Kalshi Politics — Mispricing Audit (v4, archetype-targeted)

No Mentions markets reached the survivor set. Tier-1 near-misses are logged below with their specific cut so the classifier and horizon can be judged against the record.

Trump publicly criticizes Kevin Warsh before Nov 1, 2026?

Prob: 72% NO pre-market → 76% final (blended toward 1,000–1,286-contract NO bids at 77–79¢). NO ask 81¢ > 76% → no bet.

Entry context (live, Sep 18 16:10Z): YES 19/20¢ · NO ask 81¢ · vol24h 3,736 · OI 13,973

Precedents consulted: No rows for KXDJTATTACKWARSH or multi-news-other beyond KXFOMCDISSENTCOUNT (not transferable). Cold derivation from live behaviour: markets opened Sep 16 16:45Z, Trump's Truth Social '1% or less' post landed hours later and the market did NOT resolve — the resolver's bar is criticism of Warsh himself, not of rates or 'the board'.

• CNBC Sep 16: Trump says he 'still has confidence' in Warsh, blames a 'hostile' board, told Warsh 'you might as well vote with the board because it's not going to matter'.
• Truth Social Sep 16 (via Yahoo/IBTimes): 'LOWER THE INTEREST RATES … AND FAST' — directed at the Fed, no reference to Warsh.
• FOMC Oct 27–28 sits inside the window (decision Oct 28 14:00 ET, 3 days before close); a second hike is the one live path to a by-name attack.
• Day-one tape: NOV01 printed 93¢ high on Sep 17 before settling 18–25¢; OCT01 leg now 3/5¢ — the crowd has already concluded nothing so far counts.

Reject reason: inside spread: P(NO) 76% vs NO ask 81¢ — no edge; YES side is a <40¢ tail; resolver bar ('disapproval directed at Warsh by name/title') is discretionary

Anthropic confirms IPO (S-1 effective / priced / ticker) before Nov 1, 2026?

Prob: 48% YES pre-market → 51% YES final (blended toward a 4.8k-vol24 book at Polymarket parity). NO-side P 49% vs NO ask 48¢: 1 pt, inside noise, and band-barred.

Entry context (live, Sep 18 16:10Z): YES 52/53¢ · NO ask 48¢ · vol24h 4,840 · OI 89,947

Precedents consulted: KXIPOANTHROPIC-DATE-26JUL01 resolved NO — confidential S-1 (Jun 1) met none of the three triggers. KXIPO-26-SPACEX: Kalshi settled on the Jun 11 pricing, not the May 20 public S-1 or roadshow. KXIPOANTHROPIC-DATE-26SEP01 lesson: a Kalshi rung ≈ the Polymarket 'trading by' rung dated one day later; >10 pts above the twin is headline residue.

• EDGAR full-text search Sep 18: zero S-1/S-1A filings by Anthropic since Jun 1 (only third-party filers mention it) — no public prospectus yet.
• Reuters via CNBC Sep 5: public S-1 late September, marketing mid-October at earliest, listing 'days before' the Nov 3 midterms; Bloomberg Sep 13: Nasdaq chosen; Sep 15: IPO 'still on'.
• Polymarket Sep 18: 'IPO by Oct 31' 54.5% ($890k vol); 'IPO in October' 53.5% — Kalshi NOV01 at 52/53 is at parity after the Sep 13 repricing (70¢ → 50¢ on 18.7k contracts).
• Trigger fires at pricing (evening before first trade), so a debut on Oct 29–30 pays YES and a post-midterm debut pays NO — a genuine timing coin flip.

Reject reason: 40–59¢ band requires HIGH conviction AND a Tier-1 process archetype (this is Tier-2 corporate-event); 1-pt paper edge vs Polymarket-parity book is noise; cluster already carries live NO legs (OCT01, OCT17)

Tesla/SpaceX definitive, binding merger agreement announced before Nov 1, 2026?

Prob: 4% YES pre-market → 6% final (blended toward a 10/11 book). P(NO) 94% vs NO ask 90¢: 4 pts, but ≥90¢ is a hard reject.

Entry context (live, Sep 18 16:10Z): YES 10/11¢ · NO ask 90¢ · vol24h 1,069 · OI 71,623

Precedents consulted: No rows for KXCOMPANYACTIONMERGER. Archetype rows (KXIPO/KXIPOANTHROPIC) confirm this resolver waits for the mechanical instrument, not the headline; rules_secondary here explicitly excludes Musk 'expression of intent, hypothetical, negotiation update'.

• Musk at All-In Summit (Benzinga, Sep 2026): 'who can imagine what action one might take' — intent language, expressly non-qualifying.
• SpaceX priced its IPO Jun 11, 2026 (SPCX); a definitive Tesla/SpaceX agreement within six weeks of a fresh listing has no reported process behind it.
• Kalshi ladder itself prices ~47% by May 2027 and 66% by 2028 — the crowd's mass is 2027, not October.

Reject reason: deep-favorite band: NO ask 90¢ (≥90¢ reject); YES side is a <40¢ tail

Bank of Canada holds (hike 0bps) at Oct 28, 2026 meeting?

Prob: 60% hold pre-market → 61% final (book is liquid and OIS-anchored; evidence is thin). YES ask 63¢ > 61% → no bet.

Entry context (live, Sep 18 16:10Z): YES 62/63¢ · YES ask 63¢ · vol24h 814 · OI 25,471

Precedents consulted: KXCBDECISIONCANADA Jun/Jul 2026: six-then-seven consecutive holds, two-sided guidance read as a hold tell. SEP-H0/H25 rows (Sep 2 hold, unanimous 35-economist poll): 'treat OIS as the market's view, not the Bank's signal; hike trigger is core-based (trim/median 1.9–2.0%)'. KXFEDDECISION-26JUL-H0: hawkish repricing into a live meeting was a fade — but that was a Fed that then held; the BoC's own trigger is now closer to firing.

• BNN Bloomberg Sep 18: swaps 'narrowly in favour of a hike' for Oct 28 (a coin flip); before Sep 2 they had 94% hold.
• Economists: RBC (Fan) hold; Desjardins (Bartlett) hold, first hike Q1 2027; Capital Economics (Brown) 'base case no hike in October, though likely a close call'.
• Fed hiked 25bp to 3.75–4.00% on Sep 16 (unanimous) — adds CAD/imported-inflation pressure; BoC gets Sept CPI (~Oct 20) and the BOS (~Oct 19) before the Oct 28 MPR decision.
• Kalshi H0 slid 89¢ (Sep 5) → 63¢ (Sep 18) and H25 rose 10¢ → 38¢, tracking OIS with a lag of days.

Reject reason: inside spread: P(hold) 61% vs YES ask 63¢ — no edge over an OIS-tracking book; no informational advantage over swaps on a macro coin flip

Bank of Canada hikes 25bps at Oct 28, 2026 meeting?

Prob: 39% pre-market → 38% final. Ask 38¢ = model; no bet and under the 40¢ floor regardless.

Entry context (live, Sep 18 16:10Z): YES 37/38¢ · YES ask 38¢ · vol24h 526 · OI 27,200

Precedents consulted: Same rows as H0. SEP-H0 edge note already recorded that the paper-edge side of this ladder sat under the 40¢ floor throughout Sep 2–17; still true.

• Same source set as H0; the >25bp leg (H26) is 0/1¢ and both cut legs 0/1¢ — the ladder is a clean two-outcome {hold, +25} split, as every BoC/Fed precedent row predicts.

Reject reason: <40¢ band (tails 0-for-8 lifetime); P(hike) 38% = YES ask 38¢ — zero edge; cluster runner-up to H0

Anthropic confirms IPO before Oct 17, 2026?

Prob: 95% NO pre-market → 95% final. Not eligible as a new pick.

Entry context (live, Sep 18 16:10Z): YES 4/8¢ · NO ask 96¢ · vol24h 1,292 · OI 19,684

Precedents consulted: KXIPOANTHROPIC-DATE-26JUL01 (NO), KXIPO-26-SPACEX (settled on pricing).

• EDGAR FTS Sep 18: no Anthropic S-1. Polymarket 'IPO by Oct 15' 1.65%.

Reject reason: already an active pick (EXISTING_PICKS, NO from Sep 9 @71¢); now also deep-favorite band (NO ask 96¢). Thesis intact: no public S-1 on EDGAR as of Sep 18, roadshow mid-October at earliest.

Position sizes and limits

EV for the record

Every line is a shadow-tracked reject. Positive-looking rows (Tesla/SpaceX NO +4, Anthropic NOV01 NO +1) are barred by band rules, not by EV — that is the design: EV-agnostic sizing, band-gated entry.

Cluster table (one leg per cluster)

Conviction exposure

Risk profile

Max loss $0. No correlated theses. Live prior picks (not re-underwritten here) remain: NO on Anthropic IPO OCT01/OCT17, NO on crypto market-structure bill, NO on FISA 702, NO on Hormuz normalization, YES on Suez weekly >270. Nothing in this run adds to those clusters.

Execution notes

• Nothing to execute. If the Warsh NOV01 NO ask drifts to ≤75¢ with no Warsh-directed remark, it re-enters the core band with ~1 pt of edge — still not enough; a real trigger would be a dated resolver clarification narrowing 'clear contextual reference'.
• Anthropic NOV01: a public S-1 flip dated before Sep 30 with a stated roadshow start ≤ Oct 19 would move P(YES) toward 65%; a flip after Oct 5 moves it toward 30%. Either would be a legitimate re-look — but the 40–59¢ band stays closed to a Tier-2 archetype regardless.
• Bank of Canada: re-check after the Oct 19 Business Outlook Survey and Oct 20 CPI; the H0 leg is the only side that can reach the 60–84¢ band with edge (it would need core CPI-trim/median to stay ≤2.0% while OIS holds a hike).