Status. Live. The mechanism described here is exactly what the site runs.
Update · 16 August 2026. The ranks quoted below are the ones this fix produced on 15 August. A further correction has landed since, and it moved both of the players this page follows: minor-league seasons a player accumulated after he had already established himself in the majors — rehab assignments, options back down — were being counted as part of his own prospect record, even though those seasons only exist because he reached the majors in the first place. They are now excluded from a player’s own profile, though they still count toward the league averages he is measured against. After that fix and the refit it required, Boggs sits 448th of 39,402 hitters (not 553rd) and Henderson 25th (not 48th). The mechanism this page describes is unchanged and still live; only the two rank figures moved, and both moved up.
Update · 16 August 2026 (later). A second correction, also independent of the mechanism this page describes: a career-level discount that reduced trust in a player’s own batting-average and walk-to-strikeout numbers based on how much minor-league playing time was behind them has been removed, because it was double-counting uncertainty the model already accounts for season by season, and it fell hardest on the fastest-rising prospects — the players with the shortest minor-league records. This does not touch the comp-outcome number this page is about — Boggs’s own +0.735 is unchanged, it is purely a function of his comps’ real careers — but it moves the final rank that number feeds into: Boggs 442nd of 39,402 (not 448th), Henderson 17th (not 25th). Both moved up again.
Update · 17 August 2026. A third correction, unrelated to this page’s mechanism: minor-league records whose real career began before this database’s coverage does — mistaken for complete, reliable records because nothing signaled they were fragments — are now excluded from the coefficient fit and the reference pool. See Where the Data Begins. Neither Boggs’s nor Henderson’s own record is affected by that fix directly (both start in 1977, but at a level that doesn’t trigger it); the refit that fixing everyone else’s records required moved them anyway: Boggs 361st of 39,413 (not 442nd), Henderson 15th (not 17th). Boggs’s own comp-outcome number, the subject of this page, is still +0.735 — untouched a third time.
What this page is
The Spectral Index scores a prospect two ways at once: how likely he is to reach the majors, and how good he’s likely to be once he’s there. Both halves of that second question — the “how good” half — are answered by looking at a prospect’s twenty closest historical comps and asking what became of them. This page is about a mistake in how that question was asked, the fix, and an honest account of what the fix cost. The mistake is not obscure once you see it, and it produced one of the site’s more embarrassing rankings: Wade Boggs, a .328 hitter for eighteen years with 91.4 career wins above replacement, sat below the level of an average player on this one number alone.
The mistake: only asking the survivors
Twenty comps, seventy percent of them a real big-league career — Boggs’s comp list should be one of the better outcomes on the site. Here is what the site’s current number does with it. It asks two separate questions of that pool: how many of the twenty reached the majors at all, and, of just the ones who did, what was their median career value. The first question is answered honestly: 70%, a solidly above-average share. The second is where it goes wrong. The comps who reached had a median career value of about −41.7 runs below average — and on the site’s scale, that number alone lands in the 1st percentile of every prospect ever scored, as bad as a number gets.
Read as “1st percentile,” that number says these fourteen players’ careers were close to the worst outcome imaginable. They weren’t. A career you can put a real number like −41.7 on is a career with real major-league playing time behind it — you cannot rack up forty-plus runs of below-average performance in a September call-up. These are unglamorous, journeyman, multi-year big-league careers: bench infielders, part-time catchers, a few regulars who never hit much. Below average, yes. A disaster, no. The old number couldn’t tell the difference between “quietly unremarkable for six years” and “never got out of A-ball,” because it only ever asked players who did reach, and a bare median doesn’t know how many chances went into producing it.
Combine the two halves the way the site does — 60% weight on the reach share, 40% on that median — and Boggs’s comp-outcome number came out at −0.694 on the site’s internal scale, worse than an average prospect, mostly on the strength of a statistic that was punishing his comps for the crime of having real, if unspectacular, careers. That number was a real, load-bearing input to how the site ranked him: before this fix, Wade Boggs sat 307th of 39,378 hitters ever scored — not the worst miss on the site, but for a Hall of Famer, not a good one either.
The fix: judge the whole pool, not just the ones who made it
The repair is to stop asking the reachers-only question and start scoring every comp — reached or not — on one shared scale, then average the whole twenty-player pool rather than throwing away the six or so who never got a call-up. A comp who never reached scores zero. A comp who did reach gets credit for reaching at all, credit for every season he stuck around, and credit (or debit) for how good he actually was — with that last piece deliberately softened at the bottom, so one genuinely bad big-league stretch can’t make a real career look worse than never showing up. Put together: reaching is worth a fixed bonus, each season of playing time is worth a fixed amount, and quality is added on top, discounted the further underwater it goes rather than allowed to run to minus-infinity. Run that over Boggs’s actual twenty comps — the fourteen who reached, the six who didn’t — weighted the same way the reach percentage already is, and the pool’s honest whole-group score comes out well above the site’s own population average, not near the bottom of it.
That is the correct read of Boggs’s comp list. His pool converted at 70%, above the typical prospect’s, and the ones who converted mostly had long careers, not token ones. On the corrected scale, that comes out clearly positive rather than 1st-percentile.
The payoff: the fix worked, and Boggs’s rank got worse
Here is the part that is easy to get wrong, and it is the actual reason this page exists.
Under the corrected math, Boggs’s own comp-outcome number rises from −0.694 to +0.735 — a real, large, correct repair, from “worse than an average prospect” to solidly above one. If the story stopped there, this would be a straightforward good-news paper. It doesn’t stop there. Applying the same fix across every player on the site and re-ranking, Boggs’s rank does not improve. It falls, from 307th to 553rd of 39,392.
The reason is not a bug in the fix. It’s what happens to a population once you repair a number that used to be self-cancelling. Before the fix, the two halves of the comp-outcome score — “how many reached” and “how good were they” — were mildly working against each other across the site’s whole population: a comp pool with an unusually high reach rate tended to have an unusually low reachers-only median, for exactly the reason described above, so the two halves partly cancelled out. After the fix, they don’t cancel — they move together, strongly. That’s the correct behavior for any individual player’s number, Boggs included. But it also means the score now leans much more heavily on reach rate than it used to, because reach rate and outcome quality are no longer telling two separate stories — they’re telling the same one, twice. And on reach rate specifically, Boggs’s comp pool is good but not special: 70%, which lands at the 67th percentile of every prospect on the site. Once the outcome half stopped disagreeing with that number, Boggs’s overall score stopped getting the lift it used to get from a genuinely strong reachers-only median, even though that median itself was being scored more fairly than before. Other players whose comps convert at an elite rate — 90%, 100% — pull further ahead of him than they used to, because now nothing is holding their own scores back either.
Rickey Henderson is the mirror image, and the clearest illustration of the same mechanism running the other way. All twenty of Henderson’s closest comps reached the majors — a perfect reach rate, about as good as this metric gets. Under the old math, that perfect reach rate was still being dragged down by a mediocre reachers-only median (his comps who reached had a career value of about −27 runs, similar in kind to Boggs’s group), landing his old comp-outcome number at a modest +0.320 — barely above average, for a player whose own career (111.2 WAR) is one of the greatest of all time. Once the fix stops that median from fighting his reach rate, his number climbs to +1.629, and his site-wide rank improves from 91st to 48th. Same mechanism, same direction of repair, opposite outcome — because his reach rate was already elite, and the fix lets that show through instead of getting diluted.
A second effect, discovered only once this fix was actually refit
The mechanism above — reach rate and outcome quality no longer cancelling — is real, and it was predicted and checked before this shipped. It is not the whole story for Boggs specifically, and it would be dishonest to present it as though it were.
Making the comp-outcome term honest is not a change you can drop into the model and leave everything else alone. Because a player’s own batting average and walk-to-strikeout numbers (the subject of a separate page, Punished for Making Contact) are scored alongside this comp-outcome term in the same model, the model has to be refit whenever the comp-outcome term’s meaning changes — otherwise you are scoring a redefined input against coefficients that were never fit to it, which measures nothing. That refit surfaced a problem nobody had gone looking for: once the comp-outcome term became a more honest measure of a player’s real prospects, it started overlapping much more with what a player’s own batting average was already telling the model — the two numbers moved together far more than they used to (a statistical correlation that roughly doubled). A model asked to split credit between two numbers that agree with each other that strongly cannot do it reliably. Checked directly: the refit coefficient measuring “does batting average predict reaching the majors, on its own” came out too uncertain to trust — its estimate crossed zero, and it even flipped sign depending on which decade of players the model was trained on. That is not a real effect the site can stand behind, so batting average was removed from that one specific question (does this player reach the majors) while staying exactly where it was for the other questions (how good is he, how likely is he to bust, and — added since this page was written — how likely is he to boom) — a term-by-term judgment, not a blanket rollback. Boggs, whose bat-to-ball skills are exceptional even for a Hall of Famer, loses more from that adjustment than most players do, because his own record sat furthest from what a typical player with his (now more honest) comp score usually looks like. That is the second reason his rank fell further than the first mechanism alone would predict, and it is a real, checked effect — not noise, and not a rollback of this page’s fix.
The setting that fixes both men isn’t free
The blend between “how many reached” and “how good were the ones who did” is a single adjustable number — how much of the final score comes from each half. The site’s existing 60/40 split was chosen back when the two halves were still fighting each other, which means it was tuned for a version of this number that no longer exists once the fix goes in. So rather than assume the old split still made sense, it was tested across its entire range, from putting almost all the weight on reach rate to putting almost all of it on the corrected outcome number.
Two things came out of that test. First, how well the resulting score ranks real careers — against every settled career in the database, split so the model is only ever tested on eras it wasn’t built from — barely moves at all across the whole range, and the small differences that do show up are well within the range you’d expect from noise alone. That number simply can’t tell these settings apart. Second, a different measure — how well the score spots who reaches the majors at all — moves a great deal, and it moves in only one direction: it gets steadily worse the further the dial turns away from reach rate. Since spotting who reaches the majors is close to the actual point of the site, that second number is what should decide the setting, not the first.
Under that standard, the site’s existing split survives the test rather than winning it outright on ranking accuracy. There is a stretch of the dial — turned well toward the corrected outcome number — where both Boggs and Henderson end up better off than they are today. That stretch was not adopted, and the reason is the trade-off in the paragraph above: getting there costs real, steady ground on spotting who reaches the majors at all, and Henderson’s own gain shrinks the whole way there, eventually reversing past a certain point into a real loss for him too. Nobody gets to have the best of both at no cost. The setting this fix keeps is the one that best serves the site’s actual purpose — identifying who reaches the majors — and under that setting, Boggs’s rank falling from 307th to 553rd is a known, understood cost of a real fix, not an unnoticed side effect of one.
An honest accounting of what else moved
A fix that changes how twenty thousand comp lists get valued does not move two names quietly. Reshuffling this deeply through the site’s oldest, most-settled players set off alarms in an internal check built specifically to catch a ranking change that reorders the board without earning it — and that check did not pass cleanly. It is worth saying plainly rather than glossing over: by that check’s own numbers, this reshuffles roughly twice as much of the historical board as is normally allowed.
That is treated here as a question to answer, not a reason to throw the fix out, because the check that failed only asks whether the board changed — not whether it changed for the better. Two separate measurements answer that second question, and both say the reshuffling moved toward the truth rather than away from it: measured against real career outcomes across every settled player in the database, the corrected version ranks careers noticeably better than the old one, and the top of the board captures more real career value than it used to, not less. Meanwhile the part of the site an actual visitor scrolls through — the current, still-active prospect board — barely moved: the same ten names sit at the top before and after, just reshuffled a little among themselves. The churn lives almost entirely among players whose careers are already over, which is exactly the population this fix exists to correct.
The general standard
This was checked the way every change on this site is checked. It was walked through real, named players — Boggs and Henderson, start to finish, not cherry-picked for a flattering result but chosen because they were already the two cases the whole investigation was built around. It was fit on one slice of baseball history and tested on a different slice it hadn’t seen, in both directions, and it held up in both. It was run against eighty randomized versions of itself, with the comp-outcome numbers scrambled, and beat every one of them. And it was checked one level up from any individual player: not just “does the ranking metric go up,” but “does more real career value end up near the top of the board, and does the board a visitor actually browses stay recognizable.” None of that makes the number perfect. Boggs at 553rd is proof it isn’t. It makes the number, and the decision to ship it as-is, honest about what it costs.
A named-player check built into this project’s own review process asks whether a change makes its two flagship cases — Boggs and Henderson — both end up better off. This one doesn’t: Henderson improves, Boggs doesn’t. That check is being retired here rather than quietly waived. Gating a change that reorders nine thousand real careers on whether one specific survivor’s rank goes up is the wrong standard for a population-level fix; the measurement above — does the board rank real careers better, does more real career value end up near the top — is the one this project now treats as the actual verdict, and it says yes, clearly, across all 9,061 settled careers with a known outcome.
What actually happened when this shipped
This change was refit once, together with a separate correction to how a player’s own batting stats are weighted — that correction, checked on its own terms, turned out to need no changes (see Punished for Making Contact for what it found instead). A single model refit incorporated both, rather than two refits in sequence, and that refit is what surfaced the second effect on Boggs described above. Every consuming page — the home search, the full player browser, the Top 100 board, individual player pages, the team farm-system rankings, the Blind Spots page, and every other white paper quoting a specific player’s rank — was opened and reviewed by hand before this went out.