# Kalshi Politics — Mispricing Audit (v4, archetype-targeted) — 2026-08-17

Date: 2026-08-17
Source: https://kalshi-audits.pages.dev/2026-08-17-kalshi-politics-mispricing-audit-v4

---
# Kalshi Politics — Mispricing Audit v4 · archetype-targeted

Report date 2026-08-17 · Horizon 45 days (to Oct 1, 2026) · Capital $2,500 · Universe: Stage-0 allowlist (targets.json, 60 series, generated 2026-08-17 16:06 UTC) · Advisory category: Politics

## Executive summary

**2 picks, 12 logged rejects, $254.54 deployed (10.2%), $2,245.46 cash (89.8%).**
The 45-day Tier-1 slate is almost entirely untradeable this week: of 37 Tier-1 markets inside the horizon, every one failed the mechanical cuts — dead books (vol24h 0), tail prices (ask ≤ 5¢), or both. The only near-live Tier-1 book (KXGOVTFULLFUND Sept rung, 56/62) trades 205 contracts a day on a 6¢ spread and is logged as a liquidity-floor reject. The deployable slate was five Tier-2 markets in three series. Both picks are precedent-fed, in the 60–84¢ core band, MEDIUM conviction, one leg per cluster: a telegraphed BoJ September hike at 75¢ and the NO side of an Anthropic IPO-confirmation-before-Oct-1 at ≤79¢. The Trump-Bedminster exact-trip-count cluster — the week's most tempting board — is rejected in full: the playable side sits below the 40¢ band floor, the priced side contradicts the primary-source schedule record, and the thesis is a pace-of-count bet correlated with the live KXLAGODAYS pick. A ~90%-cash week is the correct output of an efficient board, not a failure to find edge.

## Picks

### 1. Bank of Japan hikes 25bp at the September 2026 meeting BUY YES MEDIUM scheduled-process

KXCBDECISIONJAPAN-26SEP17-H25 · closes 2026-09-18 02:29 UTC · cluster: boj-sep-2026-hike · Tier 2

Prob (pre → final)80% → 79%
Entry (limit)75¢
Edge+4¢
Contracts233
Max cost$174.75

**Thesis.** The BoJ raises the policy rate 25bp to 1.25% on Sep 17–18. This is a telegraphed institutional move, not a data drift: Reuters (Aug 14, three sources) reported the Board is eyeing a September hike and considering a faster-than-twice-a-year pace — "an early rate hike has come into sight"; the July Summary of Opinions (Aug 10) flagged upside price risks and a possible faster pace; the July 31 statement warned core inflation could exceed the 2% target; and the yen stayed weak despite July's rare joint Japan–US intervention.

Precedent: KXCBDECISIONJAPAN-26JUN15-H25 resolved YES (7–1) with swaps at only ~72.6% — the store's lesson: "BoJ hold legs must be priced off the signalling, not off the bank's no-show history." KXCBDECISIONJAPAN-26JUL30-HOLD (verified this run, API-verbatim) settled YES with hawkish guidance that set up September. Counter-precedent weighed: KXFEDDECISION-26JUL-H25 — a hawkish repricing into the Fed faded to a 9–3 hold. The distinction: the Fed repricing was market-inferred from oil and CPI; this BoJ repricing is leak-driven — the institution itself speaking through its standard channel.
Resolver profile & deductions. Resolver: Kalshi on the BoJ's official policy-rate announcement; primary rate only. Cancel/delay clause → HOLD resolves YES (−1pt). Emergency inter-meeting hike does not count for this rung (rules_secondary; negligible at this horizon). No ambiguity in "Hike 25bps" for an exact-action rung. Total deductions ≈ 1pt → final 79%.

**Book & entry.** Entry context: yes 74/75, 3,621 contracts offered at 75¢, vol24h 3,251, OI 21,061. 14-day tape: 38¢ (Aug 4) → 87¢ intraday (Aug 14, leak day, 5.8k vol) → consolidated 74–76 by Aug 17 — entering on consolidation, not chasing the spike (48h move is −8¢, downward). Depth cap 1/3 × 3,621 = 1,207 ≫ 233; slippage ≤ 1¢. EV ≈ +4¢/contract pre-fee (+2.7¢ after ~1.3¢ fee).

Precedents consulted: KXCBDECISIONJAPAN-26JUN15-H25 · 26JUN15-HOLD · 26JUL30-HOLD · KXFEDDECISION-26JUL-H25 · 26JUL-H0

### 2. Anthropic does NOT confirm an IPO before Oct 1, 2026 BUY NO MEDIUM corporate-event

KXIPOANTHROPIC-DATE-26OCT01 · closes 2026-10-01 03:59 UTC · cluster: anthropic-ipo-timing · Tier 2

Prob NO (pre → final)85% → 82%
Entry (limit)≤79¢
Edge+3.7¢
Contracts101
Max cost$79.79

**Thesis.** "Confirms an IPO" is mechanically defined (rules_secondary): SEC declares the S-1 effective, OR the offering is priced, OR an exchange assigns a ticker. Anthropic's S-1 is still a confidential draft (submitted Jun 1); IPO-watch reporting has the public flip in ~September, the institutional roadshow through September–early October, and first trade targeting October on Nasdaq. In the SpaceX precedent all three triggers fired within ~48 hours of first trade — a September roadshow alone resolves nothing. The fastest bull path (public flip by ~Sep 5, the 15-day JOBS Act clock, a one-week roadshow, pricing by Sep 30) is physically possible but requires beating the company's own reported target; mega-IPOs slip later, not earlier, and Anthropic has said timing depends on market conditions.

Precedent: KXIPOANTHROPIC-DATE-26JUL01 resolved NO — the resolver ruled the confidential draft S-1 announcement was none of the three triggers, and YES got cheaper on the "Anthropic files for IPO" headline (the store's inverse-headline case). KXIPOANTHROPIC-DATE-26JUN01 resolved NO on the same strict mechanics. KXIPO-26-SPACEX confirms the trigger cluster lands at pricing/effectiveness, not at roadshow launch (Jun 4 roadshow was not a trigger; Jun 11–12 was).
Resolver profile & deductions. Rules are crisp and twice-enforced strictly. Residual risk: an early Nasdaq ticker assignment being read as a trigger days before pricing (−2pt), plus acceleration risk in the hottest IPO on the board (blend −1pt toward a churning book that touched 31–32¢ Aug 11–14). Final NO = 82%. Specialist-dense book (vol24h ~4.5k) noted — the edge claimed is specific (trigger mechanics + timeline), not sentiment.

**Book & entry.** Entry context: yes 24/25 (NO ask 76), vol24h 4,522, OI 110,396. NO-side depth is thin at the touch: ~109 contracts within 76–78¢, 304 within 79¢ → limit 79¢ (exactly 3¢ max slippage), depth cap 1/3 × 304 = 101 contracts. 14-day tape: YES 15¢ → 31–32¢ (Aug 11–14, heavy vol) → faded to 24¢; buying NO into a fading YES, not chasing. EV ≈ +3.7¢/contract pre-fee at ~78.3¢ avg fill.

Precedents consulted: KXIPOANTHROPIC-DATE-26JUN01 · 26JUL01 · KXIPO-26-SPACEX

## Position sizes & limits

| Market | Side | Limit | Contracts | Max cost | % of capital | Depth cap | Binding cap |
|---|---|---|---|---|---|---|---|
| KXCBDECISIONJAPAN-26SEP17-H25 | YES | 75¢ | 233 | $174.75 | 7.0% | 1,207 | MEDIUM cost cap (7%) |
| KXIPOANTHROPIC-DATE-26OCT01 | NO | 79¢ | 101 | $79.79 | 3.2% | 101 | Depth (1/3 × 304 visible ≤ 79¢) |
| **Deployed** |  | **$254.54** | **10.2%** |  |  |  |  |
| **Cash (a position)** |  | **$2,245.46** | **89.8%** |  | Reserve ≥ 10% satisfied |  |  |

## EV for the record

| Market | Side | Final prob | Entry | Edge | EV/contract (pre-fee) | Est. fee | Net EV (position) |
|---|---|---|---|---|---|---|---|
| KXCBDECISIONJAPAN-26SEP17-H25 | YES | 79% | 75¢ | +4.0¢ | +4.0¢ (5.3%) | ~1.3¢ | +$6.29 |
| KXIPOANTHROPIC-DATE-26OCT01 | NO | 82% | ~78.3¢ | +3.7¢ | +3.7¢ (4.7%) | ~1.2¢ | +$2.53 |

Sizing is conviction-weighted and EV-agnostic per v4; EV shown for the record only. Both edges are modest and honest — this is a consolidation-entry week, not a dislocation week.

## Cluster table

| Cluster | Legs surviving Stage 2 | Deployed leg | Runner-ups (logged) |
|---|---|---|---|
| boj-sep-2026-hike | H25 YES · HOLD NO · H25P | H25 YES @75 | HOLD NO (worse price, loses cancel clause) · H25P (structural zero) |
| anthropic-ipo-timing | 26OCT01 NO | 26OCT01 NO @≤79 | — (26SEP01 rung failed liquidity floor, vol24h 44) |
| trump-august-travel | NJGOLF -3 / -4 (both sides) | none — cluster rejected in full | -3 YES · -3 NO · -4 YES; correlated with live KXLAGODAYS-26AUG-0 pick |

## Conviction exposure

| Conviction | Positions | Cost | % of capital | Cap per pick |
|---|---|---|---|---|
| MEDIUM | 2 | $254.54 | 10.2% | 7% ($175) |
| HIGH | 0 | — | — | 15% ($375) |

No HIGH conviction available this week: nothing on the deployable slate is mechanically locked in. Both picks have a live way to be wrong (a September BoJ slip to October; an accelerated Anthropic pricing).

## Risk profile

- **Max loss if both picks lose: −$254.54 (10.2% of capital).** Cash is unaffected.
- **Correlation between the two picks: low.** One is Japanese monetary policy; the other is US IPO process mechanics. A global risk-off shock could hit both (BoJ delays, Anthropic accelerates is the _opposite_ direction — risk-off delays IPOs, which helps the NO), so joint downside requires two independent surprises.
- **BoJ pick, ways to lose:** a market shock or yen surge before Sep 17 (BoJ has skipped telegraphed moves under turmoil before); political pressure for delay; the leak being pace-jawboning rather than September-specific. The Fed's July counter-precedent is the cautionary tale — weighed and distinguished (leak-driven vs. data-inferred repricing).
- **Anthropic pick, ways to lose:** a public S-1 flip in the next ~2 weeks followed by a compressed September roadshow and a Sep 29–30 pricing, or an early ticker-assignment reading by the resolver. The 31¢ prints of Aug 11–14 may have been informed acceleration money — respected in the blend (85→82).
- **What is deliberately NOT held:** any leg of the Bedminster trip-count board (knife-edge exact counts, 63-point intraday swing on Aug 15), any second leg on a cluster, anything below 40¢ or above 84¢, and every zero-volume Tier-1 book however tempting the price.

## Execution notes

- **BoJ H25:** rest a limit at 75¢ (3,621 visible there). Do not chase above 75 — above it the edge is inside the fee. If the pre-meeting Nikkei/JiJi leak lands (typically days before Sep 17), the market will gap; no re-entry above band.
- **Anthropic NO:** work a 79¢ limit for 101 contracts; the book is thin at the touch (9 at 76¢), so expect fills across 76–79. If YES re-spikes above ~30 on a public S-1 flip headline, re-read the flip against the three triggers before adding — the precedent says filing headlines are fade material, but a _public_ flip starts the real clock.
- **Watch item (no action):** KXGOVTFULLFUND funding rungs into the Sep 30 fiscal deadline — Tier-1 archetype, currently below the liquidity floor; may become tradeable as the deadline approaches. Next run should re-check.

## Rejected (logged)

| Market | Side | Price | Model prob | Archetype | Reason |
|---|---|---|---|---|---|
| KXNJGOLFVISITCOUNT-26AUG-3 | YES | 73¢ | ~45% | multi-news-other | Pace-of-count thesis + no edge. Factbase (primary): 3 completed trips (Aug 1–2, 7–9, 14–16), Trump at the White House Aug 17, two weekends left after three straight Bedminster weekends; exactly-3 needs zero further arrivals and the "canceled vacation" reporting is from Aug _2025_, not 2026. Correlated with live KXLAGODAYS-26AUG-0 pick. Aug 15 candle: 23–86¢ in one session. |
| KXNJGOLFVISITCOUNT-26AUG-3 | NO | 28¢ | ~55% | multi-news-other | Entry band: 27–29¢ < 40¢ floor, despite positive model edge (tails 0-for-8 lifetime). |
| KXNJGOLFVISITCOUNT-26AUG-4 | YES | 27¢ | ~35% | multi-news-other | Entry band < 40¢; exact-count knife-edge; cluster runner-up. |
| KXCBDECISIONJAPAN-26SEP17-HOLD | NO | 77¢ | 81% | scheduled-process | Cluster runner-up: duplicates H25 YES at a worse price and loses the cancel/delay clause (HOLD wins if the meeting is delayed). |
| KXCBDECISIONJAPAN-26SEP17-H25P | — | 2¢ | ~1% | scheduled-process | Tail band + precedent: BoJ moves strictly in 25bp steps; ">25bp rung is a structural zero even when the bank moves" (26JUN15-H25P). |
| KXGOVTFULLFUND-26AUG-26SEP26 | — | 56/62 | — | confirmation/enact | Liquidity floor: vol24h 205 < 500, 6¢ spread. Only near-live Tier-1 book in the window; watch item. |
| KXKASHOUT-26APR-26SEP01 | — | 4¢ | — | departure/ouster | Tail band: ask ≤ 5¢ (vol24h 549 cleared the floor). |
| KXCOMPANYACTIONMERGER-27-26SEP01 | — | 4¢ | — | corporate-event | Tail band: ask ≤ 5¢ (liquid, vol24h 10,149). |
| KXUSAIRANAGREEMENT-27-26SEP | — | 1¢ | — | scheduled-process | Tail band: ask ≤ 5¢ (Oct rung likewise at 5¢). |
| KXALITOOUT-26SEP01 | — | 0.9¢ | — | departure/ouster | Tail band: ask ≤ 5¢. |
| KXHEGSETHOUT-26APR-SEP01 | — | 3¢ | — | multi-news-other | Already an active pick (NO). |
| KXPIRROOUT-26AUG-26SEP01 | — | 9¢ | — | multi-news-other | Already an active pick (NO). |

## Universe & method notes

- **Stage 0:** 60 allowlisted series swept live (public API, *_dollars fields); 89 open markets inside the 45-day horizon; 84 failed mechanical cuts (37 of them Tier-1 — overwhelmingly vol24h = 0 or ask ≤ 5¢); 5 candidates survived, all Tier-2.
- **Stage 1:** precedent store returned 25 scheduled-process rows (FOMC Jul-26, Hormuz ladders), 6 KXCBDECISIONJAPAN rows (Jun-26 meeting), 3 IPO rows (Anthropic Jun/Jul, SpaceX). No precedent existed for KXNJGOLFVISITCOUNT — cold-derived from rules_secondary + Roll Call Factbase.
- **Stage 2 discipline:** pre-market probabilities committed before opening books: BoJ H25 80%, Anthropic IPO YES 15%, NJGOLF-3 YES ~20% (blended to 45% after respecting the informed book — still no playable side in band).
- **Precedent writeback:** 1 row (KXCBDECISIONJAPAN-26JUL30-HOLD, verified API-verbatim) → precedents.json.
- **Stale-news trap flagged:** search surfaced "Trump cancels Bedminster vacation for Ukraine talks" — Aug 2025 reporting that pattern-matches the current question; rejected as evidence after date verification. Recorded here so the next run doesn't re-trip it.

Keel v4 · probability before price · one leg per cluster · cash is a position · Sources: Kalshi public API, mirror DB (resolution_precedents, trading_markets), Reuters via FXStreet (Aug 14), Japan Times (Aug 10), CNBC (Jul 31), Roll Call Factbase calendar, donaldtrump.golf tracker, Yahoo Finance / IPO-watch coverage of the Anthropic S-1.