Kalshi Politics — Mispricing Audit

v4 (archetype-targeted, precedent-fed) · Report date 2026-08-21 · Horizon 45 days · Capital $2,500 · Focus category: Politics
New picks
0
Capital deployed
$0
Cash position
100%
Logged rejects
16

Verdict

Zero picks this week — and that is the correct output, not a failure of the scan. The Stage-0 universe (70+ allowlisted series, 101 open markets inside the 45-day horizon) collapsed to just 8 mechanical survivors, of which 2 are already-live picks. Of the 6 genuine candidates, every single one has its playable side outside the v4 entry bands: the two central-bank events offer only sub-40¢ tails or ≥88¢ asks with no edge, and the two "sure thing" NO trades (Kash Patel, Tesla/SpaceX merger) sit in the banned ≥90¢ deep-favorite band. 40–70% cash is expected in an efficient week; this week it is 100%.

The finding of the week is a shadow trade, not a live one. Kalshi prices a September BOJ hike at 88–90¢ while 3-month TONA futures — institutional money on the same question — implied only ~53.2% as of Aug 20 (centralbank.watch). Our blended model says ~68%. The fade (NO on the hike leg at ~12¢) carries roughly a 20-point shadow edge but is a <40¢ tail, which is 0-for-8 lifetime and hard-banned. It is logged as a reject and will shadow-track; if this class of institutional-vs-retail divergence keeps showing paper profit, that is evidence for a future band revision — through the methodology, not through an exception tonight.

Stage 0 — Universe funnel

Candidates worked (Stage 1–3)

scheduled-processREJECTED — bandcluster: boj-sep-2026-decision

Bank of Japan, September 17–18 meeting (KXCBDECISIONJAPAN-26SEP17)

Precedents consulted: KXCBDECISIONJAPAN-26JUN15-H25 resolved YES with swaps at ~72.6% ("BoJ moves arrive telegraphed through deputy-governor speeches and press leaks"); -26JUN15-HOLD resolved NO ("price hold legs off the signalling, not the bank's no-show history"); -26JUN15-H25P ("BOJ normalizes strictly in 25bp steps — the >25bp rung is a structural zero"); KXFEDDECISION-26JUL-H25 ("a hawkish repricing into a live meeting is a fade candidate — dissent, not a rate move, is how the Committee vents pressure").

Diligence: BOJ held at 1.00% on July 31 with a Takata dissent proposing 1.25% and a warning that core inflation will run "clearly above" 2% (CNBC). A Reuters sources story (Aug 14) says the BOJ is "set to hike as soon as September" and to accelerate its tightening pace. Against that, 3-month TONA futures implied only 53.2% for a September move as of Aug 20.

Prob: 65% pre-market → 68% final (hike; nudged toward the liquid Kalshi book, anchored on TONA). Live book: H25 leg 88¢ bid / 90¢ ask, vol24h 2,511, OI 23,726.

Why rejected: Our 68% sits far below the 88¢ bid, so the live side is NO — at a 12¢ effective entry, inside the banned <40¢ tail band. The HOLD leg (YES at 12¢ ask) is the same thesis and additionally has 0.26 contracts of ask depth. The YES-hike side at 90¢ ask fails both the ≥90¢ rule and the edge test. All four expressions reject; the ~20pt NO-side divergence is logged for shadow-tracking.

scheduled-processREJECTED — band / no edgecluster: boi-aug31-2026-decision

Bank of Israel, August 31 decision (KXCBDISRAEL-26AUG31)

Precedents consulted: No series-specific rows in the store (cold derivation noted; July 6 meeting written back this run). Archetype rows (KXFEDDECISION, ladder mechanics) applied. Rules clarification common to all CB ladders: a cancelled/delayed meeting pays the HOLD leg.

Diligence: BOI cut 25bps to 3.50% on July 6 — its second consecutive cut and third of 2026 (BOI press release, Globes). Most forecasters expect a hold on Aug 31, but Governor Yaron said on Aug 18 the meeting "will be a live one" (Bloomberg). Inflation ~1.9%, inside target; BOI's own path sees ~3.0% in 12 months. Polymarket: 86% hold / 13% cut.

Prob: 80% pre-market → 83% final (hold; blended toward two agreeing liquid books). Live book: HOLD 87/88, vol24h 3,006, OI 19,161; C25 10/13, vol24h 514.

Why rejected: The only in-band expression is YES-hold at 88¢ ask, which needs >88% — our 83% says no edge (and the governor's "live" comment is exactly the kind of specific signal that caps a thin favorite). The NO-hold and YES-cut sides are 13¢ tails, <40¢ band; the cut leg's nominal 2¢ edge dies inside the 3¢ slippage allowance. No bet.

departure/ousterREJECTED — deep favoritecluster: kash-patel-out

Kash Patel out as FBI Director before Oct 1 (KXKASHOUT-26JUL-26OCT01)

Precedents consulted: KXKASHOUT-26APR-JUN01 and -JUL01 both resolved NO — "anonymous 'matter of time' sourcing is not a resolution event; nothing short of an actual vacancy pays," and "far rungs of monthly departure ladders are the most systematically overpriced." Rules require an actual departure date and vacating the role.

Prob: ~94% NO pre-market → 94% final. Book: YES 7/8, vol24h 534, OI 12,735. Precedent and price agree — but NO entry is 93¢, in the banned ≥90¢ band (max win 7.5% of risk; one resolver surprise erases ten wins). Rejected on band, not on thesis.

corporate-eventREJECTED — deep favoritecluster: tesla-spacex-merger

Tesla/SpaceX definitive merger before Oct 1 (KXCOMPANYACTIONMERGER-27-26OCT01)

Diligence: Rules demand an unambiguous, affirmative confirmation of a definitive signed agreement through official channels; Musk hypotheticals explicitly do not count. No such agreement exists or is pending. Prob: ~99% NO. Book: YES 4/6, vol24h 4,378. NO entry 96¢ — ≥90¢ hard reject. The Sep-01 sibling leg failed Stage 0 (ask ≤5¢).

Reject log (16)

TickerSidePxModel %Reason
KXCBDECISIONJAPAN-26SEP17-H25NO1232<40¢ band; TONA ~53% vs 88–90¢ book — shadow-tracked
KXCBDECISIONJAPAN-26SEP17-HOLDYES1232<40¢ band + no ask depth (0.26 contracts); cluster runner-up
KXCBDISRAEL-26AUG31-HOLDNO1317<40¢ band; in-band YES has no edge (83% < 88¢ ask)
KXCBDISRAEL-26AUG31-C25YES1315<40¢ band; 2¢ edge inside slippage
KXKASHOUT-26JUL-26OCT01NO9394Deep-favorite band (≥90¢)
KXCOMPANYACTIONMERGER-27-26OCT01NO9699Deep-favorite band (≥90¢)
KXIPOANTHROPIC-DATE-26OCT01NO53Already an active pick
KXPIRROOUT-26AUG-26SEP01NO94Already an active pick
KXGOVTFULLFUND-26AUG-26SEP2674vol24h 396 < 500 floor (only in-band price on the slate)
KXHEGSETHOUT-26APR-SEP014Existing pick + ask ≤5¢
KXUSAIRANAGREEMENT-27-26SEP1ask ≤5¢ (both legs)
KXNJGOLFVISITCOUNT-26AUG-378vol24h < 500; continuous-count thesis
KXLOCALLAW-KCROYALS-26OCT0198bid ≥95¢
KXMAMDANIEO-26AUG22-T01ask ≤5¢; pre-resolution window
KXCBDECISIONCANADA-26SEP-C2525vol24h < 500 across ladder
KXEBOLACOUNTRY-OCT26-CAR12spread >5¢ (whole 16-leg ladder)

Portfolio & risk

ClusterLegs deployedCostNote
— (none)0$0All candidates rejected at Stage 3/3.5
Cash$2,500100% of capital; reserve requirement trivially met

Conviction exposure: HIGH $0 / MEDIUM $0. EV-for-the-record: $0 deployed; the largest foregone shadow edge is BOJ-Sep NO-hike (~+20¢/contract at 12¢ entry, banned band). Execution notes: n/a — no orders.

Precedent writeback

Two settled markets in target series were reconstructed and verified this run and written to precedents.json:

Method note

Keel v4: Stage-0 allowlist universe (targets.json), probability committed before opening the book, precedent retrieval from resolution_precedents, resolver-verdict forecasting with priced deductions, book-derived no-trade band, 40–84¢ core band (85–89¢ half-size, ≥90¢ and <40¢ banned), one leg per cluster, live-book liquidity floor of 500 contracts/24h, conviction-capped EV-agnostic sizing. Cash is a position; zero picks is a valid and expected outcome in an efficient week.