What could have been known · a triumph with its error bars intact

The Mass Before the Particle

For nine years, precision measurements kept describing a particle that no detector had established. The center wandered. The interval breathed. The direct searches closed in. One narrow window stayed open.

First, rebuild one prediction exactly as it was published. Only then is there a trajectory.

Anchor · Gfitter release, 3 May 2012 · LoopFest slide 9

The last indirect profile

Gfitter's standard fit used the precision electroweak observables but omitted all direct Higgs searches. Slide 9 published MH = 94+25-22 GeV and a 152 GeV one-sided 95% upper bound.

...minimum plateau midpoint
...delta chi-square 1, lower
...delta chi-square 1, upper
...delta chi-square 3.84, upper

Computed live by linear interpolation between factual vector coordinates extracted from the official EPS. The 94 GeV mode is a rounded published value; the EPS has a flat zero-height segment, so this page reports its midpoint. The lower 95% crossing is not used because Gfitter published a one-sided upper bound.

The timestamped replay · 2003 to 4 July 2012

Move the public record

May 2012

40 GeV100160220280 GeV

Blue is the published indirect 68% interval. Its pale extension ends at the published one-sided 95% upper bound. Red hatching is the union of the latest observed direct exclusions available on that date. Endpoints are the papers' displayed precision, so tiny gaps can be rounding artifacts.

Commit before the mass is shown

Choose a band first. The outcome remains hidden.

What “the prediction” meant each year

These are not points sampled from a modern history graphic. Each row is the published high-Q² or standard indirect fit in that release. The asymmetric interval is 68%. The last column is the one-sided 95% upper limit, with the release's theory treatment included.

releasemode68% interval95% upper125.09 statusnamed location

Depth layer · the answer changes when the question changes

A point miss, an interval success, a direct survivor

...final point-estimate error
95% onlywhere 125.09 lands indirectly
...direct survivor windows at cutoff
tiefrozen rule versus historical fitters

The frozen rule is not a cleverer retrospective model. It is the fitters' own published indirect-only profile rule. The historical and frozen arms therefore tie. As a point prediction, the final 94 GeV mode loses badly to the public direct excesses already near 124 to 126 GeV. As an uncertainty forecast, it succeeds only at 95%.

The triumph is narrower and more substantial: loop-level measurements repeatedly required a light Standard Model Higgs, and the final indirect profile did not exclude the mass later measured directly. Independent searches then carved the public mass line from below and above until 125.09 GeV sat in one of the surviving windows.

The fits that look prescient are the ones that had already been told

There is a version of this story in which the prediction was a bullseye, and it is worth seeing exactly how it is made. The electroweak fit was published in two forms. The standard fit used precision measurements only: in Gfitter's July 2011 release it gave 96 GeV, with a 68% band of about +31 and -24 and a 95% upper limit near 169 GeV. The complete fit folded the direct search likelihoods in as well, and it gave 120 GeV, +12 and -5, upper limit about 143. The second one sits almost exactly on 125.

It is not a prediction. It already contains the searches that were closing on the answer, so quoting it as an indirect anticipation of the measured mass is circular: the searches told the fit roughly where the particle was, and the fit then told us roughly where the particle was. The honest indirect number is the first one, and the first one points low. The same choice moves the limit as well: adding the LEP lower bound to the 2010 profile pushed the 95% upper limit from 158 GeV to 185, because it removed likelihood from the bottom of the range rather than from the top.

The band itself was never a probability distribution over the Higgs mass. It was a profile of delta chi-squared, conditional on the Standard Model being the right theory, after fitting nuisance parameters including the top mass, the Z mass, the strong coupling and the hadronic contribution to the running electromagnetic coupling. Read as "where the Standard Model would have to put the Higgs, given everything else we had measured", it did its job. Read as "where the Higgs is", it was never that kind of statement.

A dating trap worth naming, because this page could have fallen into it. One of the fit results used here was published online before the discovery seminar, but the matching arXiv revision was uploaded afterwards. Citing the arXiv revision date would have dated this page's own information set with a document released after its cutoff, which is precisely the failure the firewall exists to prevent. Every release below is dated by its journal or collaboration publication, never by an arXiv revision.

What the record could not have told them

The precision record had only logarithmic, Standard Model-conditional sensitivity to the Higgs mass. It could not supply the unblinded 8 TeV ATLAS and CMS likelihoods released at the seminar. It could not establish that a new state existed, had spin zero, or had Standard Model couplings. It could not know the later combined on-shell mass, 125.09 GeV. No more diligent reading of the pre-cutoff fit releases could create those direct-event quantities.

Counterfactual limit. This page scores the frozen rule against the outcome that actually followed. It does not claim to know what any unchosen experimental or publication path would have produced.

The check

0used
0later than cutoff
0excluded, stated cause
0undated

Computed live: EPS coordinate transforms, profile crossings, band membership, interval unions, surviving windows, firewall partition, and the planted-future-record control.

Source transcriptions: the ten annual modes, asymmetric 68% bounds, 95% upper limits, direct-search endpoints, release timestamps, and 125.09 GeV outcome.

Uncertainties: each annual fit uses its authors' experimental and theory treatment. The anchor's 68% endpoints use delta chi-square 1; the one-sided 95% upper bound uses 3.84. Direct limits are observed 95% CL intervals at published endpoint precision. The later mass has 0.21 GeV statistical and 0.11 GeV systematic uncertainty.

Free choices: archive begins with the continuous annual LEPEWWG sequence in 2003; mass display domain is 40 to 280 GeV; direct exclusions are latest-per-experiment rather than accumulated superseded curves; boundary membership is inclusive; the fit mode is the midpoint of a digitized zero-height plateau.

Every archive record and its bucket
bucketdaterecordwhy
Sources and scope

Scope is the complete timestamped annual indirect release sequence from December 2003 through the discovery cutoff, joined to a selected ledger of observed direct-search combinations. It is not the complete history of electroweak fits before 2003, nor every channel-level direct search. One 2011 joint-conference combination is visibly excluded because this build did not recover a primary publication-time record with stable numerical endpoints. The undated Gfitter history graphic is counted but never admitted.