Kalshi Elections — Mispricing Audit

Report date 2026-06-22 · Horizon 45 days (markets closing on/before 2026-08-06) · Capital $1,000 · Methodology v3 (conviction-weighted, EV-agnostic, 15% cluster cap, entry-band screen)

Verdict 100% CASH

0 picks · 10 logged rejects · $1,000 held in cash

The Elections category inside a 45-day window is, this week, a textbook efficient-and-illiquid book. After pulling every active market, reading the rules, checking 14-day price history and live order books, and grounding probabilities in primary news, no contract clears the v3 screens. The markets split cleanly into three buckets, none deployable:

Per the v3 rule that cash is a position and a near-empty book in an efficient week is the correct answer — not a failure — I am deploying nothing and holding the full $1,000. Padding the book with a thin-edge deep favorite or a regime-change coin-flip on auto-copy real money is exactly what the resolved track record punished.

1. How this was researched

Selection mode: category-match. $CATEGORY="Elections" matched trading_events.category directly and returned 22 active markets closing within 45 days — comfortably enough that no theme/keyword fallback was needed.

Markets in category/horizon: 22 Survived Stage 1 cleanly: ~0 Diligenced: 10 Final picks: 0

2. Markets reviewed

The full Elections slate within the horizon. "Effective entry" is the price you'd pay for the side a buyer would naturally take. Every row is a reject — reasons in §4.

TickerMarketYES bid/askVol 24hOILeanDisposition
KXTRUMPOUT27…AUG01Trump leaves office before Aug 12 / 313,2162.17MNOPASS · priced
KXDIAZOUT-MDC…AUG01Díaz-Canel leaves office before Aug 122 / 2529812,202NO?REJECT · live crisis
KXDIAZOUT-MDC…JUL01Díaz-Canel leaves office before Jul 14 / 51,07486,238NOPASS · priced/thin
KXELECTIRAN…JUL01Iran holds presidential election before Jul 10 / 330731,284NOPASS · priced/thin
KXLAIOUT-LCHI…JUL01Lai Ching-te leaves office before Jul 10 / 401,806NOREJECT · illiquid
KXFETTERMANSWITCH…AUG01Fetterman leaves Democratic party before Aug 10 / 608,419NOREJECT · illiquid
KXFETTERMANSWITCH…JUL01Fetterman leaves Democratic party before Jul 10 / 106,400NOREJECT · illiquid/priced
KXMILLSPLATNER…AUG01Mills endorses Platner before Aug 19 / 1602,255NOREJECT · spread/cluster
KXMILLSPLATNER…JUN29Mills endorses Platner before Jun 291 / 302,997NOREJECT · illiquid
KXPRATTCONCEDE…JUL01Spencer Pratt concedes LA mayoral primary16 / 225394,367NOREJECT · spread/ambig
KXAPCALLLAMAYOR…JUN30AP calls 2026 LA mayoral primary by Jun 30~0 / ~1001,0102,010REJECT · priced/illiquid
KXGENERICBALLOT…T5.2Generic ballot above 5.2% (Jun 26)55 / 58125304REJECT · coin-flip/illiquid
KXTRUMPNUMSTATES…E1–E8Distinct US states Trump visits in June (ladder)0–55≤19950–675REJECT · illiquid ladder

3. Top picks — detail and thesis

No picks this run. Every Stage-2 candidate failed either the liquidity filter, the "edge ≤ slippage/fees" test on deep favorites, or the "price already moved / can't out-forecast a live situation" reject gate. The detailed reasoning for each is in §4 — in a zero-pick run the rejects are the report.

The two markets that came closest to a pick, and why each failed, are worth reading as the core finding:

The near-miss that was a trap — KXDIAZOUT-MDC-26AUG01 · would-be BUY NO @ ~0.78

On the surface this screamed value: Díaz-Canel was re-elected to a second five-year term by Cuba's National Assembly and told NBC's Meet the Press on April 10, 2026 that "stepping down is not part of our vocabulary." A re-elected autocrat leaving office in six weeks looks like a ~95%+ NO — so NO at 78¢ looks like +17¢ of free edge. It is a trap. There is an active 2026 Cuban crisis: a fuel/economic collapse, protests across Havana, Artemisa, Holguín and Santiago on June 2–3, fresh US sanctions on Díaz-Canel (June 5), and an explicit Trump-administration regime-change campaign targeting Cuba "by end of 2026." The 22¢ YES is informed pricing of a live ouster scenario, not stale noise — and the price has whipsawed 3¢ ↔ 39¢ repeatedly over the last two weeks (it re-spiked to 28–36¢ on June 18–20). I have no edge over a fast-moving headline market, and 24h volume (298) fails the liquidity floor anyway. Reject.

The liquid one with no edge — KXTRUMPOUT27-27-26AUG01 · would-be BUY NO @ ~0.977

The only Elections market with real depth (OI 2.17M, 24h volume 13,216, 0.6¢ spread). But NO sits at 97.1–97.7¢ and has been rock-stable at 2–3¢ YES for the entire 14-day window. My estimate for Trump leaving office (death/removal/resignation) before Aug 1 is ~1.5–2%, so true NO ≈ 98% vs buying at 97.7 — an edge of roughly +1¢, inside the fee/slippage band. Risking 97.7¢ to make 2.3¢ is poor risk/reward, and the rules' death clause resolves at last traded price rather than YES, capping even the tail's payoff. Efficiently priced. Reject.

4. Recommended $1,000 portfolio

Allocation: $0 deployed · $1,000 (100%) cash. There is nothing to size, so the cluster and conviction tables are reported empty — by design, this is the auditable evidence that the screens cleared the book rather than the screens being skipped.

LineContractsCostMax payoutEV $% of capital
(no positions)0$0$0$00%
Cash reserve$1,000$1,000100%
Total deployed0$0$0$00%

Cluster exposure (cap: 15% / $150 per cluster)

ClusterCost% capitalvs 15% cap
— (none deployed)$00%OK

Conviction exposure (caps: HIGH 15% / MEDIUM 7% / LOW 3% per pick)

TierCost deployed% capital
HIGH$00%
MEDIUM$00%
LOW$00%

Risk profile

Execution notes & watchlist triggers

5. What I rejected and why

All 10 diligenced candidates below are machine-logged and shadow-tracked to settlement. If the rejects collectively perform as well as a real pick book would have, that itself is a finding about whether the Elections screen adds value this week. Cuts made specifically by the Stage-3.5 entry-band screen (coin-flips and the ladder) are flagged.

KXDIAZOUT-MDC-26AUG01 — Díaz-Canel leaves office before Aug 1 LIVE-CRISIS TRAP

Would-be side NO @ ~0.78. Reject: active 2026 Cuban crisis (protests June 2–3, US sanctions June 5, explicit Trump regime-change push) means the 22¢ YES is informed, not stale; price whipsawed 3¢↔39¢ over two weeks. No edge over a fast-moving headline market, and 24h volume 298 < 1,500.

my P(stays)≈73% · market NO≈77% · YES bid/ask 22/25 · depth ~410 · OI 12,202 · cluster diaz-canel-out

KXTRUMPOUT27-27-26AUG01 — Trump leaves office before Aug 1 PRICED

Would-be side NO @ ~0.977. Reject: efficiently-priced deep favorite. NO 97.1–97.7¢ vs true ~98% — edge ≈ +1¢, below slippage/fees. Risking 97.7¢ to make 2.3¢ is poor risk/reward; the death clause caps even the tail. >90¢ band, thin by construction.

my P(stays)≈98% · market NO≈97.5% · YES 2/3 · depth 7,736 · OI 2.17M · vol24h 13,216 · cluster trump-leaves-office

KXDIAZOUT-MDC-26JUL01 — Díaz-Canel leaves office before Jul 1 PRICED / THIN

Would-be side NO @ ~0.95. Reject: deep NO favorite at 95–96¢, edge ≈ +1–3¢ for him surviving 9 more days; rides the same diaz-canel-out cluster as the headline-risk Aug rung; vol24h 1,074 < 1,500.

my P(stays)≈95% · market NO≈95.5% · YES 4/5 · depth 14,632 · OI 86,238 · cluster diaz-canel-out

KXELECTIRAN-26JUL01 — Iran holds a presidential election before Jul 1 PRICED / THIN

Would-be side NO @ ~0.97. Reject: Pezeshkian was elected July 2024; the next regular presidential vote is ~2028. No election before Jul 1, 2026. NO already 97–99¢ (no edge room) and vol24h 307 < 1,500.

my P(no election)≈98% · market NO≈97–99% · YES 0/3 · OI 31,284 · cluster iran-election

KXLAIOUT-LCHI-26JUL01 — Lai Ching-te leaves office before Jul 1 ILLIQUID

Would-be side NO @ ~0.96. Reject: Taiwan's president, no transition path in 9 days. NO already ~96–99¢ (no edge room); vol24h 0 and a wide YES 0.1–4¢ / NO 96–99.9¢ market — unfillable.

my P(stays)≈99% · YES 0/4 · OI 1,806 · vol24h 0 · cluster lai-leaves-office

KXFETTERMANSWITCH-26MAY-AUG01 — Fetterman leaves the Democratic party before Aug 1 ILLIQUID

Would-be side NO @ ~0.94. Reject: Fetterman has repeatedly said he is staying a Democrat. NO already ~94¢+ (no edge room); vol24h 0 and a 0–6¢ YES market — illiquid, unfillable.

my P(stays Dem)≈96% · YES 0/6 · OI 8,419 · vol24h 0 · cluster fetterman-party-switch

KXMILLSPLATNER-26NOV03-AUG01 — Mills endorses Platner before Aug 1 SPREAD / CLUSTER

Would-be side NO @ ~0.84. Reject: vol24h 0 and a 7¢ bid-ask spread (>5¢ filter). Also rides the mills-platner-endorse thesis already covered by active published picks (JUL01 NO, PLATNERDROPOUT NO) — deploying here would compound a correlated, already-live cluster.

my P(no endorsement)≈85% · YES 9/16 · OI 2,255 · vol24h 0 · cluster mills-platner-endorse

KXPRATTCONCEDE-27JAN01-JUL01 — Spencer Pratt concedes LA mayoral primary before Jul 1 SPREAD / AMBIGUOUS

Would-be side NO @ ~0.82. Reject: novelty market; vol24h 539 < 1,500 and a 6¢ spread (>5¢ filter). "Concede" is ill-defined — a primary candidate has no obligation to formally concede — so resolution carries definitional ambiguity risk on top of illiquidity.

my P(no concession)≈78% · YES 16/22 · OI 4,367 · vol24h 539 · cluster la-mayoral-primary

KXGENERICBALLOTVOTEHUB-26JUN26-T5.2 — Generic ballot above 5.2% ENTRY-BAND CUT (COIN-FLIP)

Would-be side YES @ ~0.58. Reject (Stage 3.5): coin-flip band (~55–59¢) with no high-conviction edge — fails the entry-band screen outright. Compounded by vol24h 125 and OI 304, it is effectively untradeable.

my P≈55% · market≈57% · YES 55/58 · OI 304 · vol24h 125 · cluster generic-ballot

KXTRUMPNUMSTATES-26JUL01-E1…E8 — Distinct US states Trump visits in June (ladder) ENTRY-BAND CUT + ILLIQUID

Would-be side YES on the modal leg (E6 @ ~0.51). Reject (Stage 3.5 + liquidity): mutually-exclusive ladder with negligible liquidity (per-leg OI 50–675, vol24h ≤199). The modal legs sit in the coin-flip band with no defensible count edge, and the whole set is unfillable. Checked the leg-sum for a structural arb — none worth the spread.

E5 14¢ · E6 51¢ · E7 24¢ · per-leg OI 50–675 · vol24h ≤199 · cluster trump-states-visited

6. Sources

Data sources: Kalshi live mirror DB (categories, rules, 5-min snapshots) + Kalshi public API (live order books) + web search for primary-source news, as of 2026-06-22.

Disclaimer: Probabilities here are subjective estimates, not guarantees. Prediction-market contracts resolve to $1 or $0 — a losing position goes to zero. This is research and methodology documentation, not financial advice. Methodology v3: position size is gated by conviction tier and entry band, never by EV; edge_cents/ev_pct are recorded for calibration only because the resolved track record showed stated EV to be anti-predictive. This run deployed no capital.