Touché /
Methode

Fragile-Pattern-Blocklist (master reference)

META/fragile-patterns.md

Master, human-readable reference. The skills grep this file. Every counter is checked against these patterns by a separate Kontrahent (opponent) pass before it ships — the builder never grades its own counter. A live own-goal kicked this off: a Bitcoin energy counter ("Bankensystem zieht mehr Strom") that dies instantly to "Banking wird von Milliarden genutzt, Bitcoin nicht" (true per-capita). That is the canonical aggregate-vs-per-capita failure.

The patterns

Pattern Trigger shape Default verdict
aggregate-vs-per-capita Compares absolute totals across differently-sized groups (no denominator). KILL — unless the per-capita math is run and the point still wins.
whataboutism "And what about X?" — deflection away from the claim. KILL — redirect to the claim.
survivorship "Hat alle Prognosen / jeden Crash überlebt" — only the survivors are visible. KILL — it hides the ones that failed.
cherry-picked-timeframe Selective start/end date that flatters the point. KILL — unless the full series is shown.
appeal-to-future-tech "Wird sich künftig lösen / die Technik wird besser." KILL — it concedes the present.
naked-aggregate-quote A bare quota on a young / small / recent group. KILL — use cohort / parity / trajectory instead.
is-ought Cites current law against a reform ("ist doch legal / steht so im Gesetz"). REFRAME as a burden-shift — do NOT kill the fact.

Canonical example (aggregate-vs-per-capita)

Counter: "Das Bankensystem zieht mehr Strom als Bitcoin." Rebuttal: "Banking wird von Milliarden genutzt, Bitcoin nicht." (true per-capita)

The absolute total has no denominator, so the rebuttal is true AND strong AND unanswerable with the Aktenlage → KILL the point. Run the per-capita math first; only ship the comparison if it still wins on a per-user basis.

Kill-vs-keep rule (stated once)

KILL a point only when its strongest rebuttal is:

  1. true — factually correct, and
  2. strong — it actually lands, not a quibble, and
  3. unanswerable with the Aktenlage — nothing in our evidence beats it.

All three must hold. If even one fails, do NOT kill — answer it (build the rebuttal-answer into the counter) or reframe it. A patched-but-still-weak point is an open own-goal; drop it (weglassen, per gungho) rather than ship it half-armored. Rank surviving counters by rebuttal-resistance and lead with the unbreakable one.

The is-ought exception

is-ought is the one pattern that is reframed, never killed. When an opponent cites current law against a reform ("das ist doch legal", "steht so im Gesetz"), they have confused IS with OUGHT: that something is currently legal says nothing about whether it should be. Do NOT drop the underlying fact. Instead, reframe as a burden-shift — the opponent now has to argue that the status quo is right, not merely that it exists. The fact stays; the frame moves.