How we check figures

How the figures are checked

Every figure keeps the definition its publisher used. Where a later, better source corrects an earlier value, the record uses the corrected value and keeps the change visible; where credible publications disagree, both values stay visible with the treatment applied. Read the corrections and published differences

Evidence status codes

CodeMeaning
CONFLICTunresolved contradiction shown explicitly
Ddestination booklet/list
Pprimary school/FOI/examinations source
Rreconstructed with assumptions
Ssecondary press/league/aggregator

Definitions used by the ledgers

Definition

Acceptances

Offers firmly accepted/finally accepted on the UCAS/apply-centre basis. Keep cycle-year convention explicit.

Definition

Applications

Pupils/application records submitted. Largest admissions quantity; not evidence of success.

Definition

Attribution lag

For a leaver in Y: admitted roughly Y−5, GCSE Y−2, UCAS Y−1. Credit a head only when the full relevant stewardship window is identified.

Definition

Destinations

Leaver-list/booklet placements. A list, not a rate unless a matching denominator is supplied.

Definition

Matriculations

Pupils who actually enrolled. May differ from acceptances due to deferral, conditions, and changes of plan.

Definition

Missingness

Use null/blanks; never delete an asymmetrical row or fabricate a comparator.

Definition

Offers

Conditional or unconditional offers made. A leading indicator; not final attendance.

Definition

Pandemic

CAG 2020 and TAG 2021 rows are quarantined from normal examined-year trend analysis.

Definition

Strict cycle

Applications/offers/final acceptances for one UCAS/apply-centre cycle. Quarantined from destinations/current-cycle offers.

Grade mappings

Grade mapping

GCSE

Old A*–G: A*, A*–A, A*–B. New 9–1: grade 9 has no exact old-scale equivalent; 9–8 is the nearest old-A* top-equivalent; 9–7 is A*/A-equivalent; A*–B requires grade 6.

Grade mapping

KCS internal IB headline

For the internal combined KCS headline only: IB 7→A*, 7–6→A*/A, 7–5→A*/B.

Grade mapping

PRE U

D1–D2 is A*-band; D1–D3 is the honest A*/A-equivalent; D1–M1 is the published/folded measures and is not honest A*/A.

Rules the presentation follows

  1. Every chart/filter must expose basis (offer, acceptance, matriculation, destination, exam entry), period, denominator, confidence, and source/version.
  2. Default comparisons must be metric-pure; cross-basis overlays require an explicit warning and different visual encoding.
  3. Show competing values rather than silently choosing when conflicts remain.
  4. Use corrected cells from the corrections table; never revive superseded headline figures.
  5. Keep v13 full-table rows available even when v14.6 elides them.
  6. Treat 2020 CAG and 2021 TAG as quarantined.
  7. Do not infer missing Cambridge splits or SPGS 2025 A*/A.
  8. For Winchester 2015, keep the coherent 443-entry original Pre-U table separate from the later 444-entry accounts version. For 2017/2019 D1–M1, primary raw-count variants control and later league candidates remain distinct. Keep the 2021 mixed A-level/Pre-U TAG measures out of both pure trends.

Which compilation controls

The source compilations are layered: a later companion does not overwrite a detailed annual table it merely elides.

OrderVersionControlsDoes not control
1v14.6seven-school companion synthesis, latest post-v14 corrections, Wayback KCS/Westminster destination anchorsall detailed annual tables that it elides; any later-secondary candidate that regresses against a v13.9 primary raw-count lock
2v13 consolidated readercomplete strict spines and full annual/detail tables elided in v14.6
3v13.9 master + addendumWinchester primary image/raw-count locks and analytical calculations
4v14.0–v14.2 addendaprospectus reconciliation, GCSE corrections, league-measures evidence
5legacy attachmentnothing where a later locked correction exists; retained for lineage and audit

Lineage of the compilation

  1. Legacy clean/standardised

    initial Winchester/Eton/St Paul's/KCS compilation; now superseded cell-by-cell

  2. Reader consolidation

    consolidated readers, KCS layers, audit/cross-check additions

  3. Source-recovery deltas

    archive routes and historical research leads; most deltas explicitly made no CSV change

  4. Analytical milestone

    long-arc rate analysis, decline decomposition, two-measures computation, like-for-like comparison

  5. Attribution

    headmaster stewardship ledger and lag model

  6. Primary lock

    Winchester 2018/2019 image tables, 2019 GCSE correction, pandemic quarantine, legacy league benchmarks

  7. Correction addenda

    prospectus offer/matriculation reconciliation; 2016/2019 GCSE locks; third-party D1–M1 league-measures exhibit

  8. Seven-school companion

    Westminster/KCS/SPGS/Wycombe integration, full GCSE repairs, Wayback destination locks; v14.6 is latest but not sufficient alone

Oxford and Cambridge records

Acceptances or admissionsOxford labels vary between Accepts and students admitted; Cambridge acceptance means unconditional firm after confirmation. Never label this as offer rate.
Applicationssubmitted completed applications counted by the named source population
Cycle mapping evidenceOxford FOI/202505/554 explicitly says UCAS Cycle 2025 is for entry in 2025 or deferred 2026 (and likewise 2022-2024). Cambridge cycle reports explicitly use the intended start year. An archived KCS January 2013 claim exactly matches Oxford and Cambridge cycle-2013 offer rows. Do not add one to cycle_year.
Cycle yearuniversity admissions cycle named for intended entry year; Cambridge reports explicitly say applications begin in September of the previous calendar year
Destinationschool leaver's eventual university; a different unit from applications/offers/acceptances
Entry yearintended start year; deferred-entry applicants may start one year later
Offersconditional or unconditional offers unless the source states otherwise
SuppressionNever midpoint. Oxford <3/blank means bounded unknown below 3; Cambridge older releases use <4 or <3; unidentified Cambridge 2025 source reportedly rounds 1–4 up to 5.

US and overseas university records

  • Aggregate rows are explicitly flagged; never sum them with institution rows.
  • Offers, acceptances, firm places, destinations and presence-only claims remain separate metric types.
  • Null institution counts mean presence-only, never infer count=1.
  • US-only, North-American and all-overseas panels remain distinct.
  • Conflicting source versions have conflict_group and canonical_for_analysis=false where unresolved.
Source evidenceOpen full record ↗