# Kalshi Politics — Mispricing Audit (v4, archetype-targeted) — 2026-08-20

Date: 2026-08-20
Source: https://kalshi-audits.pages.dev/2026-08-20-kalshi-politics-mispricing-audit-v4

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

2026-08-20 · Keel v4 · Category focus: Politics · Horizon: 45 days ·
    Capital: $10,000 · Universe: targets.json (70 allowlisted series, generated 2026-08-20 12:06 UTC)
  

Verdict: 0 picks · 7 logged rejects · $0 deployed · 100% cash ($10,000)

    A zero-pick week, and correctly so. The Stage-0 mechanics reduced 101 open target-series markets
    to six with live books, and every survivor then failed on band, conviction, edge, or fillable depth.
    The single best raw edge of the run — NO on a government-wide funding law before Sep 11 at ~55¢
    against a 64% model — is exactly the trade v4's band rules exist to gate: a MEDIUM-conviction
    thesis in the HIGH-only 40–59¢ band, in a book born this morning with $36 of fillable depth.
    It is logged and will shadow-track. Cash is a position.
  

## Stage funnel

| Stage | In | Out | What happened |
|---|---|---|---|
| Stage 0 — universe & mechanics | 101 | 6 | Open markets in allowlisted series closing by Oct 4. Cuts: vol24h<500 (83), YES ask ≤5¢ (11), YES bid ≥95¢ (1) — live numbers, not file snapshots. |
| Band pre-screen | 6 | 4 | Tesla/SpaceX merger (5/7¢) and Pirro-out (8/9¢) unplayable on both sides by band rules; logged, no diligence spent. |
| Stages 1–2.5 — precedents, diligence, resolver profile | 4 | 4 | Precedent retrieval (12 rows scheduled-process, 11 confirmation/enact, 3 corporate-event), primary-source news via three parallel researchers, pre-market probabilities committed before opening any book. |
| Stage 3–4 — reject-before-keep, bands, sizing | 4 | 0 | All four rejected: no-edge-at-ask (2), band/conviction mismatch (1), sub-40¢ tails and liquidity floor on remaining ladder legs. |

## Picks

**None.** No market on the slate offered a deployable combination of (a) an in-band entry,
(b) the conviction tier that band requires, (c) ≥3–5 points of blended edge after resolver-risk
deductions, and (d) a book that can fill meaningful size within 3¢ of slippage. Per v4 doctrine,
40–70% cash is expected in an efficient week; this week it is 100%.

## Rejected — the full log (shadow-tracked)

### 1 · KXGOVTFULLFUND-26AUG-26SEP11 — Government-wide funding law before Sep 11

confirmation/enact · Tier 1cluster: fy27-appropriations
best raw edge of the runREJECT: band/conviction mismatch + no depth

**Shadow side:** NO at ~55¢ · Prob: 33% YES pre-market → 36% final (NO = 64%) · Conviction: MEDIUM

**Thesis.** Both chambers have already passed government-wide CRs (House: clean, through Dec 4, passed Jul 21 220-205;
  Senate: through Dec 11, passed Aug 8, 90-6, with an OMB grants-rule rider the House GOP dislikes — NEFI, Aug 11).
  For YES before Sep 11, the House must clear the Senate bill in its only pre-Labor-Day window (Aug 31–Sep 3, per the
  House Press Gallery: “next votes expected Monday, Aug 31”), with _no announced floor plan_ (verified Aug 20 against
  the Press Gallery and CBS), and the President must sign by Sep 10 with zero deadline pressure — the Senate is adjourned
  to Sep 14, so an amended bill cannot ping-pong in time. That is a fade of an unannounced fast-track, not a lock.

**Why rejected.** The 40–59¢ band deploys HIGH conviction only (the DHS-at-47¢ lesson: process already in motion).
  A live 4-day fast-track path caps this at MEDIUM. Independently, the book was born _today_ (first-ever trades
  Aug 20: 1,232 contracts) and visible NO depth at ≤55¢ is 200 contracts → 1/3 depth cap = 66 contracts ≈ $36 of cost.
  There is no position to build. Logged so the shadow record can test whether the band rule is leaving money on the table.

Precedents: KXRECNCH-26-JUN06 [no] — “one chamber clearing a bill does not compress the other chamber's floor
  schedule”; KXRECNCH-26-JUN03 [no] — “leadership-announced target dates are the optimistic tail”; KXBILLS-HOUS [yes] —
  the 10-day presentment clock enacts without a signature.
  Entry context: 45/49¢, vol24h 1,232, OI 227, NO depth@≤55¢ = 200.

### 2 · KXGOVTFULLFUND-26AUG-26SEP18 — Government-wide funding law before Sep 18

confirmation/enact · Tier 1cluster: fy27-appropriations
REJECT: <40¢ band; cluster runner-up

**Shadow side:** NO at ~34¢ · Prob: 48% YES pre-market → 55% final · Conviction: LOW

NO at 30–34¢ is a tail purchase — the banned band (tails 0-for-8 lifetime), regardless of the apparent 11-point
  shadow edge against a first-day book quoting 66/70. Even if in band it would lose the one-leg-per-cluster contest to SEP11.

Entry context: 66/70¢, vol24h 654, OI 600.

### 3 · KXGOVTFULLFUND-26AUG-26SEP26 — Government-wide funding law before Sep 26

confirmation/enact · Tier 1cluster: fy27-appropriations
REJECT: liquidity floor; zero edge

**Shadow side:** YES at 78¢ · Prob: 78% YES pre-market — ask 78, market fair

The rung the evidence actually likes (both chambers done, White House SAP supportive, election-year incentive,
  ~12–15% residual shutdown tail after two shutdowns this fiscal year) — but vol24h is 400 (<500 hard floor, the
  98-of-136-zero-volume lesson) and the ask offers no edge anyway.

Entry context: 72/78¢, vol24h 400, OI 4,381.

### 4 · KXCBDISRAEL-26AUG31-HOLD — Bank of Israel holds at the “August” meeting

scheduled-process · Tier 2cluster: boi-sep2026
specialist-dense book (vol24h 9.7k)REJECT: no edge at ask; unexplained 48h move

**Shadow side:** YES at 85¢ · Prob: 70% YES pre-market → 78% final · Conviction: LOW

**Resolver fact established (the real work here).** The BoI moved this decision from Aug 31 to **Sep 1**
  (Reuters, Jul 6 — Jackson Hole conflict). The rulebook auto-resolves “No change” YES only if the meeting is delayed
  _past expiration_, and expiration is Sep 7 — so the clause does **not** fire; the market settles on the actual
  Sep 1 decision. Pre-market: P(hold) ≈ 62% (July CPI 1.5% y/y, lowest since May 2021; Hapoalim and Meitav leaning cut
  post-CPI; Governor Yaron on Aug 18: the decision is “live”; against that, hot Q2 GDP +15.4% and election-year fiscal
  risk make hold modal), plus ~20% weight on a resolver reading the August meeting as “cancelled” → auto-YES, → 70%.

**Why rejected.** Blended to 78% against a specialist-dense book — still below the 85¢ ask, so YES has negative
  model edge; NO at 15–17¢ is a banned tail. The price also jumped 76→88→83 on Aug 19 (2,443 contracts) on no public
  catalyst I could identify — the v3 48-hour rule bars deploying against an unexplained move. The precedent that would
  justify buying a hold rung (KXFEDDECISION-26JUL-H0: buy it when a repricing has knocked it _below_ par) points the
  wrong way: this rung is priced above my number, not below.

Precedents: KXFEDDECISION-26JUL-H0 [yes], KXFEDDECISION-26JUL-H25 [no].
  Entry context: 83/85¢, vol24h 9,727, OI 15,601. Sources: Reuters via investing.com (Jul 6), Bloomberg (Aug 18),
  Globes CPI (Aug 14), CBS Q2 GDP via Globes (Aug 16).

### 5 · KXIPOANTHROPIC-DATE-26OCT01 — Anthropic confirms an IPO before Oct 1

corporate-event · Tier 2cluster: anthropic-ipo
specialist-dense book (vol24h 6.2k)REJECT: insufficient edge post-fade

**Shadow side:** NO at 79¢ · Prob: 15% YES pre-market → 20% final (NO = 80%) · Conviction: LOW

**Evidence.** EDGAR-verified Aug 20: no public S-1, no effectiveness, no pricing, no assigned ticker — the three
  mechanical triggers this resolver honors (series precedent; on SpaceX it did not fire at the public flip, only in the
  final pre-listing 48h). Reporting: The Information (Aug 18) — listing “as soon as late September,” governance terms
  still being restructured last week; FT (Aug 13) — October at ~$2T is modal. Only an aggressively early late-September
  pricing resolves YES.

**Why rejected.** The market already ran this exact fade: YES spiked 15¢→32¢ on the FT story (Aug 13–14) and
  decayed back to 22–25¢ by Aug 20. Fillable NO is 78–80¢ (6 contracts at 78, 146 at 79, 5.2k only at 80) against a
  blended 80% — 0–2 points, inside the 3¢ slippage tolerance, in a book flagged specialist-dense. A prior-run Keel NO
  on the SEP01 leg (now 0/1¢) already carries this cluster's exposure.

Precedents: KXIPOANTHROPIC-DATE-26JUL01 [no] — the confidential filing met zero triggers (inverse-headline
  case, 130k contracts); KXIPO-26-SPACEX [yes] — triggers fire together at pricing.
  Entry context: 22/25¢, vol24h 6,213, OI 115,568. Sources: EDGAR full-text search, anthropic.com/news (Jun 1),
  CNBC (Jul 15), FT via Fortune (Aug 13), The Information via relays (Aug 18).

### 6 · KXCOMPANYACTIONMERGER-27-26OCT01 — Tesla/SpaceX definitive merger before Oct 1

corporate-event · Tier 2cluster: tesla-spacex-merger
REJECT: deep-favorite band

YES 5/7¢ is a sub-40¢ tail; NO at ~93–95¢ sits in the ≥90¢ reject band. Rules demand unambiguous confirmation of a
  definitive signed agreement — Musk intent statements explicitly excluded. Unplayable both ways; no diligence spent.

Entry context: 5/7¢, vol24h 7,027, OI 53,440.

### 7 · KXPIRROOUT-26AUG-26SEP01 — Pirro leaves as US Attorney (DC) before Sep 1

multi-news-other · Tier 2cluster: pirro-dc-usattorney-exit
REJECT: deep-favorite band

YES 8/9¢ tail vs NO ~91–92¢ deep favorite — both banned. The record's worst live loss was a 94¢ “lock” that broke
  in 48 hours; that memorial stands. A prior-run Keel NO on this exact ticker already shadow-tracks it.

Entry context: 8/9¢, vol24h 1,218, OI 37,852.

## Portfolio & sizing

| Position | Cost | % of capital | Note |
|---|---|---|---|
| Cash (reserve) | $10,000 | 100% | No pick met band + conviction + edge + depth simultaneously. ≥10% reserve rule trivially satisfied. |

**Cluster table:** one cluster (fy27-appropriations) produced three ladder legs; under the one-leg rule only
SEP11 could ever have deployed, and it failed conviction-band and depth. All legs logged. No conviction exposure;
EV-for-the-record: $0 deployed, $0 at risk.

## Risk profile & execution notes

- **Existing picks respected:** KXCRYPTOSTRUCTURE-26JAN-OCT (NO) does not appear above; not re-researched.
- **Cluster overlap disclosure:** other live strategy signals already hold NO on KXIPOANTHROPIC-DATE-26SEP01 and NO
      on KXGOVTSHUTDOWN-26OCT01 (fy27-appropriations adjacent). Nothing in this run adds correlated exposure.
- **What would change the verdict this week:** (a) a Johnson announcement scheduling — or ruling out — the Senate CR
      for the Aug 31 week would move GOVTFULLFUND-SEP11 NO to HIGH conviction or kill it; (b) a public Anthropic S-1 flip
      dated for an October roadshow would make OCT01 NO a clean core-band pick if the price lags; (c) a Reuters poll for the
      Sep 1 BoI decision showing a near-unanimous hold would explain the Aug 19 move and re-fair the HOLD leg.
- **Fresh-book caution:** both GOVTFULLFUND September legs printed their first trades today; quotes there are
      price discovery, not consensus.

## Method (v4) — what was actually done

- Stage 0 from `targets.json` only: 101 open markets in 70 allowlisted series → 6 mechanical survivors
      (live vol24h ≥ 500, spread ≤ 5¢, not pinned in either tail).
- Stage 1 precedent retrieval before any opinion: 26 precedent rows read across scheduled-process,
      confirmation/enact, corporate-event archetypes; each surviving thesis cites the rows it leans on.
- Stage 2 probability-before-price: `prob_pre_market` committed to disk from primary-source evidence
      (three parallel researchers barred from consulting market odds) _before_ orderbooks or candlesticks were opened;
      blended numbers shown on every card.
- Resolver clarifications read from live `rules_secondary` on every candidate; the BoI meeting-move/expiration
      analysis and the Anthropic trigger mechanics were decided on rules text + precedent, not vibes.
- Precedent writeback: 1 row (KXIPOANTHROPIC-DATE-26AUG01, EDGAR-verified) → `precedents.json`.

  Keel v4 · run 2026-08-20 · report: kalshi-politics-mispricing-audit-v4 · machine-readable picks at
  /tmp/out/picks.json · precedent writeback at /tmp/out/precedents.json ·
  Data: kalmari.app mirror (schema-verified), Kalshi public API (*_fp/*_dollars fields),
  primary-source web diligence. Prices in cents; entry = cost of stated side.