Kalshi Politics — Mispricing Audit (Sentinel, managed)

Report date 2026-07-27 · Category: Politics (category-match mode) · Horizon 45 days · Capital $1,000 · Methodology v3.3 (conviction-weighted, EV-agnostic sizing; resolver risks priced per v3.1; sub-35¢ tails hard-rejected per v3.2; LOW conviction never deploys per v3.3)

1. How this was researched

Mode: category-match. trading_events.category = 'Politics' matched 352 active markets across 143 events closing within 45 days on the live Kalshi mirror (a theme-match fallback was not needed). Stage-1 mechanical screens cut 335: 271 below the 1,500-contract 24h-volume floor, 36 announcer-noise markets (Truth-Social-post counts, "who will Trump talk to", press-briefing counts), 14 fully priced (≥95¢/≤5¢), 12 already live picks in the feed (excluded up front per the duplicate rule, incl. KXHORMUZWEEKLY-26JUL26-T40, KXHEGSETHOUT-26APR-AUG01, KXKASHOUT-26APR-AUG01), and 2 on spread. The 17 survivors collapsed into 13 diligenced candidates after treating ladder legs of one event as a single thesis.

Each candidate got: rules text from the mirror, a 14-day daily price history from the Kalshi candlesticks API, a live orderbook snapshot at pick time, and primary-source news diligence (five parallel research passes). Two datasets did heavy lifting: the IMF PortWatch daily chokepoint feed (the actual resolver for all Hormuz markets — pulled directly from the PortWatch ArcGIS service) and RealClearPolitics' own backend JSON for the approval average and its daily history. Every pick's probability is a forecast of the named resolver's verdict at resolution time, with resolver risks priced as explicit point deductions (Stage 2.5).

2. Markets reviewed

TickerQuestionYES bid/ask24h volVerdict
KXHORMUZNORM-26MAR17-B260901Hormuz 7-day avg >60/day before Sep 18/956,478PICK — BUY NO
KXAPRPOTUS-26JUL31-40.2RCP approval below 40.2 on Jul 316/77,083PICK — BUY NO
KXCLARITYVOTE-26JUL-AUG08Senate recorded vote on CLARITY Act by Aug 859/6110,374PICK — BUY YES
KXHEGSETHOUT-26APR-SEP01Hegseth out as SecDef before Sep 18/83,822PICK — BUY NO
KXHORMUZWEEKLY-26JUL26-T30>30 Hormuz transit calls Jul 20–2688/895,090REJECT — resolver blind
KXHORMUZWEEKLY-26JUL26-T50>50 Hormuz transit calls Jul 20–2614/1514,910REJECT — resolver blind
KXHORMUZPEAK-26JUL26-T10Peak single day >10, Jul 20–2665/701,665REJECT — resolver blind + spread
KXTRUMPNUMSTATES-26AUG01-E10Exactly 10 distinct states in July71/7321,328REJECT — fairly priced
KXAPRPOTUS-26JUL31-40.9RCP approval 40.8–41.0 on Jul 3131/323,091REJECT — coin-flip band
KXMAMDANIEO-26AUG01-T0Mamdani non-emergency EO this week12/133,025REJECT — tail band (edge logged)
KXLAKECONF-26MAY12-SEP01Kari Lake confirmed before Sep 15/82,196REJECT — tail band (edge logged)
KXPARDONSTRUMP-26JUL-1710–24 people pardoned in July91/922,225REJECT — no edge + definitional
KXELECTIONEMERGENCY-26SEP01Election-emergency EO before Sep 110/111,927REJECT — fairly priced
KXTRUMPUFC-26AUG-DJTTrump attends UFC 33014/152,000REJECT — tail band, fair

3. Picks — detail and thesis

Pick 1 — KXHORMUZNORM-26MAR17-B260901 · BUY NO @ 0.92 · HIGH CONVICTION

Will the 7-day moving average of Strait of Hormuz transit calls (IMF PortWatch) be above 60 before September 1, 2026?

My prob (NO)96%
Market implied (NO)92%
Entry band>90¢ deep favorite
Clusterhormuz-recovery
Contracts150 ($138.00)
Edge+4¢ / +4.3%

Mispricing thesis. The market pays 8–9¢ for a recovery that has a physical floor under it. The resolver needs the PortWatch 7-day average to exceed 60 transit calls/day at some point before Sep 1. PortWatch is printing 6–17/day (its last seven prints, through Jul 19, sum to 85 — an average of ~12/day). Even the best post-crisis stretch — the late-June Islamabad-Memorandum ceasefire — peaked at a 7-day average of ~32/day, barely half the threshold. To resolve YES, a durable deal must be signed almost immediately, both the Iranian and US blockades lifted, mines swept, war-risk insurance repriced from 3–10% of hull value back toward 0.25%, and the world fleet must return — all by ~Aug 24 so a qualifying 7-day window can complete. Insurers are on record that this takes months.

Evidence.

Resolver profile (Stage 2.5). Resolver: IMF PortWatch 7-day moving average, resolving YES on any qualifying print before Sep 1 — a first-occurrence mechanism, so the NO side must survive every print for 5 weeks. Priced risks: −2 pts (PortWatch data revision or counting artifact spiking a weekly average). Evidence-based P(NO) ≈ 98 − 2 = 96.

Tail risks. Cleanest loss: a comprehensive US–Iran deal within ~2 weeks followed by an unprecedentedly fast fleet return, or a PortWatch data artifact. Note the market resolves on data prints, not peace headlines — a deal alone does not flip it.

Liquidity/entry. yes_bid 8 / yes_ask 9; NO fillable at 92¢ against 3,043 contracts at top of book; 24h volume 56,478; OI 854,333. Deepest book in the category.

Pick 2 — KXAPRPOTUS-26JUL31-40.2 · BUY NO @ 0.94 · HIGH CONVICTION

Will Trump's approval be below 40.2 at 11:00 AM ET on July 31, 2026, per RealClearPolitics?

My prob (NO)97%
Market implied (NO)94%
Entry band>90¢ deep favorite
Clustertrump-approval-rcp
Contracts120 ($112.80)
Edge+3¢ / +3.2%

Mispricing thesis. The RCP average is 40.9 today (verified three ways, including RCP's own backend JSON, poll window 7/6–7/26). Resolving YES requires a fall of ≥0.7 points in four days. RCP's own daily series shows a mean absolute daily move of 0.11 points; over the last 180 days only one day moved ≥0.5 (max 0.6) and typical 4-day drift is 0.2–0.4. And the single most likely composition change before Friday — Pew's 34% (fielded 7/6–12) aging out of the window — mechanically moves the average up 0.3–0.5. The 6¢ YES bid is Iran-war-headline demand ("string of polls below 40" — Forbes, Jul 27) paying for a move the average's arithmetic can't plausibly deliver.

Evidence.

Resolver profile (Stage 2.5). Resolver: the RealClearPolitics average, snapshot at exactly 11:00 AM ET Jul 31 — a snapshot mechanism (the genre that produced this program's costliest v1 loss, so it was priced deliberately). What can change the snapshot: which polls sit in the trailing window at that instant and any new entries. Priced risks: −1 pt (bulk poll-window turnover / RCP methodology quirk), −1 pt (shock event dumping several instant polls at 37–38 before Friday). Evidence-based P(NO) ≈ 99 − 2 = 97.

Tail risks. Cleanest loss: a catastrophic news shock in the next 96 hours (major war escalation with US casualties) producing multiple overnight polls in the 37s and RCP ingesting them before 11 AM Friday.

Liquidity/entry. yes_bid 6 / yes_ask 7; NO fillable at 94¢ against 3,414 contracts; 24h volume 7,083; OI 16,411. Market opened Jul 25; YES ticked 4¢ → 7¢ — no adverse 48h move against the NO side beyond 3¢ of tail demand.

Pick 3 — KXCLARITYVOTE-26JUL-AUG08 · BUY YES @ 0.61 · MEDIUM CONVICTION

Will the Senate conduct any recorded vote on the CLARITY Act before Aug 8, 2026?

My prob (YES)65%
Market implied (YES)61%
Entry band60–90¢ favorite
Clustercrypto-legislation
Contracts114 ($69.54)
Edge+4¢ / +6.6%

Mispricing thesis. This is a rules-vs-headline gap — the crowd is pricing the bill's troubles, but the contract pays on any recorded vote, and the rules explicitly count a cloture vote on the motion to proceed, even one that fails. Thune's Jul 23 comments ("I don't think we'll be able to get them done… I would like to at least get Clarity started. We'll see where the votes are") cratered this market from 73¢ to 39¢ — but he ruled out passage, not floor action. Punchbowl's week-of-Jul-27 floor lookahead states a crypto market-structure "procedural vote [is] expected regardless of Democratic agreement on ethics restraints." A failed cloture vote resolves this YES.

Evidence.

Resolver profile (Stage 2.5). Resolver: occurrence of a qualifying recorded vote before Aug 8 — an unambiguous, publicly observable mechanism (Senate roll-call record); definitional boundaries are unusually well-specified in the rules. Priced risks: −3 pts (floor congestion — Russia sanctions vehicle, CR, nomination packages — lets leadership punt to September if the ethics deal collapses by ~Jul 30), −2 pts (no cloture motion on CLARITY actually filed as of Jul 27). Evidence-based P ≈ 70 − 5 = 65.

Tail risks. Cleanest loss: ethics talks collapse this week and Thune quietly shelves the MTP rather than burn two floor days on a doomed cloture vote — his "we'll see where the votes are" is a ready off-ramp.

Liquidity/entry. yes_bid 59 / yes_ask 61; 287 contracts fillable at 61¢; 24h volume 10,374; OI 48,813. Related exposure note: the live feed already carries NO on the same bill becoming law (KXCRYPTOSTRUCTURE, prior runs) — that pairing is coherent (vote happens, bill still fails), but readers should know both ride crypto-legislation news.

Pick 4 — KXHEGSETHOUT-26APR-SEP01 · BUY NO @ 0.92 · MEDIUM CONVICTION

Will Pete Hegseth be out as Secretary of Defense before Sep 1, 2026?

My prob (NO)96%
Market implied (NO)92%
Entry band>90¢ deep favorite
Clusterhegseth-tenure
Contracts76 ($69.92)
Edge+4¢ / +4.3%

Mispricing thesis. Personnel-departure markets structurally overprice exits, and this one carries 8% for a 5-week window in which every observable is anti-departure: Hegseth testified to Congress on Jul 21 running the Iran war's $67B supplemental request, announced a new force-wide policy Jul 15, and published a "warriors not bureaucrats" piece Jul 21. There is no July reporting of Trump wavering, no successor shortlist, no investigation at a decision point. Wartime secretaries mid-funding-fight don't get quietly swapped in August.

Evidence.

Resolver profile (Stage 2.5). Resolver: formal, permanent vacation of the office before Sep 1 (early-close market). Rules are tight: temporary leaves, suspensions, recusals do not count — which protects NO against a "Hegseth sidelined" headline resolving against us. Death clause resolves at last fair price (roughly neutral). Priced risks: −1 pt (abrupt Trump firing in a volatile administration). Evidence-based P(NO) ≈ 97 − 1 = 96.

Tail risks. Cleanest loss: Trump fires him over a new scandal or an Iran-war setback with zero advance reporting — the only live path, and inherently unforecastable.

Liquidity/entry. yes_bid 8 / yes_ask 8 (1¢-tight book); NO fillable at 92¢ against ~323 contracts at top of book, 3,000+ within a cent; 24h volume 3,822; OI 93,974. Cluster note: the live feed already holds NO on the Aug 1 leg of this same event from a prior run — same underlying thesis, so this run keeps the new leg at MEDIUM size ($70, 7%) rather than stacking it.

4. Recommended $1,000 portfolio

Sizing is set by conviction tier + entry band only (HIGH ≤ $150/pick, MEDIUM ≤ $70/pick, 15% per-cluster cap). EV columns are recorded for calibration and did not drive contract counts — the resolved v1/v2 record showed stated EV was anti-predictive.

#TickerSideLimitConviction / bandContractsCostMax payoutEdge ¢EV %
1KXHORMUZNORM-26MAR17-B260901NO92¢HIGH / >90¢150$138.00$150.00+44.3
2KXAPRPOTUS-26JUL31-40.2NO94¢HIGH / >90¢120$112.80$120.00+33.2
3KXCLARITYVOTE-26JUL-AUG08YES61¢MEDIUM / 60–90¢114$69.54$114.00+46.6
4KXHEGSETHOUT-26APR-SEP01NO92¢MEDIUM / >90¢76$69.92$76.00+44.3
Total deployed$390.26$460.004.4 blended
Cash held (cash is a position)$609.74 (61%)Expected dollar edge ≈ +$17.20

Cluster exposure (cap: 15% of capital = $150 per cluster)

ClusterPicksCost% of capitalCap
hormuz-recovery#1$138.0013.8%15%
trump-approval-rcp#2$112.8011.3%15%
crypto-legislation#3$69.547.0%15%
hegseth-tenure#4$69.927.0%15%

The Hormuz Jul 20–26 weekly markets were rejected outright, so hormuz-recovery holds only the Sep-1 normalization leg — no correlated Hormuz double-bet. Clusters #3 and #4 each overlap a thesis already live in the feed from prior runs (crypto-law NO; Hegseth Aug-1 NO); both were deliberately kept at MEDIUM size.

Conviction exposure

TierPicksCost deployed% of deployed
HIGH2$250.8064%
MEDIUM2$139.4636%
LOW0$0 (never deploys, v3.3)0%

Risk profile

Execution notes

5. What I rejected and why

All ten Stage-2 survivors below are machine-logged in picks.json and shadow-tracked to settlement — if the rejects outperform the picks, the screen adds nothing and we want to know. Three cuts came from the Stage-3.5 entry-band screen; note the discipline cost real modeled edge on Mamdani and Lake.

TickerWould-be side @ priceMy probReject reason
KXHORMUZWEEKLY-26JUL26-T30YES @ 89¢85Resolver blind. PortWatch's last daily print is Jul 19 — the entire Jul 20–26 resolution window is unpublished (8-day lag), and PortWatch (10–17/day through Jul 19, no downtrend) diverges sharply from Lloyd's/NBC trackers showing a collapse to ~1 vessel/day. Modeled edge negative at 89¢. Stage-2.5 deductions exceed the ~15-pt cap.
KXHORMUZWEEKLY-26JUL26-T50NO @ 86¢82Same blind window; my scenario range for the PortWatch total (≈35–55) spans the threshold. This exact resolver produced the v3.2 tail loss (printed 28 vs a tracker's 51). Fair-priced; no defensible edge.
KXHORMUZPEAK-26JUL26-T10YES @ 70¢72Best statistical case of the family (needs one day >10; Jul 18–19 printed 17 and 15 with no new shock at the boundary) — but the same unpublished-window problem, a 5¢ spread, and a 2.8k-OI book. Entry-band screen note: survives the band but fails Stage 2.5's resolver-deduction cap.
KXTRUMPNUMSTATES-26AUG01-E10YES @ 73¢72Fairly priced. Verified count: 8 distinct states through Jul 26; Michigan today makes 9; announced Jul 31 Camp David (MD) cabinet meeting makes 10. But the split hinges on whether Andrews AFB transits count as Maryland under VISITAREA rules, and Camp David already cancelled once (May 26). No edge at 73¢.
KXAPRPOTUS-26JUL31-40.9YES @ 32¢30Entry-band screen. Genuine coin flip between the two middle bands (average 40.9 now, estimate ~41.0 Friday, Pew-dropout skews up). Coin-flip/tail pricing without HIGH conviction and ≥15¢ edge is a mandated reject — this is the band that hit ~33% historically.
KXMAMDANIEO-26AUG01-T0YES @ 13¢25Entry-band screen — sub-35¢ tail, hard reject (v3.2), no exceptions. Logged loudly: modeled +12 pts of edge (empirical non-emergency-EO weekly hit rate ≈28%; he signed EOs Jul 20 and Jul 24). The shadow track will tell us what this discipline costs.
KXLAKECONF-26MAY12-SEP01YES @ 8¢35Entry-band screen — sub-35¢ tail, hard reject (v3.2). The largest modeled edge in the audit: reported out of SFRC 12–10 on Jul 22 after Cornyn's PEPFAR deal, and pre-recess en bloc packages have historically swept in dozens of calendar ambassadors. Prime shadow-track candidate for the reinstatement argument.
KXPARDONSTRUMP-26JUL-17YES @ 92¢89No edge (model 89 < price 92) plus an unpriceable definitional trap: DOJ's list shows 17 July-3 grants including corporate entities vs 11 natural persons — whether companies count as "persons," and a Friday-Jul-31 news-dump batch, could each flip the 10–24 band.
KXELECTIONEMERGENCY-26SEP01NO @ 90¢91Roughly fair. The activist draft EO is real and Trump's Jul 16 election-security speech tracked its framing, but his announced path is legislative (SAVE America Act) and he publicly denied emergency plans; ~1¢ of edge is not deployable.
KXTRUMPUFC-26AUG-DJTYES @ 15¢16Entry-band screen (tail) and fairly priced anyway. ~23% base rate at numbered PPVs, Philadelphia fits his August geography, but he skipped the far bigger McGregor card Jul 11 and no attendance reporting exists.

6. Sources

Data sources: Kalshi live mirror (read-only SQL), Kalshi public API v2 (live orderbooks and candlesticks at pick time), IMF PortWatch ArcGIS feed, RealClearPolitics backend JSON, and the primary-source reporting linked above. All probabilities are subjective estimates by the auditing model (v3.3 methodology); prediction-market contracts can and do resolve to zero, and losing the entire stake on any pick is a normal outcome. Prices move — the entry context recorded per pick is the honest baseline for judging fills. This report is part of a public performance track record; picks are cleared for auto-copy and rejects are shadow-tracked to settlement. Nothing here is financial advice.