Kalshi Politics — Mispricing Audit (Sentinel, managed)

Report date 2026-08-03 · Methodology v3.3 (conviction-weighted, EV-agnostic sizing; resolver risk priced; sub-35¢ tails and LOW conviction never deploy) · Horizon 45 days · Capital $1,000 · Picks count toward the public track record and are cleared for auto-copy.

1. How this was researched

Mode: category-match. trading_events.category = 'Politics' matched directly — 335 active markets closing before 2026-09-17 in the live Kalshi mirror. No theme-keyword fallback was needed.

Pipeline: (1) latest snapshot per market from market_snapshots; mechanical screen cut anything with volume24h < 1,500, spread > 5¢, price ≥ 95¢ / ≤ 5¢, or already live in the feed (26 existing pick lines excluded up front, incl. the KXCPI/KXPAYROLLS/Hormuz-weekly and CLARITY picks). 22 survivors. (2) Seven of the 22 warranted full Stage-2 diligence (rules text, 14-day intraday history, live orderbook, primary-source news); the other 15 were sub-35¢ ladder legs / tails that die at the v3.2 band screen regardless of thesis. (3) Stage 2.5 resolver profiling and the Stage 3.5 conviction + entry-band screens produced 1 final pick and 13 logged rejects. This was an efficient week in Politics: most liquid action sat in exact-count ladders (Truth Social posts, golf-trip counts, approval-strike bands) where either the band screen or resolver ambiguity kills deployment. A one-pick, 85%-cash book is the methodology working, not a shortfall.

2. Markets reviewed

Ticker	Market	YES bid/ask ¢	Vol 24h	Disposition
KXLAGODAYS-26AUG-0	Trump makes 0 Mar-a-Lago trips in Aug 2026	92 / 93	2,649	PICK — BUY YES · HIGH
KXAGCONF-26JUN05-SEP01	First announced AG pick (Blanche) confirmed before Sep 1	84 / 85	29,263	Reject — price already moved today (18→89)
KXEOWEEK-26AUG08-0	>0 executive orders signed Aug 2–8	87 / 89	2,029	Reject — unverifiable catalyst; modeled edge negative
KXAPRPOTUS-26AUG07-39.2	RCP approval below 39.2 at 11am ET Aug 7	65 / 67	5,221	Reject — knife-edge at strike; resolver unauditable
KXHORMUZWEEKLY-26AUG09-T55	>55 Hormuz transit calls Aug 3–9 (PortWatch)	72 / 74	1,859	Reject — resolver deductions >15 pts (PortWatch history)
KXJUDGECOUNT-26AUG-1	≥2 federal judges confirmed in Aug 2026	29 / 31	1,776	Reject — YES side is sub-35¢ tail (hard reject)
KXNJGOLFVISITCOUNT-26AUG-3	Exactly 3 Bedminster trips in Aug 2026	58 / 62	2,161	Reject — trip-count definitional ambiguity
KXNJGOLFVISITCOUNT-26AUG-2 / -4	Exactly 2 / exactly 4 Bedminster trips	25/26 · 5/7	1,744 · 1,684	Reject — sub-35¢ tails
KXAPRPOTUS-26AUG07-39.3 / -39.6	RCP approval in 39.2–39.4 / 39.5–39.7 band	15/16 · 12/13	2,400 · 2,864	Reject — sub-35¢ tails
KXTRUTHSOCIAL-26AUG08 (×7 legs)	Trump Truth Social post-count bands, week of Aug 2–8	5–23	2,142–8,616	Reject — tails + posting-volume noise
KXTRUMPUFC-26AUG-DJT	Trump attends UFC 330	10 / 11	4,872	Reject — sub-35¢ tail
KXLAGODAYS-26AUG-1	Exactly 1 Mar-a-Lago trip in Aug 2026	7 / 8	12,222	Reject — sub-35¢ tail (complement of pick)
KXHORMUZWEEKLY-26JUL26-T70	>70 Hormuz transit calls Jul 20–26	15 / 18	1,638	Reject — tail; closes Aug 4
KXACTBLUETOP-26AUG07-JOSS	Ossoff top ActBlue fundraiser Aug 7	6 / 8	1,581	Reject — tail + unauditable resolver

3. Top picks — detail and thesis

Pick 1 — KXLAGODAYS-26AUG-0 · BUY YES @ 0.93 · HIGH CONVICTION

Will Trump make 0 trips to Mar-a-Lago as President in Aug 2026? — closes 2026-09-01 14:00 UTC

Market: 93¢ (implied 93%)

My probability: 97%

Edge: +4¢ (~+4.3% on cost)

Entry band: >90¢ deep favorite

Cluster: trump-august-location

Mispricing thesis. This is a seasonal near-mechanical zero priced as if it carried 7% event risk. Mar-a-Lago closes for the season after Mother's Day and does not reopen until fall; Trump spends August at Bedminster, NJ — the market's own sibling ladders (Bedminster trip-count, Truth Social) are all priced off a Bedminster August. Across every August of both Trump presidencies the Mar-a-Lago trip count was zero. The 7¢ discount is a rent payment from capital that won't hold a 4-week position, not a probability statement.

Evidence.

Mar-a-Lago closes annually for the summer after Memorial Day/Mother's Day; Trump relocates summer operations to Bedminster ("Summer White House") — Daily Beast, locatetrump.com, Yahoo/Palm Beach Post.
The seasonal reopening is an autumn event (political HQ returns to Mar-a-Lago "with the onset of autumn" — Fox News); no public schedule shows any Florida/Mar-a-Lago event in August 2026.
Flag checked and cleared: the Feb 22, 2026 Mar-a-Lago Secret Service shooting — Trump was in Washington; it has no bearing on August movements.
Market context: the sibling leg KXLAGODAYS-26AUG-1 ("exactly 1 trip") trades 7–8¢; the ladder is internally consistent with ~zero expected trips.

Resolver (Stage 2.5). Source: Kalshi resolves on the public record of presidential movements (pool reports / flight tracking); mechanism: count of trips over the full Aug 1–31 window; rules text: "If Trump makes exactly 0 trips to Mar-a-Lago in Aug 2026, then the market resolves to Yes." Unambiguous count, no snapshot risk. Priced risks: −2 pts (a Florida hurricane/disaster trip in peak season that includes a brief, countable stop at the closed property — the only plausible path to a non-zero count; historically even post-hurricane Florida visits, e.g. Irma 2017, did not touch Mar-a-Lago). Evidence-based 99% − 2 = 97%.

Tail risks (cleanest way to lose). A major hurricane strikes Palm Beach in August and Trump inspects damage at the property itself; or a one-off security/family event pulls him to the closed club. That is the entire loss case — 29 days of it.

Numbers. 161 contracts @ 93¢ limit = $149.73 cost (15.0% of capital — HIGH-tier cap). Max payout $161.00 (+$11.27). EV = 0.97 × $161 − $149.73 = +$6.44 (+4.3%). Recorded for calibration only — size was set by tier + band, not EV.

Liquidity / entry context. Live book at signal: YES 92 bid / 93 ask; 298 contracts offered at 93 (269 more at 94, 500 at 95) — our 161 fill at the limit with zero slippage. Vol24h 2,649; OI 4,374 and building (839 → 4,374 over three days).

Price history. Listed Aug 1; has traded a tight 87–94 band and sat at 92–93 for three consecutive sessions on rising OI. No 20¢ moves; nothing to suggest the market knows something the schedule doesn't.

4. Recommended $1,000 portfolio

#	Ticker	Action	Limit ¢	Contracts	Cost $	Max payout $	EV $	EV %	Tier / band (set the size)
1	KXLAGODAYS-26AUG-0	BUY YES	93	161	149.73	161.00	+6.44	+4.3	HIGH · >90¢ deep favorite
Cash held					850.27	85.0% of capital — cash is a position; the slate didn't earn more
Total				161	1,000.00	161.00	+6.44	+4.3 blended on deployed

EV cents / EV % are recorded for the calibration ledger only; per v3 they did not drive contract counts. Sizing = conviction tier (HIGH → 15% cap = $150) + entry band.

Cluster exposure

Cluster	Cost $	% of capital	Cap
trump-august-location	149.73	15.0%	15%

Note: the rejected Bedminster golf-count and Mar-a-Lago sibling legs would have shared this cluster; the 15% cap is fully consumed by the single strongest expression of the thesis — by design.

Conviction exposure

Tier	Cost deployed $	% of capital
HIGH	149.73	15.0%
MEDIUM	0.00	0%
LOW	— never deploys (v3.3)	—

Risk profile

Worst case: −$149.73 (−15.0% of capital) if Trump makes any Mar-a-Lago trip in August. Model probability of this: ~3%.
Best case / most likely (~97%): +$11.27 (+1.1% on total capital, +7.5% on deployed, in 29 days).
Concentration: one thesis, one resolver, capped at exactly the 15% cluster limit. No leg can flip the run's sign beyond that cap.

Execution notes

Limit 93¢, day-repost; 298 contracts are offered there now — do not lift 94+. If the offer fades, 92 bid-join is acceptable; the position is not urgent.
Invalidation triggers (exit or hedge): any pool report placing Trump in Palm Beach; a named hurricane on a Palm Beach track (watch NHC from mid-August); any announced Mar-a-Lago event before Sep 1. The complement leg (-1 at 7–8¢) is the natural hedge if a Florida trip becomes likely.
Watchlist: KXAGCONF (re-enter thinking if it retraces to ≤80 before Wednesday's committee vote with the deal intact); KXJUDGECOUNT YES if cloture is actually filed on judicial nominees this week (band rule still forbids it below 35¢ — watchlist only); next week's Hormuz strip only if PortWatch prints become auditable.

5. What I rejected and why

All 13 rejects below are machine-logged in picks.json and shadow-tracked to settlement — if the rejects outperform the picks, the screens add nothing, and we want to know that. Entry-band cuts are marked ⛔.

Diligenced rejects (Stage 2 survivors)

KXAGCONF-26JUN05-SEP01 (Blanche confirmed AG before Sep 1) — would be BUY YES @ 85, my prob 88%. The whole thesis printed today: Cornyn and Tillis flipped after DOJ agreed to terminate the $1.8B anti-weaponization fund and narrow the IRS-settlement scope (NPR, Aug 3; The Hill; CBS), committee vote re-set for Tuesday. The market repriced 18→89 intraday on 29k volume. The remaining 15¢ prices exactly the risk I'd price — a floor vote slipping past the Aug 10 recess (recess through Labor Day week, and the Sep 1 close makes recess = loss). Residual edge ≤3¢, inside my own forecast noise. Stage-3 "already moved in 48h" screen: reject.
KXEOWEEK-26AUG08-0 (>0 EOs Aug 2–8) — no deployable side. Spiked 65→90 today on a catalyst I could not verify anywhere public: whitehouse.gov shows no executive order since Jul 24 (Smithsonian EO) — late July produced only proclamations and memoranda, which do not count. 2026 cadence is ~1.45 EO/week and slowing into vacation season; my base-rate P(≥1) ≈ 70–75% vs an 89¢ ask — modeled edge on YES is negative. My lean is actually NO at 13¢ (modeled +15¢) — and that is precisely the anti-predictive high-EV tail v3.2 exists to kill. ⛔ NO side sub-35¢: hard reject.
KXAPRPOTUS-26AUG07-39.2 (RCP approval below 39.2, Aug 7 snapshot) — would be BUY YES @ 67, my prob ~57%. The RCP average sat at exactly 39.2 as of Aug 2 (RCP 2nd-term approval, via search index — the resolver page itself returns 403 to automated fetch and I could not audit the poll window). "Below 39.2" loses on a tie. Resolution = one aggregator's discretionary poll-inclusion arithmetic at an exact timestamp, four days out; a single R-friendly tracker entering the window moves the average more than the strike distance. Resolver deductions alone breach the Stage-2.5 cap. Reject: knife-edge on an unauditable resolver. (Sibling bands 39.3/39.6 at 12–16¢: ⛔ tails.)
KXHORMUZWEEKLY-26AUG09-T55 (>55 Hormuz transits Aug 3–9) — would be BUY YES @ 74. The resolver is IMF PortWatch, and this program's own resolved record includes a 75% loss where PortWatch printed 28 against a commercial tracker's 51. Weekly counts are also revision-prone within the settlement window. Priced resolver deductions exceed 15 pts → Stage-2.5 automatic reject. The feed additionally already carries live Hormuz-cluster exposure (T50 weekly YES, NORM NO) from prior runs.
KXJUDGECOUNT-26AUG-1 (≥2 judges confirmed in Aug) — would be BUY YES @ 31, my prob ~45%. ⛔ Sub-35¢ tail: hard reject regardless of the modeled +14¢. For the record, the evidence is genuinely split: four judicial nominees are pending on the floor and pre-recess confirmation pushes are reported (Nomination Notes), but Thune filed no judicial cloture heading into this final pre-recess week, whose floor time is committed to the Clayton DNI vote and the Blanche fight (Punchbowl, The Hill). The NO side at 71¢ is the favorites band but my probability sits against it — no deployable side.
KXNJGOLFVISITCOUNT-26AUG-3 (exactly 3 Bedminster trips) — no confident side, price 58–62. Exact-count market whose unit ("a trip") is undefined across a multi-week residential stay punctuated by DC returns — a one-arrival-count dispute flips adjacent legs. Definitional dispute I can't price with confidence → Stage-2.5/3 reject. Legs -2 and -4: ⛔ tails.

Entry-band screen cuts (⛔ sub-35¢, v3.2 — no exceptions)

KXTRUTHSOCIAL-26AUG08-B149 (modal post-count bin, 22–23¢) and six sibling legs — tails, and posting-volume noise besides; no forecastable edge on posting cadence.
KXTRUMPUFC-26AUG-DJT (Trump attends UFC 330, 10–11¢) — tail.
KXLAGODAYS-26AUG-1 (exactly 1 Mar-a-Lago trip, 7–8¢) — tail; complement of Pick 1.
KXHORMUZWEEKLY-26JUL26-T70 (15–18¢, closes Aug 4) — tail on a flaky resolver with hours to run.
KXACTBLUETOP-26AUG07-JOSS (6–8¢) — tail, plus "top ActBlue fundraiser" has no public auditable leaderboard.

6. Sources

Data: Kalshi live mirror (read-only SQL over trading_events, trading_markets, market_snapshots, queried 2026-08-03) and the Kalshi public trade API (/markets, /orderbook, live at signal time). Methodology v3.3: conviction-weighted, EV-agnostic sizing; 15% cluster cap; sub-35¢ tails and LOW-conviction leans are logged, shadow-tracked rejects — never deployed. All probabilities are subjective estimates; every contract can resolve to zero; nothing here is investment advice.