Kalshi Elections — Mispricing Audit
Report date: June 15, 2026 · Category: Elections · Horizon: markets closing within 45 days · Capital: $1,000
This was a thin week for the Elections book. Of 27 active markets closing inside the horizon, almost all are either fully priced (the AP-call markets at 99¢), illiquid with no tradeable side (Iran/Taiwan leave-office contracts), or efficiently priced around a knife-edge (Trump approval). One thesis survived diligence: the June 17 FOMC will almost certainly hold rates, but a fractured committee makes a unanimous hold unlikely — so "hold + ≥1 dissent" at ~36¢ is cheap and "hold + 0 dissents" at ~64¢ is rich. Both picks express that single view; the report is candid that the portfolio is therefore concentrated, and holds a large cash reserve rather than padding with thin-edge trades.
1. How this was researched
- Universe. Pulled all
active markets whose event category = 'Elections' closing between Jun 15 and Jul 30, 2026 from the live Kalshi mirror (join trading_markets → trading_events). 27 markets.
- Screen. Cut fully-priced markets (YES bid ≥95¢), markets with no tradeable side in the live orderbook, "what will X say"-style noise, and contracts already live as picks from prior runs. Liquidity here is thin across the board —
volume24h tops out at ~1,500 — so I leaned on open interest and live order-book depth as the better liquidity proxies rather than the standard 1,500-share volume floor, and say so explicitly for each pick.
- Diligence per candidate. Resolution rules text from the mirror; 14-day intraday price history from market_snapshots; live top-of-book depth from the Kalshi public API; and primary-source news (Fed reporting, polling aggregators) via web search/fetch.
- Probability & edge. A subjective true-probability estimate with explicit reasoning and stated tail risks; edge quantified in cents and expected return at the achievable fill.
- Reject test. Is the rules text unambiguous? Is the evidence public? Has the price already moved on the thesis in the last 48h? Can I fill without ≥3¢ slippage? Anything failing one of these is in §5, not the picks.
- Excluded as already-live picks (would duplicate the feed): KXTRUMPCOLLINS-26JUN16, KXMILLSPLATNER-26NOV03-JUL01, KXPLATNERDROPOUT-26.
2. Markets reviewed
Prices are YES bid/ask from the latest snapshot (Jun 15) or live orderbook. Several near-identical strikes (TRUMPNUMSTATES E1–E8, three AP-call dates) are collapsed to one row. Tickers already live as picks are not shown.
3. The picks
Ordered by conviction. Conviction reflects how defensible the edge is (source quality + clarity of resolution rules + order-book depth), not the raw EV. Both picks express one thesis — see the concentration note in §4.
Pick 1 — KXFEDCOMBO-26JUN-0-T0 BUY YES @ 0.38 High conviction
Market: June FOMC — "No change AND >0 dissents"
Now: 31/36¢
True prob: ~58%
Edge: +22¢
Return at fill: ~61%
Resolves: Jun 17, 2026
Mispricing thesis
The market is ~97% sure the Fed holds on June 17 (the 25bp-cut legs price ~4–8¢ combined), and I agree. The disagreement is on dissents: this leg ("hold with at least one dissent") trades ~34–36¢, implying only a ~35% chance the hold is anything but unanimous. Given how openly split this committee is, that is too low. A unanimous hold requires every voter — including a governor who has dissented repeatedly for cuts — to fall in line at the same meeting. I put that at well under half, so ≥1 dissent at ~58%.
Evidence
- The committee just dissented four ways. The April 29, 2026 FOMC produced an 8–4 vote — the most dissents on a single decision since October 1992. Gov. Stephen Miran dissented for a 25bp cut; three regional presidents objected to the statement's easing-bias language. (CNBC, Apr 29 2026; DeFiRate.)
- Miran is a permanent voter. As a Board governor he votes at every meeting, and his dissents are substantive (he wants cuts), not collegial. For this leg to lose given a hold, even Miran has to stop dissenting — unlikely on a policy he's pressed repeatedly.
- The macro backdrop forces a split. May headline CPI ran 4.2% y/y and unemployment is ~4.3%; futures now lean toward a year-end hike, not a cut. A hold leaves both the hawks (who want tightening) and Miran (who wants easing) unsatisfied — exactly the setup that produces dissents. (Kraken economic brief, Jun 10 2026.)
- New chair ≠ guaranteed unity. June 17 is Kevin Warsh's first meeting as chair. First meetings sometimes draw consensus, but that courtesy doesn't bind an entrenched policy dissenter, and a hawkish Warsh holding rates does nothing to win over the doves. (DeFiRate; Kraken.)
Tail risks
- Cleanest way to lose: a "new-chair honeymoon" unanimous hold — Warsh rallies the committee and even Miran abstains from dissenting. Possible, but it cuts against four-way April dissent and Miran's track record.
- Surprise cut. If the Fed cuts (this leg requires "no change"), it resolves NO regardless of dissents. Market-implied ~4%; baked into the 58% estimate.
- Definitional: low. "Dissents >0" resolves off the official recorded FOMC vote — there is no judgment call.
Numbers · liquidity · price history
- Fill: best YES ask 36¢ with ~500 contracts resting (the next offer gaps up to ~68¢), so size the immediate clip near 500 and scale the rest with limit orders ≤38¢ over the 48h into the meeting. OI 7,958; vol24h 1,360.
- EV: at 36¢, fair ~58¢ → +22¢ edge, ~61% return; payout $1.00. At a 38¢ limit, edge +20¢.
- History: traded ~30¢ in early June, sagged to ~18¢ on Jun 10, then re-rated to ~34¢ by Jun 15 on rising volume (vol24h ~1,400). The thesis is partly recognized — but the last-48h move is flat (~+1¢) and 34–36¢ is still far below ~58¢, so material edge remains.
Pick 2 — KXFEDCOMBO-26JUN-0-0 BUY NO @ 0.43 Medium conviction
Market: June FOMC — "No change AND 0 dissents"
Now (YES): 59/66¢
True P(YES): ~38%
Edge (NO): +20¢
Return at fill: ~48%
Resolves: Jun 17, 2026
Mispricing thesis
The mirror image of Pick 1, and a slightly broader bet. "Unanimous hold" trades ~64¢ YES; the same evidence says a clean, dissent-free hold is a minority outcome (~38%). Buying NO at ~42¢ wins on any result except a unanimous hold — i.e. the likely hold-with-dissent and any surprise cut — so NO fair value is ~62¢.
Evidence
- Identical to Pick 1: April's record four dissents, permanent dissenter Miran, a 4.2% CPI backdrop splitting hawks from doves, and a new chair who can't satisfy both wings with a hold.
- Why it's labelled medium, not high: it is the same underlying thesis as Pick 1 (positively correlated, not a hedge). I down-weight conviction to flag that, and to discourage double-sizing the dissent view. The edge itself is well-grounded; the structural concentration is the reason for the lower badge.
Tail risks
- Cleanest way to lose: the same unanimous-hold scenario that sinks Pick 1. If that happens, both picks lose together — the portfolio's primary risk (§4).
Numbers · liquidity · price history
- Fill: NO is bought against the resting YES bids — ~1 contract at the 38¢-equivalent, then real depth (~500) around a 42¢ NO cost; set a 43¢ limit. OI 9,594 (the deepest leg of the set).
- EV: at 42¢, NO fair ~62¢ → +20¢ edge, ~48% return; payout $1.00.
- History: YES ran 75–81¢ on Jun 8–10, then fell to ~65¢ by Jun 15 as the dissent leg re-rated — the same repricing seen from the other side, confirming the move is dissent-driven.
4. Recommended $1,000 portfolio
Deployed $501 of $1,000. Blended EV ≈ 55% on capital at risk; expected dollar edge ≈ $275 if my probabilities are right. Costs use the signal prices (36¢ / 42¢); filling up to the stated limits trims edge by ~2¢/contract.
Why the reserve is ~50%, not 10%
This is deliberate, not laziness. The category is genuinely thin this week and my one defensible thesis is liquidity-capped (~500 contracts per leg at good prices before the book gaps). Forcing the remaining capital into the rejects below would mean either paying >3¢ slippage or buying near-zero-edge trades — i.e. padding. The reserve is dry powder for (a) scaling the FOMC legs via patient limits as depth refreshes into Jun 17, and (b) the July election calendar.
Risk profile
- Worst case (~38% — unanimous hold): both picks lose, −$501. The two picks are the same thesis, so they fail together — this is the dominant risk and the reason deployment is capped at ~half of capital.
- Best / most-likely case (~58% — hold with ≥1 dissent): both win. Pick 1 returns +$480 (750×$0.64), Pick 2 +$319 (550×$0.58) = +$799 on $501.
- Surprise cut (~4%): Pick 1 loses (−$270), Pick 2 wins (+$319) → roughly +$49 net — the one scenario where the two legs partially offset.
- Concentration: 100% of deployed capital is on a single event (June FOMC) and a single variable (dissents). There is no diversifying second thesis this week — I rejected the candidates that could have provided one (Cuba, approval) on edge grounds rather than manufacture diversification.
Execution notes
- Limit discipline. Pick 1: bid 36–38¢, never chase past 38¢ (the book gaps to ~68¢). Pick 2: bid 42–43¢. Work both as resting limits over the 48h into the meeting; partial fills are fine.
- Same-thesis cap. Treat Picks 1 + 2 as one position for sizing. Do not add a third FOMC leg on top — that is the same bet a third time.
- Watchlist trigger. KXTRUMPVH-26JUN19-T39.9 — if the live VoteHub average moves clearly above ~41% (a >1pt cushion over the 39.9 line) before Jun 19, YES at ~80¢ becomes a defensible add; at ~40% it stays a pass.
- What invalidates the thesis. Any credible pre-meeting signal that Warsh has lined up a unanimous hold (e.g. Miran publicly softening, or reporting of consensus) — exit both. Resolution is fast (Jun 17, ~2pm ET), so there is little time decay to manage.
- Hedge. None needed beyond the natural partial offset on a surprise cut; the position is small relative to capital by design.
5. What I rejected and why
- KXTRUMPVH-26JUN19-T39.9 — Trump approval >39.9% on Jun 19 (~79¢ YES). VoteHub — the resolution source — currently reads ≈40%, i.e. essentially on the line. Broader aggregates straddle it (≈38.6% blended; some trackers 37–42%). With the average sitting within ~0.1pt of the strike and only four days to go, the outcome is poll-noise, and the market's 79% already prices the distance-to-threshold about as well as I can. No defensible edge — tantalizing proximity, but not actionable.
- KXDIAZOUT-MDC-26JUL01 — Díaz-Canel leaves office by Jul 1 (~5–6¢ YES). Tempting as a "buy NO at 95¢" near-certainty, but there is an active 2026 Cuba crisis with real replacement speculation (Al Jazeera, Mar 31; NPR, Apr 10 — he publicly refused to step down). That makes 5–6¢ a defensible price for a live tail, not an obvious mispricing, and my edge (≈+2–3¢) sits inside the noise — against an ugly 3–5% chance of losing 95¢. Rejected on risk/reward.
- KXELECTIRAN-26JUL01 & KXLAIOUT-LCHI-26JUL01 — Iran election / Lai Ching-te leaves office by Jul 1. Both are correctly priced near zero (no scheduled election in Iran before 2028-29; Lai's term runs to 2028). But there is no cheap NO side in either book — the YES bid is 0, so a NO buyer pays ~100¢. Real-but-untradeable. Watchlist only.
- KXPRATTCONCEDE-27JAN01-JUL01 — Spencer Pratt concedes the LA mayoral primary (~54¢). Resolution turns on whether a reality-TV candidate issues a concession statement — a "will X say" market explicitly out of scope, and an unforecastable coin-flip on a personal social-media post. No defensible edge.
- KXTRUMPNUMSTATES-26JUL01-E1…E8 — # of states Trump visits in June. A mutually-exclusive range set, but the legs trade essentially zero volume (E5/E6 at 32–38¢ with vol24h ≈0). Untradeable without moving the market against myself; no public edge on a travel-count tally.
- KXFEDCOMBO-26JUN-25C-0 / 25C-T0 — the two "25bp cut" legs. Correctly priced cheap (~4–8¢) and consistent with a ~96–97% hold. No tradeable NO side (empty YES book), and I have no edge calling a surprise cut. Pass.
- AP-call (KXAPCALLLAMAYOR-*), KXMILLSPLATNER-…-JUN29, KXFETTERMANSWITCH. Either fully priced (AP-call strikes at 98–99¢) or dead (0 volume, no two-sided book). Nothing to trade.
- Already-live picks (excluded up front to avoid feed duplicates): KXTRUMPCOLLINS-26JUN16, KXMILLSPLATNER-26NOV03-JUL01, KXPLATNERDROPOUT-26.
6. Sources