Skeptikolle

Track record – we claim in advance and grade ourselves in public afterwards

You probably came here to debunk one of our claims. Good. This page is built for exactly that: calibration, the full warning archive with its misses, and the list of claims we killed ourselves.

Kalibraatio, sanoimme X, toteutui Y

rises monotonically · misses in the count too

Riskipiste → poissaolon/virallisen tiedon toteuma

all players, two seasons · n = 15 581 ● measured

matala riskipiste5 %

keskitaso24 %

korkea52 %

The curve rises from 23 % to 65 % with the risk score; if it did not rise, the whole product would be worthless. That is why it is public.

Pre-flag-sweep, tapahtumataso (sama vamma = 1)

measured 16.7.2026 ● measured

pre-flag75 % (9/12)

varoitustaso63 % (12/19)

clear-base~17 %

Core numbers from the same data: the fitness concern showed before the official word2.2× (N = 15.581) · breakout before output 1.9× (n = 5 005) · uutinen × tapahtuma 4.77× (n = 511) ● measured

Varoitusarkisto

40 warnings · misses included and marked

#R24-001William Saliba (Arsenal), substituted on 31.08. already at min 4 → no game in the next✓ hit

#R24-002Jurriën Timber (Arsenal), substituted on 14.03. already at min 37 → no game in the next✓ hit

#R24-003Benjamin White (Arsenal), substituted on 13.12. already at min 30 → no game in the next✓ hit

#R24-004Benjamin White (Arsenal), substituted on 10.05. already at min 27 → no game in the next✓ hit

#R24-005Martin Ødegaard (Arsenal), substituted on 13.09. already at min 17 → no game in the next✓ hit

#R24-006Martin Ødegaard (Arsenal), substituted on 04.10. already at min 29 → no game in the next✓ hit

#R24-007Noni Madueke (Arsenal), substituted on 21.09. already at min 45 → no game in the next✓ hit

#R24-008Gabriel Martinelli Silva (Arsenal), substituted on 17.01. already at min 45 → no game in the next✓ hit

#R24-009Kai Havertz (Arsenal), substituted on 25.04. already at min 33 → no game in the next✓ hit

#R24-010Cristhian Mosquera (Arsenal), substituted on 28.09. already at min 45 → no game in the next✓ hit

#R24-011Cristhian Mosquera (Arsenal), substituted on 03.12. already at min 43 → no game in the next✓ hit

#R24-012Viktor Gyökeres (Arsenal), substituted on 01.11. already at min 45 → no game in the next✓ hit

#R24-013Matty Cash (Aston Villa), substituted on 04.03. already at min 45 → no game in the next✓ hit

#R24-014Tyrone Mings (Aston Villa), substituted on 28.09. already at min 31 → no game in the next✓ hit

#R24-015Youri Tielemans (Aston Villa), substituted on 13.09. already at min 45 → no game in the next✓ hit

#R24-016Emiliano Buendía Stati (Aston Villa), substituted on 26.10. already at min 28 → no game in the next✓ hit

#R24-017John McGinn (Aston Villa), substituted on 18.01. already at min 17 → no game in the next✓ hit

#R24-018Boubacar Kamara (Aston Villa), substituted on 23.08. already at min 15 → no game in the next✓ hit

#R24-019Amadou Onana (Aston Villa), substituted on 30.12. already at min 45 → no game in the next✓ hit

#R24-020Harvey Elliott (Aston Villa), substituted on 28.09. already at min 45 → no game in the next✓ hit

#R24-021Douglas Luiz Soares de Paulo (Aston Villa), substituted on 01.11. already at min 13 → no game in the next✓ hit

#R24-022Kyle Walker (Burnley), substituted on 21.02. already at min 45 → no game in the next✓ hit

#R24-023Maxime Estève (Burnley), substituted on 30.12. already at min 14 → no game in the next✓ hit

#R24-024Lucas Pires Silva (Burnley), substituted on 06.12. already at min 42 → no game in the next✓ hit

#R24-025Axel Tuanzebe (Burnley), substituted on 02.02. already at min 45 → no game in the next✓ hit

#R24-026Jacob Bruun Larsen (Burnley), substituted on 28.02. already at min 45 → no game in the next✓ hit

#R24-027Hannibal Mejbri (Burnley), substituted on 21.03. already at min 17 → no game in the next✓ hit

#R24-028Loum Tchaouna (Burnley), substituted on 03.12. already at min 45 → no game in the next✓ hit

#R24-029Adam Smith (Bournemouth), substituted on 30.08. already at min 25 → no game in the next✓ hit

#R24-030Tyler Adams (Bournemouth), substituted on 15.12. already at min 4 → no game in the next✓ hit

#R24-031David Brooks (Bournemouth), substituted on 28.02. already at min 45 → no game in the next✓ hit

#R24-032Ryan Christie (Bournemouth), substituted on 09.05. already at min 41 → no game in the next✓ hit

#R24-033Junior Kroupi (Bournemouth), substituted on 27.12. already at min 45 → no game in the next✓ hit

#R24-034Bafodé Diakité (Bournemouth), substituted on 27.12. already at min 45 → no game in the next✓ hit

#R24-035Rico Henry (Brentford), substituted on 03.03. already at min 20 → no game in the next✓ hit

#R24-036Aaron Hickey (Brentford), substituted on 21.02. already at min 44 → no game in the next✓ hit

#R24-037Kevin Schade (Brentford), substituted on 01.02. already at min 41 → no game in the next✓ hit

#R24-038Mikkel Damsgaard (Brentford), substituted on 25.01. already at min 36 → no game in the next✓ hit

#R24-039Antoni Milambo (Brentford), substituted on 17.08. already at min 45 → no game in the next✓ hit

#R24-040Vitaly Janelt (Brentford), substituted on 21.02. already at min 45 → no game in the next✓ hit

huti3 pre-flag cases out of 12 did not lead to an absence or an official notice (event level); included in every count, not hidden✗ huti

huti7 warning-level cases out of 19 did not materialise; the warning level is the weaker observation, and we say so out loud✗ huti

Early substitution → no game in the next · season 2024-25 ● measured· the warning id (#R24-NNN) travels with the shared card, so every claim can be traced back here.

Team hypotheses – how the summer numbers held

resolved hypothesis = outcome · the faded ones too

No team hypothesis resolved yet – hypotheses set in the summer resolve as the season runs, when a new measurement moves the state. This section also shows the faded and the dead, with the same weight.

What we do not claim: the graveyard of observations

dead observations in public · nobody else tells you this

📅

Calendar / fixture congestion

k. 10.7.2026

”Predicted everything, hit nothing.”

~1.1× ● measured

🏃

Raaka minuuttikuorma

k. 10.7.2026

”Died as a fatigue gauge, woke as a durability mark.”

inverse ● measured

📉

A run of three weak points matches

k. 16.7.2026

”Just regression to the mean.”

0.97× stratifioitu ● measured

A gravestone means: we measured it, it did not work, and we do not sell it. The “rising/falling” stamp comes only from underlying data, never from points runs.

All observations, in one view

one engine · many signal families

2.2×

before the official flag

Fitness observation

Accumulating early substitutions + news point to a shrinking role.

William Saliba (Arsenal), taken off as early as min 4 → no game next

● measured

Have your say🫣 through the fingers 24 %

1.9×

before the output arrives

Form / breakout

Expected goal involvements (xGI) lead the realised.

Yéremy Pino Santos, xGI 10.6 vs 4 realised ('due')

● measured

Have your say🧤 safe hands 39 %

4.77×

uutinen × tapahtuma

Cross-checked sources

Two observations weak alone become strong together.

each alone only ~1.6×

● measured

Have your say🧠 cool head 50 %

2.48

Arsenal

Squad risk (team level)

Player-level observations composed into a team.

avg. regulars under fitness concern / round

● measured

Have your say🫣 through the fingers 26 %

–

reactions on this page · dominant mood in the pulse

Community mood

The page's own soft mood number: community mood per story.

starting numbers ○ illustrative; the cold start is bridged with public-source sentiment (threshold 500 reactions/day); weight in the reading only once reliability is measured

○ live / rakentuu

Have your say🧠 cool head 45 %

This inventory used to live on the front page; now the front-page thread shows only the changes, and the full picture lives here.

Miten luemme

four principles, briefly

The rise is predictable, the fall is not, and we say so.When a player creates more than they convert (gap between created and returned > 2, output ≤ 1), the points level of their next starts rose in 54.7 % of cases (1.7×) and none fell, but n = 53 is a small sample and will grow with further seasons before the claim hardens. The weaker gap (1–2): a rise in 37 % (1.15×, n = 492).● measured

We do not draw a down-arrow the data does not carry.The points drop is not predicted for continuers – the drop happenspoissaolona, which the fitness path predicts (72 % / 3.6×, n = 1 086). That is why the player card has only ↗ (rise) and ⤵ (absence risk); there is no down-arrow, because the measurement does not support one.● measured

Every number carries its mark.● measured = real data or a measured history run; ○ illustrative = not yet measured. The card shows one number; the rest (sample sizes, seasons, confounds) live in the ●'s tooltip and on this page.

Misses are published right where the hits are.The warning archive shows both, and dead findings get a gravestone. Email is no place to embarrass yourself; this page is the place to prove honesty.

The house glossary, the only place these words appear

the terms live in the code, in this box and in the Analyst-level term layer; in stories and this page's body text we use plain language (warning · outcome · hit · observation · mark)

kuittia warning checked against the outcome: we said it in advance, and after the game what happened was logged, hit or not. Archive ids (#R24-001…) point to these rows.

signaalian observation or mark that stands out in the data, measured against history before it earns weight in a reading.

lippuofficial absence data, the league's own availability list.

kuvioa recurring situation in a player's rows, for example a run of early substitutions without an explanation.