind it, or run ahead of it

The Baseline's one-word read on where a hitter's line sits against the contact behind it, taken straight off his park-adjusted luck gap. OWED — his contact has outrun his results, and the box score owes him. SQUARE — the line and the bat agree. PAID — his results have outrun his contact, and he has been paid for balls that were not struck well enough to keep paying. The cut is ±0.030 of contact value, about six-tenths of a standard deviation of the league, and it is a fixed line rather than a share of the field — a share would guarantee the same number of lucky hitters every night whether or not the league had any. It is a description, not a call: a call gets filed to the Ledger with a metric, a threshold and a window.

How much contact value a park turns the same batted ball into, against what the launch physics expected

Take every ball put in play at a park, ask the launch-speed / launch-angle grid what that ball is worth league-wide, and average the difference. Positive: this park converts the same contact into more than the physics expected. Negative: it eats it. The batted-ball quality arriving at every park is close to league average — what differs is what the park does with it. The scale is value per ball in play on the wOBA weights — an out 0, a single 0.9, a double 1.25, a triple 1.6, a home run 2. It is NOT runs. It carries no favor on its own: a park that inflates contact is good news for the hitter and bad news for the pitcher, so we do not rank anyone on it. It is REGRESSED toward zero on a 2,000-ball prior, because a single park-season is only about 3,900 balls in play and the raw average of that many is mostly noise — which is why the count is always printed beside it. It is the same number the engine subtracts to park-adjust every hitter's luck gap; there is one definition, not a public one and a private one.

Results on contact minus what the launch physics support, park-adjusted

Actual results on contact minus what the launch physics say those batted balls should have produced, adjusted for park. Positive: results running ahead of the contact quality underneath them — fade shape. Negative: contact getting underpaid — breakout shape. The gap is mostly transient. It tends to close, in both directions, which is exactly what makes it a signal. The scale: value per ball in play on the wOBA weights — an out 0, a single 0.9, a double 1.25, a triple 1.6, a home run 2. So a gap of 0.042 is 0.042 of wOBA-on-contact. It is NOT 0.042 runs, and our own One Chart called it runs until 2026-07-14.

Higher means the bat is real and the numbers haven't caught up yet

It combines contact quality (how hard he hits it) with the luck gap (how little he has been paid for it) into one number, scaled so that higher means more room to catch up. On our pre-registered out-of-sample holdout, the top quintile of this score outgained the bottom by +1.15 annualized WAR. Two boundaries, and we mean both. It ranks who is likely to OUT-PERFORM his current value — not who is best; the best players in baseball are already paid for their contact and score low here. And it is a rank, not a prediction: it does not say he is due, because "due" claims a thing will happen and claims it will happen soon, and we do not say that without filing a call — a metric, a threshold, a deadline, and a public verdict. The gap tends to close. It does not owe him anything.

What his contact was worth — the value his batted balls earned, per ball in play

What a hitter's contact DESERVED. Take every ball he put in play, ask what a ball struck that hard, at that angle, in that park is worth on average, and average the answers. It is the line he earned, before a single fielder, bounce or gust had a say in whether he got it. The scale is value on contact, where an out 0, a single 0.9, a double 1.25, a triple 1.6, a home run 2 — so it is not runs and it is not batting average; a .450 here means his average ball in play was worth about half a double. Set it against GOT and the difference is the luck gap.

What his contact actually produced, per ball in play — on the same scale as Deserved

What a hitter's contact actually PRODUCED — the same average, over the same balls, using what really happened to each one instead of what it was worth. Deliberately on the identical scale as DESERVED, because the only reason both are printed is so you can subtract one from the other: that difference is the luck gap, defined below, and it is the quantity every board on this site ranks. Above his deserved line he has been paid more than he struck for; below it, less.

A Triple-A hitter's contact graded on the major-league value grid

For AAA hitters: their batted balls graded on the MLB value grid — as if each ball in play had been struck in a big-league park. It projects contact quality across the promotion. Honest boundary: our signals are holdout-validated at the MLB level; the AAA application is the frontier. It is a TRANSLATION, not the thing: it does not share a scale with a major-leaguer's DESERVED and the engine refuses to difference the two.

Our pitch-quality grade — stability-validated, not outcome-validated. Stuff is not command.

Our pitch-quality grade, built from physical pitch characteristics and calibrated on MLB pitches. It is stability-validated (year-over-year r = 0.87) but not yet outcome-validated, which is why pitchers are watchlist rows, never calls. And stuff is not command.

Runs allowed per nine — ALL runs, not just earned. Higher is worse, and it is not ERA.

Runs allowed per nine innings — every run that scores while he is on the mound, earned or not. It is the engine's run scale, and it is deliberately NOT ERA: ERA forgives the runs an official scorer rules unearned, and we don't, because a run is a run to the team trying to win. For the same innings a pitcher's RA9 sits at or above his ERA, so do not read the two as one number. It is one of the inverted metrics — higher is worse, however English reads it — and it is what an arm call is judged on: the runs he actually gives up across the window.

Fielding-independent pitching — what his strikeouts, walks and homers alone say he earned, on our RA9 scale.

Fielding-Independent Pitching — what a pitcher's own strikeouts, walks, and home runs say he earned, stripped of the seven fielders who handle everything hit into play. The idea, called DIPS, is that a pitcher controls those three true outcomes far more than what happens once a ball is put in play, so they carry more information about his next runs than his runs themselves do. We lift our FIP onto the RA9 scale with a season constant (about +3.5) applied before any comparison, so a 4.48 FIP reads straight against a 4.48 RA9 — do not set it against a FanGraphs FIP, which sits on the lower ERA scale and shows a different number. Higher is worse. Our study H-ARMGAP found it carries real information about a pitcher's next runs beyond his ERA, but out of sample it came in co-equal to ERA, not ahead of it — which is why we file arm calls and rank no pitcher board.

RA9 minus FIP — how far his runs have run below (lucky) or above (unlucky) his own components.

A pitcher's RA9 minus his FIP — the mirror of the hitter's luck gap, with the sign flipped because runs are the inverted kind of number. NEGATIVE means his runs have run BELOW his fielding-independent line: he has prevented more than his strikeouts, walks, and homers alone predict, because the sequencing, the strand rate, and the defense behind him have broken his way. That is the lucky, fade-shaped case, and the engine expects the runs to climb back toward the components. POSITIVE is the reverse — runs above the arm, unlucky, a buy shape. We only read the gap as luck when his FIP sits mid-pack; when his runs AND his components are both extreme it is mechanics rather than luck, and the engine sets it aside. It is the whole reason we can file a fade on a pitcher in the best run-prevention season of his life.

Run expectancy across the 24 base-out states — the run accounting under every number we print

The run accounting underneath everything we print. Before each plate appearance an inning already has a value: the runs an average team scores from that base-and-out state to the end of the half. Bases empty with nobody out is worth about half a run; bases loaded with nobody out, about 2.3. There are twenty-four such states — three out-counts times eight arrangements of runners — which is where the 24 comes from. RE24 credits a batter with the change he causes, plus any runs that actually scored on the play. A single that lifts the inning from 0.50 to 1.10 pays him +0.60. A strikeout that drops it from 0.50 to 0.27 charges him −0.23. That is the whole idea, and it is why RE24 sees what a batting average cannot: the same strikeout costs about 0.9 runs with two in scoring position and one out, and about 0.16 with the bases empty. Same line in the box score, six times the damage. We compute it ourselves from public play-by-play, rebuilding the run-expectancy table from each league's own season so that MLB and AAA never share a baseline; our 2024 values correlate at 0.975 with FanGraphs'. A single game's RE24 is what we call night value. It works the same on the mound, from the other side of the ball: a pitcher is scored by the swings in expectancy the batting team causes, so a run he keeps off the board is a run he's worth. We flip the sign so a good start reads positive — a hitter's value and a pitcher's live on one runs scale.

What his night was worth in runs (RE24). Negative is normal.

What a player's night was worth in runs, by our own RE24 accounting. Every plate appearance moves the inning's run expectancy — the runs an average team scores from that base-and-out state to the end of the half — and the player is credited the difference, plus any runs that scored on the play. Raise it and you're paid; make an out and you're charged. A pitcher's night value is run prevention on the same scale: the runs he kept off the board versus an average start — same currency as a hitter's, opposite side of the ball. Negative is normal. An out always costs expectancy, so an 0-for-4 is negative before anyone asks how it looked, and roughly half of any night's lines sit below zero — zero is an average night, not a bad one. What varies is the situation: a strikeout with two in scoring position and one out costs about 0.9 runs; the same strikeout with the bases empty costs about 0.16. Same out in the box score, six times the damage. (A sacrifice fly is a small negative too: it scores a run but burns an out and leaves the inning worth less than it found it.) That gap is the whole reason the Overnight ranks by this instead of by hits. The run-expectancy table is rebuilt from each league's own season, so MLB and AAA never share a baseline.

What a called ball or strike is worth — and it depends on the count. The engine behind an umpire's run favor on a Blue Note.

A called ball or strike is not worth the same everywhere — its value depends on the count. A called strike on 3-2 ends the at-bat in a strikeout; the same strike on 0-0 barely moves the needle. So we built a table of what a plate appearance is worth from each of the twelve ball-strike counts (0-0 through 3-2), with a walk waiting on the fourth ball and an out on the third strike, averaged over roughly seven million pitches since 2015. The table has the shape every fan already feels in their gut: 3-0 is the hitter's count and worth the most, 0-2 the pitcher's and worth the least. When an umpire misses a call, we price the miss by the gap between the count his call produced and the count the correct call would have — so a missed strike three, a strikeout erased and a batter handed life, is worth far more than a missed ball on 3-0. That gap, summed and signed to the team it helped, is the run favor printed on every Blue Note. It is the same linear-weight run accounting as RE24, read down the count instead of across the bases — and, like everything here, we compute it ourselves from public play-by-play. Nobody handed it to us.

Sample size — how many batted balls the signals are measured on

Sample size: how many batted balls a hitter's signals are measured on this season. More is more trustworthy; under 100 we flag the row ourselves — see the dagger below.

Sample size for the stuff grade — tracked pitches

Sample size for the stuff grade: how many tracked pitches the grade is computed from. Unlike a results stat, this one settles fast — stuff is measured from the physical shape of a pitch, not from what hitters did to it, so a few hundred pitches is already a grade rather than a rumor. That is why no thin-sample dagger hangs on it.

The row hasn't earned its numbers: under 20 IP, 100 balls in play, or 20 games

A dagger on a sample means the row hasn't earned its numbers: under 20 innings, under 100 balls in play, or under 20 games. The line beside it is a rumor, not a reading — an ERA over fourteen innings will tell you almost anything you want to hear. We could quietly rank those rows and let you assume they were solid. We flag them instead, and leave them on the board, because hiding a thin row is the same lie as trusting it.

Where an open call is now, against where the model expected it by now — a progress reading, not a forecast

A call says a number will move by a threshold before a date. This is how far it has actually moved, against how far the model expected it to have moved BY NOW. The expectation is measured in BALLS IN PLAY, not in days, and that is the whole trick: the threshold itself was set by assuming a season-to-date average can only move as far as the new evidence dilutes it. So a hitter who has not batted is not behind — he is waiting, and we say so. Grading him against a calendar would judge him for days in which he never swung. Under 25 balls in play we print no grade at all: we simulated 1,162 calls under this rule across our six training seasons, and below that sample the reading disagrees with the eventual verdict a third to nearly half the time (33% at 10–24 balls, 42% under 10) — a coin flip with a color. IT IS NOT A PREDICTION. Even above the floor it disagrees with the final result about a quarter of the time. It tells you where he is, not how it ends. The verdict is the verdict, and it lands on the window's last day whatever this says.

His line over the trailing two weeks, with the date it runs through

The player's line over the trailing two weeks, straight from MLB's public data. The window prints the date it runs THROUGH, because fourteen days ending we-won't-say-when is a duration and not a window — and a duration can never look stale. If the nightly enrichment step fails it warns rather than dying, so the line can be a run or two behind; the through-date is how you can tell, and it is why we print it. Absent when a player hasn't appeared in the window.

How a ballpark changes an outcome, against a league average of 100 — and there is never just one of them

An index for how often something happens in a park against how often it happens league-wide, where 100 is average, 120 is twenty percent more and 85 is fifteen percent less. THERE IS NO SUCH THING AS ONE PARK FACTOR, and printing a single number is the reason we show a whole row. Coors Field this season is 119 for runs and 111 for home runs and 237 for TRIPLES — the thin air is a small part of it and the enormous outfield is most of it, and a reader handed 'Coors: 119' would have been told the least interesting thing about the building. THE WINDOW IS PART OF THE NUMBER. A single season and a three-year rolling view are different measurements of the same park and disagree by about three points typically; both are printed, each labeled with the years it covers. A park without three seasons of history has no rolling figure at all and we print that absence rather than filling it with 100 — the park that happens to affect is Sutter Health Park, which is the most extreme park in baseball. AND ONE SEASON IS NOISY. Across 2022–2026 a park's own single-season contact index moves by a median of three points with the walls standing still; Globe Life Field ran 103, 107, 95, 90 and 98. Each park page prints its own band so a single figure is never read as more precise than it is. Our own park effect, printed underneath, is a DIFFERENT quantity on a different scale — value on contact per ball in play, the amount we subtract from every luck gap measured there — and it cannot be added to or subtracted from any index. The two can even point opposite ways, because the index counts all offense while ours is only the residual against what the batted balls deserved.

Where he stands among qualified hitters — and the pool is named, counted and dated.

Four of his contact numbers, each drawn as a place among every qualified major-league hitter rather than as a raw figure. The pool is the same one the boards are selected out of: hitters with at least 100 balls in play this season. Its size is printed on the module, because a percentile without a denominator is a rank pretending to be a measurement. RIGHT IS ALWAYS BETTER, on every row, and that is why one row is drawn flipped: the luck gap is a number where higher is worse (his results have outrun his contact), so a hitter with the lowest gap in the league sits at the right-hand end. The printed value falls while the dot moves right, and the module says FLIPPED on that row rather than letting the geometry quietly argue with the number. The direction is not decided by this page — it is read out of the engine's own metric registry, the same declaration every board colors its arrows from. We print 1 to 99 and never 0 or 100: the very hardest hitter in the league is at the top of the men measured so far, which is not the same as a ceiling.

The hollow dot — where he stood 30 days ago, against the league as it was then.

The solid dot is today. The hollow dot is the same number thirty days ago, and the line between them is the move. It is a real rebuild of that evening, not today's answer with his recent games removed: the value grid, the park adjustments and the qualifying field are all recomputed from the balls that had actually been hit by then. That distinction is worth up to five percentile points on the luck-gap row, because the grid and the park factors were different numbers a month ago. SOME HITTERS HAVE NO HOLLOW DOT, and that is information rather than a gap: it means he had not reached 100 balls in play a month ago, so there is no earlier place to mark. The module says so in words. A zero would have claimed he sat at the bottom of the league, which is a different fact entirely. The earlier field is also smaller for the same reason, so both counts are printed — his place then was a place among fewer men.

The scale our contact numbers are in — not runs. A single is 0.9, a homer 2.0, an out 0.

The scale the luck gap and the overlay are measured in, and the reason a gap of 0.042 is not 0.042 runs. Weighted On-Base Average pays each outcome what it is actually worth to an offense, instead of pretending a single and a home run are the same event (batting average) or that a home run is four singles (slugging). Our contact numbers use the wOBA weights: an out 0, a single 0.9, a double 1.25, a triple 1.6, a home run 2. So 'value on contact' is per ball in play, on that scale. IT IS NOT RUNS, and we printed it as runs on our own One Chart until 2026-07-14 — a run value would put a single near 0.47 and a homer near 1.40, which is a different number entirely. The number was right and the word beside it was wrong, which is the only kind of error that gets past everybody.

How well this year predicts next year. 1.0 = perfectly; 0 = not at all.

The letter r, printed all over this site, and the number that decides which of our signals you should trust. It measures how tightly two things move together, from 1.0 (when one goes up the other always does) through 0 (knowing one tells you nothing about the other) to −1.0 (they move exactly opposite). It is the whole argument in one letter. A hitter's hard-contact number this year predicts his next year at r = 0.88 — close to a fingerprint. His luck gap predicts his next year at r = 0.27 — close to nothing. That is why we bet on the first and against the second, and it is the single fact the engine rests on. It is not a percentage and it is not a probability. It says how much of the pattern travels, not how often we are right.

The win-scale sibling of RE24 — how much each play swung the team's chance of winning. Our own calc.

Win Probability Added — the win-scale sibling of RE24. Where RE24 credits a batter the RUNS his play added to the inning, WPA credits him the WIN PROBABILITY it added to the game — and the same plate appearance is worth far more in a tie in the ninth than up eight in the second, because only the late-and-close one actually moves who wins. It is leverage made into a number. We compute it ourselves from public play-by-play, the same pipeline as our run expectancy; cross-checked against FanGraphs for 2024 qualified batters it correlates at 0.905. It is context, not a signal the boards rank on: a season's WPA says who has already helped win games, not who is mispriced going forward. That is the whole distinction the site is built on — this is a result, and the boards trade in what the result hasn't caught up to yet.

The share of games that have been won from exactly this situation — always stated for the HOME team

Given the inning, the half, the outs, who is on which base and the score, win expectancy is the share of real games that ended in a win from exactly that situation. Ours is empirical, not modeled: we count what happened across every season in our lake and read the answer off the pile, with a coarser inning/score table standing in where a situation is too rare to trust on its own. Extra innings since 2020 are kept separate, because a tenth inning that starts with a runner already on second is not the same game state as one that doesn't. IT IS ALWAYS THE HOME TEAM'S NUMBER. 0.62 means the home side wins 62% of the time from here — so the very same 0.62 is bad news for the visitors, and a road hitter having a wonderful night drives it DOWN. It describes a GAME, never a player, and nobody is ranked on it.

How much the game can still swing on this one pitch. The average moment is 1.00

How much is riding on this moment. We take every play that has ever left a given situation, measure how far it moved win expectancy on average, and scale the whole thing so the typical moment in a typical game comes out at 1.00. Two out in the ninth of a one-run game runs high; up eight in the second runs near zero. IT IS A WEIGHT, NOT A GRADE. A high number is not a compliment and a low one is not a criticism — it is the size of the stage, not the quality of the performance, and it says nothing about the man standing on it. It is the multiplier underneath every clutch statistic on this site, which is why we print it rather than quietly folding it in.

How much the win probability moved, all night, added up — and scaled to a nine-inning game

Add up every swing in win expectancy across a whole ballgame — each one counted regardless of which way it went — and you have the Game Excitement Index. A game that was 4-0 after two and stayed there barely moves; a game that changed hands four times moves constantly, and the number is the total distance traveled. It is scaled to nine innings, because otherwise it would quietly be a ranking of how LONG games were: a thirteen-inning game accumulates more of everything simply by lasting longer. A game shortened below seven innings gets no index at all rather than a scaled-up one, because multiplying a game we only half-saw is how you manufacture excitement out of rain. IT DESCRIBES A GAME, NEVER A PLAYER. A wild night was not a good night for anyone in particular — the losing pitcher was in it too — so nobody is ranked on it and it is not colored like a signal. Across the 2026 season the middle of the distribution sits near 2.6; a quiet game runs under 1.5, a memorable one over 4.1.

One divided by the winner's worst moment — how close the team that won came to losing

Find the lowest the eventual WINNER's win probability ever got, and take one divided by it. A team that bottomed out at a 1-in-8 chance and won anyway scores 8. A team that led wire to wire still bottoms out somewhere near where it started, so it scores about 2 — which is the number's way of saying there was no comeback, not that there was a small one. WE CAP IT, AND WE SAY SO. The arithmetic runs to enormous figures — one 2026 game computes to 263 — but the low end of a win-expectancy table is built on a few dozen observed games, and at those odds a single extra game in the pile moves the answer by hundreds. A cell reading zero means nobody came back from there in about fifty tries; it does not mean nobody can. So past the point the reciprocal stops measuring anything we print a plus sign instead of a decimal point, and when we rank comebacks we rank them on how low the winner actually got — never on this number, whose ordering at the top is an artifact of the division rather than a fact about the games.

The single play that moved the game's outcome furthest, and what was riding on it

Of every plate appearance in the game, the one that moved win expectancy the furthest. It is not the loudest play or the longest ball — a two-run single with two out in the ninth of a one-run game outweighs a three-run homer in a blowout, because only one of them changed who was going to win. The leverage printed beside it is how much was riding on that moment before it happened — see WPA and leverage index, both on this page.

The top exit velocity a hitter produced in one game — his best swing, not his typical one

The fastest a ball came off his bat in a single game, in miles per hour. One number, from one swing: the night's best, not its average. IT IS NOT EV95. Our EV95 is a whole SEASON's 95th percentile — where a hitter lives when he really connects, across hundreds of batted balls. This is one game's single hardest ball. Both are printed in miles per hour and they are not the same quantity, so the engine refuses to subtract one from the other: a man whose season EV95 is 104 and who hit 107 tonight has not gained three miles an hour of anything. When we forecast this number in advance, it is graded against the actual hardest ball of that same game — like against like, in the metric's own unit.

Who moved on the boards since last night, and which way

What changed on the boards overnight — who climbed, who fell, who arrived, who dropped off. The boards are rebuilt every night from that night's games, so a rank is a photograph, not a verdict, and the Drift is the movement between two photographs. READ THE DIRECTION AGAINST THE BOARD, NOT AGAINST YOUR INSTINCT. A climb on the Breakout board means the engine likes him more. A climb on the Fade board means it likes him LESS — he is further into the territory we bet against. We printed a Fever Report that got this exactly backwards (No. 004, and the correction rides above it forever), which is why the color on every arrow follows the engine's read rather than the arrow's direction.

Tracked, not flagged — the four Index wings

Tracked, not flagged. Four wings, all data-defined and refreshed nightly: the top 25 qualified bats by season WAR (MLB's calculation), the top 10 pitchers by the same — tracked ungraded, because our arm signals aren't outcome-validated yet — the consensus top 25 prospects (FanGraphs Board, snapshotted weekly), each shown next to what our engine grades, and the milestone chasers, who join while the chase is hot and drop out when it cools (their pages say so, and the countdown is on the Chase). Every Index page carries a computed explanation of why no board flagged him. A flag means the engine found a mispricing; the Index means we're watching. The distinction is the product.

What X is saying about him this week. Chatter, not signal.

Short summaries of what X is saying about a player — on the field and on cardboard — over the trailing week. Built by searching X posts date-scoped to that window (summarized by Grok, xAI); when there are fewer than a handful of relevant posts, the track reads quiet rather than inventing a take, and a summary is never allowed to draw on anything older than the window. Every strip carries its as-of date. Tones: ▲ hot, △ warm, ▼ cold, ▬ quiet. It is chatter, not signal — the boards do not read it and it does not move a call.

Wins Above Replacement — MLB's own calculation, not ours

Wins Above Replacement — a player's total value versus a replacement-level fill-in, denominated in wins. Rough guide: 2 is a solid regular, 4 an All-Star case, 6+ an MVP conversation. Not ours: the season WAR on player pages is MLB's own calculation, pulled from the same public API as everything else and refreshed nightly. When a ledger call cites WAR it uses FanGraphs' version and says so — the flavors differ by a few tenths, which is why we always name the source.

On-base plus slugging — the league's own line, not an FVR signal

On-base plus slugging — the back-of-the-card summary of what a hitter has actually produced. Rough guide: .700 is light, .800 is a good regular, .900+ is a middle-of-the-order bat. Not ours: it is MLB's own line, and it is not an input to any FVR signal. It sits on the boards because every board ranks a gap between what a hitter's contact deserves and what he has gotten — and OPS is the half of that gap the market can already see. On the breakout board, a fat EV95 beside a thin OPS is the entire thesis in one row.

Feet per second on his best runs — Statcast's, not ours. Printed on the Fade board; never ranked on.

How fast a man runs, in feet per second, on his best competitive runs — Statcast's measurement, not ours. It sits on the Fade board because it explains that board: our expectation model grades a batted ball on how it was struck, and it cannot see legs, so a fast man beats it every year (the correlation between sprint speed and our luck gap has run about +0.3 in each of seven seasons). His gap is fat partly because he can run. You should be able to see that. What we do NOT do is rank on it: we built the corrected board — the luck gap with speed subtracted out — tested it, and it picked WORSE fade candidates than the plain gap did, in both test years. So the number is printed and the engine keeps its hands off it. Rough scale: 27.3 ft/s is average, 30+ is elite, under 26 is a catcher. Both receipts are on the methodology page, the one that worked and the one that didn't.

Earned run average — the league's own line, not an FVR signal

Earned run average — runs allowed per nine innings. Not ours, and deliberately printed next to the stuff grade on the AAA arms board: stuff is not command, so a high grade beside a fat ERA is exactly the thing our pitcher signals cannot see. We show it rather than hide it. Treat both with the sample beside them — an ERA over twenty innings is a rumor.

A falsifiable prediction: metric, threshold, window, timestamp

A falsifiable prediction: filed with a metric, a threshold, a window, and a timestamp. Nothing about a call is edited after filing.

The quantity a call is judged on

The quantity a call is judged on. Currently three kinds: results on contact vs baseline, WAR added from filing (FanGraphs WAR accumulated after the filing timestamp), and MLB debut by window end. “Results on contact” is his actual value on contact — what the bat really produced — measured against where it stood at the filing timestamp. It is NOT park-adjusted, and this page said it was until 2026-07-14. The boards RANK on the park-adjusted gap, so that a ballpark never decides who gets flagged; a call RESOLVES on what actually happened to him. Those are different jobs and we should have said so.

The number the metric must clear for the call to be a HIT

The number the metric must clear for the call to be a HIT.

Where the metric stood at the moment of filing

Where the metric stood at the moment of filing — the starting line the threshold is measured from.

The date the call resolves. No extensions.

The date the call resolves. No extensions, no exceptions.

The engine computes the metric and stamps HIT or MISS

When the window closes, the engine computes the metric and stamps HIT or MISS. Nobody argues, and nobody grades their own homework.

Resolved for us; resolved against us; window still open

Resolved in our favor; resolved against us; window still open. The record on every page counts all three, because a hit rate with the misses removed is marketing, not measurement.

Errors are appended in public; the original filing stands

If we discover an error mid-window, the correction is appended in public and noted on the call. The original filing stays as filed.