Kalshi Politics — Mispricing Audit (Anchor, NO-side structural)

Report date 2026-09-30 · strategy=Anchor · audit_version=anchor · horizon 45d · capital $2,500 · category focus: Politics (any allowlisted series in scope) · universe: targets.json generated 2026-09-30T16:37Z (120 rows / 115 unique series, min OI 5,000)

1pick (NO side)

23logged rejects (5 Stage-2, 18 Stage-0)

$28.56capital deployed (1.1%)

$2,471.44cash held (98.9%)

1. Summary

Anchor only takes the side the 552-pick record supports: NO on event markets, in the 60–89¢ effective-entry band, with a thesis whose only losing path is a new discrete event. This run's 45-day window left a thin slate: 867 open markets across the 115 allowlisted series, 94 inside the horizon, 6 through the mechanical cuts, 3 real Stage-2 candidates, and one pick. The single pick is NO on the Anthropic IPO “before Nov 1” rung at 84¢, sized to a depth-capped 34 contracts (about 1% of capital). Cash is the position this week: 98.9% held.

Disclosure. This audit was produced by an Anthropic model. The Anthropic IPO thesis rests only on public sources named below (EDGAR, WSJ via PYMNTS/Investing.com, Reuters via CNBC, Fortune, Yahoo Finance) and Kalshi's own rules and precedents; no non-public information was used or available.

2. Universe funnel (Stage 0)

Step	Markets	Note
Open markets in allowlisted series (public API sweep)	867	115 unique series, all tiers
Close or expected expiration within 45 days	94	most Tier-1 departure/enact markets close Dec 31 or later and are out of horizon
Cut: live volume24h < 500	70	largest bucket; near misses were Hegseth DEC01 (454), Zelenskyy OCT01 (436), CA Voter ID (267)
Cut: YES ask ≤ 5¢ (deep favorites)	12	McConnell/Pelosi resign, Impeach JAN01, AGI Q3, Warsh, Ebola OCT26 rungs
Cut: spread > 5¢	3	OpenAI personal agent (2 rungs), US-China AI hotline OCT01
Cut: close > 1h after expected expiration	3	CA wealth tax, Impeach MAR01/JAN28 (API expected_expiration set to 2026-11-03)
Stage-2 candidates	6	Anthropic IPO NOV01, diesel export ban, BoC OCT H25 + H0, Iran deal NOV, Tesla/SpaceX NOV01

Mentions markets, cumulative-count and level markets: none survived to Stage 2 (the Warsh “publicly criticize” market is logged as a Mentions-style Stage-0 reject). Nothing in $EXISTING_PICKS (KXCRYPTOSTRUCTURE-26JAN-OCT) is in the allowlist, so no collision.

3. Pick

BUY NO @ 84¢mechanism: no-vehicleconviction: MEDIUMcorporate-eventcluster: anthropic-ipo-timing

`KXIPOANTHROPIC-DATE-26NOV01` — When will Anthropic officially announce an IPO? · Before Nov 1, 2026

YES condition. For YES to pay, the SEC must declare Anthropic's S-1 effective, or the offering must be priced, or a securities exchange must assign it a ticker, before Nov 1, 2026 04:59Z, resolved by Kalshi against those three mechanical triggers (rules_secondary; no dated clarifications).

Thesis. No vehicle: no public S-1 exists on EDGAR as of Sep 30, so the SEC's 15-day public-filing clock before any roadshow has not started; WSJ (Sep 18) reports a November debut built around Q3 numbers, and none of Kalshi's three triggers (S-1 effective, priced, ticker assigned) can be reached before Oct 31 without an unreported acceleration.

Prob: 8% pre-market → 12% final. Pre-market number committed before opening the book (notes file 16:43Z). The blend moved up, not down: the leaked draft shows a print-ready document, so a public flip in early October is plausible enough to keep ~10 pts on the tail.

Vehicle / calendar / already-happened.

Vehicle: none. EDGAR full-text search for all forms mentioning Anthropic, Jun 1–Sep 30, returns only third-party filers (Augurey Ventures series funds, ETF trusts); no S-1 or S-1/A by Anthropic. Yahoo Finance (Sep 25): “As of September 26, the SEC EDGAR database contains no public Anthropic filing … A prospectus that has not yet appeared in EDGAR cannot support an October roadshow.”
Calendar: the SEC's non-public review policy requires the registration statement to be public at least 15 days before the roadshow. From a public flip on Oct 1 the earliest roadshow start is Oct 16; SpaceX (precedent KXIPO-26-SPACEX) took 22 days from public S-1 (May 20) to pricing (Jun 11). WSJ (Sep 18, via PYMNTS and Investing.com): Anthropic “plans to launch its initial public offering in November … chosen to allow Anthropic to share its third-quarter financials”; “timing could change.” Fortune (Sep 29): the IPO “after several delays, could occur after the U.S. midterm elections in November.”
Already happened? No. The Sep 28–29 “prospectus” coverage (Reuters via CNBC; TechCrunch; Fortune “leaked IPO prospectus”) was a leaked draft reviewed by Reuters, not a filing. The rung spiked 6¢ → 27¢ (Sep 29) and 41¢ intraday (Sep 30) on those headlines and has faded to 15/16¢ as EDGAR stayed empty. That is the 48h move, and its catalyst is identified and non-qualifying.

Resolver profile. Kalshi resolves on three mechanical triggers. Precedents JUL01/JUN01: a confidential draft submission and a company press release saying it “filed to go public” are non-events. KXIPO-26-SPACEX: a proposed symbol in an S-1 is not “a ticker assigned”; Kalshi settled on pricing. Residual risk that a Nasdaq listing-approval notice is read as ticker assignment a day or two before pricing: priced at 2 pts. Total deductions 2 pts (well under the 15-pt reject line).

Cleanest tail. Anthropic flips the S-1 public in the first week of October and runs a compressed, SpaceX-speed roadshow to price Oct 27–30 ahead of the Nov 3 midterms. This reverses the reported plan (November, Q3 numbers) and needs a decision the company has not signalled, so it is a new discrete event rather than the current trend (delay) continuing. Priced at ~10 pts; the pick survives with it in.

Entry context (16:38Z). YES 16/17¢ · NO ask 84¢ with 103 contracts visible (YES bids 0.16×103.65); next NO asks 85¢ (40), 86¢ (150), 87¢ (750) · volume24h 7,909 (specialist-dense: >5,000) · OI 97,731 · 30-day range 87¢ → 5¢ → 16¢.

Precedents consulted: KXIPOANTHROPIC-DATE-26JUL01 (NO), -26JUN01 (NO), -26SEP01 (NO, re-verified this run), KXIPO-26-SPACEX (Jun 15 settlement on pricing). Quoted lesson (JUL01): “treat ‘filed’ and ‘confirmed’ as separated by months of process.”

4. Position sizes and limits

Ticker	Side	Limit	Contracts	Cost (= max loss)	Max win	Binding cap
`KXIPOANTHROPIC-DATE-26NOV01`	NO	84¢	34	$28.56	$5.44	Depth: 1/3 of 103 contracts visible at ≤84¢ (flat 10% cap would allow 297)
Total deployed				$28.56 (1.1%)	$5.44	Reserve ≥10% satisfied: $2,471.44 cash

Sizing is flat and EV-agnostic (MEDIUM = HIGH = 10% of capital in cost, halved in the 85–89¢ band). 84¢ sits in the core 60–84¢ band, so full size applies before the depth cap. Do not chase to 85–87¢ for size: the 85–89¢ band halves the allowance and the incremental depth (40 + 150 contracts) does not change the picture.

5. EV for the record

Ticker	Side	Price	P(YES) final	Fair (side)	Edge	EV / contract	EV %
`KXIPOANTHROPIC-DATE-26NOV01`	NO	84¢	12%	88¢	+4¢	+4.0¢	+4.8%

EV = 0.88×16 − 0.12×84 = +4.0¢ on 84¢ at risk. Recorded, not used for sizing.

6. Cluster table

Cluster	Legs considered	Deployed	Runner-ups / notes
anthropic-ipo-timing	NOV01 (NO), OCT24 (NO, 98¢)	NOV01 NO	OCT24 is deep-favorite band; DEC01/JAN01 rungs close outside the 45-day horizon
boc-oct-2026	H25 (NO), H0 (YES)	none	both rejected: swaps-implied 44% hike vs Kalshi 34–36¢ makes NO-H25 and YES-H0 negative-edge
diesel-export-ban	KXDIESELEXPORTBAN-26 (NO)	none	live vehicle; no edge inside a 4¢ spread
us-iran-nuclear-deal	NOV (NO)	none	deep-favorite band (95¢)
tesla-spacex-merger	NOV01 (NO)	none	deep-favorite band (92¢), 3.6 contracts at touch

7. Risk profile

Concentration: one position, $28.56 at risk (1.1% of capital); worst case is a full loss of that amount if Anthropic prices before Oct 31.
Correlation: none with the existing live pick (crypto market-structure bill NO).
Event path: the position is short a single discrete acceleration event. Watch EDGAR for a public S-1/S-1A: a public flip before Oct 9 raises P(YES) toward 25–30% and the pick should be re-underwritten; a flip after Oct 12 makes an October pricing mechanically implausible under the 15-day rule.
Resolver: 2 pts for an early ticker-assignment read; rules are mechanical and precedent-tested.
Liquidity: exit via the 3,335-contract NO bid at 83¢ costs ~1¢; the book is deep on the NO-bid side and thin on the NO-ask side.

8. Execution notes

Limit BUY NO 84¢ × 34 on KXIPOANTHROPIC-DATE-26NOV01. Marketable against the 103 contracts at the touch; do not lift 85¢+ for size.
Slippage ≤ 3¢ satisfied (0¢ at 34 contracts). Price-level structure is tapered deci-cent; 84¢ is a valid 1¢ tick.
Cancel-and-re-underwrite if a public Anthropic S-1 appears on EDGAR before fill, or if YES trades ≥ 22¢ on volume (implies a dated roadshow report).
Cash: $2,471.44 (98.9%) held. An efficient week; 70–100% cash is the expected Anchor outcome.

9. Stage-2 rejects (side, price, probability, reason)

Market	Archetype	Side	Price	P(YES)	Cluster	Reason
`KXDIESELEXPORTBAN-26` US diesel export ban in effect before Nov 3, 2026?	multi-news-other	NO	83¢	22%	diesel-export-ban	inside spread — live vehicle, not status quo: Trump Sep 22 ('let's not send out the diesel') and Sep 27 ('very seriously' looking at a ban); a WH official denied a ban Sep 23, Sec. Wright opposes a blanket ban, Sen. Cruz relayed 'no ban' Sep 26. Prob 30% pre-market -> 22% final vs 17-21¢ market: no NO edge at 83¢ across a 4¢ spread; 'ban' vs partial restriction adds resolver risk.
`KXCBDECISIONCANADA-26OCT-H25` Bank of Canada hikes 25bps on Oct 28, 2026?	scheduled-process	NO	66¢	44%	boc-oct-2026	inside spread — CORRA OIS implies 44% hike / 56% hold (BlueGamma, Sep 29) vs Kalshi 34-36¢; NO at 66¢ pays ~10¢ above swaps-implied fair. Precedent 26SEP-H0: 'the swaps print, not the economist survey, is the reference for the no-trade band on this series'. Tail is the energy-inflation trend continuing (continuation tail).
`KXCBDECISIONCANADA-26OCT-H0` Bank of Canada holds (0bps) on Oct 28, 2026?	scheduled-process	YES	64¢	56%	boc-oct-2026	yes-side thesis — a hold is a macro decision forecast, neither already-happened nor a process-in-motion step the resolver pays on; YES at 64¢ vs swaps-implied 56% is also negative edge.
`KXUSAIRANAGREEMENT-27-26NOV` US agrees to a new Iran nuclear deal before Nov 1, 2026?	scheduled-process	NO	95¢	4%	us-iran-nuclear-deal	deep-favorite band — NO costs 95¢ (max win 5%). Thesis is sound (06/17/26 clarification: the Islamabad MoU fails criterion 1; precedent 27-26JUL) but the band excludes it.
`KXCOMPANYACTIONMERGER-27-26NOV01` Tesla/SpaceX definitive combination agreement before Nov 1, 2026?	corporate-event	NO	92¢	5%	tesla-spacex-merger	deep-favorite band — NO costs 92¢; 3.6 contracts at the touch, unfillable at size.

Diligence notes on the Stage-2 rejects

Diesel export ban (NO @ 83¢, 30% → 22%). YES condition: a Source Agency must report that a US ban on diesel exports has gone into effect before Nov 3. Anchor-shaped (a new action by the executive), but there is a live vehicle: Trump Sep 22 (“I've said let's not send out the diesel”), Bessent examining “a full or partial ban”, a White House official denying a ban on Sep 23 (OilPrice, Sep 23), Sen. Cruz relaying “no ban” to refiners Sep 26, then Trump again “very seriously” looking at it Sep 27 (Bloomberg) and Sep 28 (CNBC); Sec. Wright opposes a blanket ban. Diesel $6.38/gal (EIA, week of Sep 28), midterms in five weeks. That is not a status-quo tail with a clean discrete trigger; it is an open political decision, and 22% vs a 17–21¢ market leaves nothing across the spread. Definitional risk (“ban” vs partial restriction or destination-specific limits) adds ~5 pts.
Bank of Canada H25 (NO @ 66¢) / H0 (YES @ 64¢), 44% hike. Precedents 26JUN/26JUL (12 rows) and the Sep 28 note on 26SEP-H0 are explicit: this ladder trades the hike leg 10–15 pts below CORRA swaps, so NO on the hike overpays even inside the Anchor band. BlueGamma (CORRA OIS, as of Sep 29): hold 56% / hike 44% / +11bp; BNN Bloomberg (Sep 18) put swaps near 50% after the Fed's own hike. Sell-side (RBC, Desjardins, Capital) still says hold; Scotia/National say hike. Catalysts before Oct 28: Sep jobs (~Oct 9), BOS (~Oct 19), Sep CPI (~Oct 20). The NO tail is energy inflation continuing, a continuation tail. YES-H0 is a macro forecast and fails the side rule.
Iran deal NOV (NO @ 95¢, 4%). Thesis intact (06/17/26 clarification rules out the Islamabad MoU; precedent 27-26JUL) but 95¢ is outside the band by rule.
Tesla/SpaceX NOV01 (NO @ 92¢, 5%). Deep-favorite band and 3.6 contracts at the touch.

10. Stage-0 rejects logged (mechanical cuts with volume24h ≥ 500)

Market	Archetype	Side	Price	P(YES)	Cluster	Reason
`KXOAIPERSONALAGENT-27JAN01-26OCT01` OpenAI releases a personal AI agent before Oct 1, 2026?	multi-news-other	NO	72¢	32%	openai-personal-agent	spread > 5¢ (28/37) — Stage-0 cut; also inside 7d
`KXUSCHINAAIHOTLINE-27-26OCT01` US-China AI incident hotline agreed before Oct 1, 2026?	multi-news-other	YES	86¢	83%	us-china-ai-hotline	spread > 5¢ (80/86) — Stage-0 cut; also inside 7d
`KXOAIPERSONALAGENT-27JAN01-26NOV01` OpenAI releases a personal AI agent before Nov 1, 2026?	multi-news-other	NO	59¢	46%	openai-personal-agent	spread > 5¢ (41/52) — Stage-0 cut
`KXCAWEALTHTAX-26` California one-time billionaire tax initiative passes?	confirmation/enact	NO	68¢	32%	ca-2026-ballot-measures	close after expiration — close_time 2027-11-03 vs expected_expiration 2026-11-03 (Stage-0 cut); election-day data shape, prob = market mid
`KXIMPEACH-27-MAR01` President impeached before Mar 1, 2027?	legal-action	NO	72¢	28%	trump-impeachment	close after expiration — close 2027-03-01 vs expected_expiration 2026-11-03 (Stage-0 cut); prob = market mid
`KXIMPEACH-28-JAN01` President impeached before Jan 1, 2028?	legal-action	YES	67¢	66%	trump-impeachment	close after expiration — close 2028-01-01 vs expected_expiration 2026-11-03 (Stage-0 cut); prob = market mid
`KXRETIREMM-26` Mitch McConnell resigns early?	departure/ouster	NO	97¢	3%	mcconnell-resign	deep-favorite band — YES ask 3¢, NO costs 97¢ (Stage-0 cut)
`KXRETIREPELOSI-26` Nancy Pelosi resigns early?	departure/ouster	NO	97¢	3%	pelosi-resign	deep-favorite band — YES ask 3¢, NO costs 97¢ (Stage-0 cut)
`KXIMPEACH-27-JAN01` President impeached before Jan 1, 2027?	legal-action	NO	98¢	3%	trump-impeachment	deep-favorite band — YES ask 4.4¢ (Stage-0 cut)
`KXAGICO-COMP-26Q3` Any company announces AGI in Q3 2026?	multi-news-other	NO	100¢	1%	agi-announce	deep-favorite band — YES 0/1¢ (Stage-0 cut); also inside 7d
`KXHEGSETHOUT-26AUG-26NOV01` Hegseth out as SecDef before Nov 1, 2026?	multi-news-other	NO	97¢	4%	hegseth-out	deep-favorite band — YES ask 4¢ (Stage-0 cut); API expected_expiration also inside 7d
`KXDJTATTACKWARSH-26SEP-26OCT01` Trump publicly criticizes Warsh before Oct 1, 2026?	multi-news-other	NO	100¢	1%	trump-warsh	inside 7d; also deep-favorite band (YES 0/1¢) and a Mentions-style market (Stage-0 cut)
`KXIPOANTHROPIC-DATE-26OCT24` Anthropic confirms an IPO before Oct 24, 2026?	corporate-event	NO	98¢	1%	anthropic-ipo-timing	deep-favorite band — YES 2/3¢ (Stage-0 cut); same thesis as the NOV01 pick, no payout left
`KXIPOOPENAI-26OCT01` OpenAI IPO before Oct 1, 2026?	corporate-event	NO	100¢	1%	openai-ipo	inside 7d; deep-favorite band (Stage-0 cut)
`KXEBOLACOUNTRY-OCT26-RWA` Human Ebola case reported in Rwanda before Oct 1, 2026?	multi-news-other	NO	100¢	2%	ebola-oct26	inside 7d (Stage-0 cut); YES 0/5¢
`KXEBOLACOUNTRY-OCT26-ZAMB` Human Ebola case reported in Zambia before Oct 1, 2026?	multi-news-other	NO	100¢	2%	ebola-oct26	inside 7d (Stage-0 cut); YES 0/5¢
`KXEBOLACOUNTRY-OCT26-KEN` Human Ebola case reported in Kenya before Oct 1, 2026?	multi-news-other	NO	100¢	2%	ebola-oct26	inside 7d (Stage-0 cut); YES 0/5¢
`KXEBOLACOUNTRY-OCT26-ETHI` Human Ebola case reported in Ethiopia before Oct 1, 2026?	multi-news-other	NO	100¢	1%	ebola-oct26	inside 7d (Stage-0 cut); YES 0/3¢

For Stage-0 rejects no view was formed; P(YES) is the market mid at the sweep. A further 70 in-horizon markets were cut for volume24h < 500 and are not listed.

11. Precedents consulted

KXIPOANTHROPIC (3 rows + KXIPO SpaceX row): JUL01 NO, JUN01 NO, SEP01 NO. Lessons quoted above. One row written back to precedents.json for SEP01 (leaked-prospectus headlines are not filing events; NOV01 tape 53 → 6 → 27 → 16¢).
KXCBDECISIONCANADA (12 rows): JUN/JUL/SEP ladders; the SEP-H0 and SEP-H25 notes carry the swaps-vs-Kalshi reference used to reject the October legs.
KXUSAIRANAGREEMENT (2 rows): 27-26JUN NO, 27-26JUL NO; definitional two-prong test.
KXDIESELEXPORTBAN, KXCOMPANYACTIONMERGER: nothing in the store; derived cold from live rules_secondary.

12. Sources

Kalshi public API: /markets (rules_primary/secondary, *_fp fields), /markets/{ticker}/orderbook, /series/KXIPOANTHROPIC/markets/KXIPOANTHROPIC-DATE-26NOV01/candlesticks (period 1440, Aug 30–Sep 30).
SEC EDGAR full-text search (efts.sec.gov), queries “Anthropic” all forms Sep 20–30 and S-1/S-1A Jun 1–Sep 30, run 2026-09-30; data.sec.gov submissions for Anthropic-named CIKs (all third-party feeder funds).
PYMNTS, Sep 18, 2026, citing WSJ: “plans to launch its initial public offering in November … timing could change.” Investing.com, Sep 18, 2026, citing WSJ: “targeting a November debut … pushing back its prospective timeline past an expected October target.”
Yahoo Finance, Sep 25, 2026: “The Anthropic S-1 That Was Expected After Labor Day Still Hasn't Arrived.” Fortune, Sep 29, 2026: “Anthropic's leaked IPO prospectus …” CNBC/Reuters Sep 28 and TechCrunch Sep 28 (prospectus contents). CNBC/Reuters Sep 5 (mid-October marketing timeline, pre-delay).
OilPrice.com, Sep 23, 2026: White House official denies diesel export ban; Wright quotes. CNBC Sep 28 and Bloomberg Sep 27: Trump “very seriously” considering. Gokhshtein Sep 26 (Cruz relay). EIA weekly diesel price.
BlueGamma Bank of Canada forecast page (CORRA OIS, as of Sep 29, 2026). BNN Bloomberg Sep 18, 2026. Money.ca / MoneySense October 2026 previews (sell-side split).

Anchor methodology notes: probability committed before price (16:43Z notes file); side rule, continuous-number, continuation-tail and rules-reading checks applied in order; one leg per cluster; flat sizing with depth cap; cash is a position.