Kalshi Science & Technology — Mispricing Audit
Report date 2026-06-19 · horizon 45 days · portfolio capital $1,000 · category Science and Technology
Final picks: 2 (both on one underlying — SpaceX's June launch count). The category is thin this week; I am not padding.
One-line thesis. SpaceX has flown 9 Falcon 9 missions in the first 19 days of June 2026 (~0.47–0.56/day) and has ≥4 more already manifested for June 20–30. The Kalshi launch-count ladder prices “exactly 14” at ≈69% and assigns only ≈8.5% to ≥16 and ≈3.5% to ≥17. That upper tail is too thin given the demonstrated cadence. Today's rally already re-priced the safe rungs (>12, >13), so the remaining edge sits in the cheap, un-moved tail: KXSPACEXCOUNT-26JUN-15 and -16 YES.
1 · How this was researched
- Universe. Pulled all active markets in trading_events.category = "Science and Technology" closing within 45 days via the read-only DB mirror — 220 markets across ~25 series (AI-model leaderboards, GPU compute-price ladders, FDA approvals, nuclear criticality, Ebola/alien novelty markets, SpaceX/Blue Origin launches).
- Liquidity screen. Joined each ticker to its most-recent market_snapshots row and cut anything with volume24h < 1,500, bid-ask spread > 5¢, already fully priced (YES bid ≥ 95¢ / ask ≤ 5¢), or "what will X say"-style noise. Most of the 220 died here — the category is dominated by 1-cent leaderboard longshots and 0-volume GPU/biotech ladders.
- Exclusions. Dropped every ticker already live in the feed from prior runs (KXAISPIKE-*, KXFRONTIER-*, KXGROK-*, KXLLM1-26JUN30-A, KXSPACEXCOUNT-26JUN-14, KXCLAUDE-MYTH-*, KXSPACEXSTARSHIP-13-*, etc.) so nothing here duplicates an existing pick.
- Diligence. For survivors: pulled full rules + rules_secondary (resolution source) from the DB and Kalshi's public API, 14-day intraday price history, live order-book depth, and primary-source news. The launch-count thesis was built off the canonical Wikipedia April–June 2026 spaceflight list (citing nextspaceflight.com) and Spaceflight Now.
- Resolution source check. KXSPACEXCOUNT resolves on FAA source-agency launch data (per rules_secondary) — i.e. FAA-licensed SpaceX launches. In June 2026 that is effectively the Falcon 9 manifest; Starship is grounded pending the Flight-12 mishap review and contributes nothing.
2 · Markets reviewed
YES bid/ask are live top-of-book from Kalshi's public API at ~15:30 UTC on 2026-06-19; vol24h from the DB mirror's latest snapshot. The full 220-market universe is dominated by sub-1,500-volume leaderboard/compute-price ladders that never reach this table.
3 · The picks
Read this before the picks. Both picks express the same view — that the market's "exactly 14 launches" consensus under-weights SpaceX's habit of adding un-manifested Starlink flights late in the month. They are positively correlated and both resolve to zero in the modal (≤14-launch) world. Size them as one high-variance tail position, not two independent bets. Conviction reflects how defensible each edge is — the resolution is clean and countable, but the edge rests on a cadence model that disagrees with a liquid market, so neither earns HIGH.
Pick 1 — KXSPACEXCOUNT-26JUN-15 · BUY YES @ 0.11 limit · MEDIUM CONVICTION
“More than 15 SpaceX launches in June 2026” (i.e. ≥16). Live YES 8 / 9¢ · last 8¢ · closes 2026-07-01 14:00Z (expected resolution 2026-06-30).
Current ask 9¢My true prob ≈21%Edge at 11¢ fill ≈+10¢EV ≈+91%Launches needed 7 more in 11 days
Mispricing thesis. SpaceX has flown 9 Falcon 9 missions through June 19 (June 3, 4, 7, 8, 11, 12, 15, 17, 19). Reaching 16 needs 7 more launches in the 11 remaining days — a rate of 0.64/day. That is above their sustained average, so it is genuinely a tail. But the market prices it at only 8.5%, which is hard to square with a June run-rate of 0.47–0.56/day and a back-third of the month that is exactly when SpaceX clears its Starlink backlog. A defensible Poisson on the honest full-June rate (λ≈5.2 additional) puts ≥16 at ≈27%; I haircut that to ~21% for schedule-capping/under-dispersion. Even the haircut leaves the 9¢ ask more than double-cheap.
Evidence (cited below).
- 9 launches, June 1–19, each confirmed as Falcon 9 Block 5 in the Wikipedia April–June 2026 launch list (sourced to nextspaceflight.com): Starlink 17-47, 10-43, 17-47, 10-43, 17-44, a 12 Jun Starlink, 17-54, BlueBird 8–10, and NROL-179 (19 Jun, 08:40 UTC — already flown).
- ≥4 more already manifested for the remaining window: two Falcon 9s on June 21 (incl. Starlink 17-28), Starlink 17-45 on June 25, and Globalstar 2-R Mission 1 (June, TBD). A named manifest 11 days out is a floor — SpaceX routinely adds Starlink flights inside a week's notice.
- Fleet is healthy. No Falcon 9 grounding is in effect; the only active SpaceX stand-down is Starship (Flight 12 mishap, 22 May 2026), which does not touch the Falcon 9 line that drives this count (Spaceflight Now, 27 May 2026).
- The market itself rallied the lower rungs today — >13 jumped 69→84¢ on the NROL-179 launch — yet left >15 and >16 untouched, leaving an internally inconsistent kink (P(=14)≈69%, then a cliff).
Tail risk (the cleanest single thing that loses). A Falcon 9 anomaly / FAA stand-down anytime in the next 11 days — SpaceX was grounded three separate times across 2024–25 — would cap the count near 13–14 and send this to zero. Secondary: a quiet Eastern-Range / pad lull into the July-4 weekend, or Globalstar slipping to July, leaving them short of 16.
Limit 0.11Size 600 contractsCost ≈$66Max payout $600EV ≈+$60
Liquidity / history: YES-ask depth is thin — ≈148 contracts at 9¢ and ≈500 more at 11¢, so a 600-contract fill at an 11¢ limit stays inside the 3¢-slippage rule but cannot scale much further. Price has held a tight 4–10¢ band for 14 days on heavy volume (vol24h 7,080), i.e. the tail has not moved on today's launch news.
Pick 2 — KXSPACEXCOUNT-26JUN-16 · BUY YES @ 0.06 limit · LOW CONVICTION
“More than 16 SpaceX launches in June 2026” (i.e. ≥17). Live YES 3 / 4¢ · last 4¢ · closes 2026-07-01 14:00Z.
Current ask 4¢My true prob ≈12%Edge at 6¢ fill ≈+6¢EV ≈+100%Launches needed 8 more in 11 days
Mispricing thesis. The deeper-tail extension of Pick 1: ≥17 launches needs 8 more in 11 days (0.73/day). A genuine longshot, but the market's 3.5% is well below a Poisson read (≈16% on the honest rate) and my conservative ~12%. The same un-manifested-Starlink argument applies, just one launch further out. I keep this small and call it LOW conviction precisely because it leans hardest on the cadence model rather than on the named manifest.
Evidence. Same launch log and manifest as Pick 1. The contract has the deepest book in the ladder (~5,000 YES contracts available within 3¢ of best), which is why it can carry more dollars than the thinner >15 rung despite lower conviction.
Tail risk. Identical to Pick 1 but with less margin: any single missed launch (anomaly, weather stand-down, range conflict, Globalstar slip) is enough to keep the count under 17. This is the first to die in any slowdown.
Limit 0.06Size 2,000 contractsCost ≈$120Max payout $2,000EV ≈+$120
Liquidity / history: ≈970 contracts at 4¢, ≈560 at 5¢, ≈1,900 at 6¢ — a 2,000-contract fill blends to ~5–6¢ well inside the slippage rule. Price has sat at 1–4¢ for two weeks (vol24h 3,100); unmoved by today's lower-rung rally.
4 · Recommended $1,000 portfolio
This is a capacity-constrained deployment. The only genuinely mispriced contracts in the category are these two thin tails; pushing more capital in means paying through the edge (>3¢ slippage), which the screen forbids. So the bulk of the book stays in cash by necessity, not by choice — see execution notes.
Risk profile
- Worst case (most-likely): SpaceX finishes June with ≤15 launches → both picks resolve NO → lose the full $186 deployed (1.9% of capital). The market's modal outcome (exactly 14) lives here, so expect to lose the stake more often than not.
- Middle: exactly 16 launches → Pick 1 pays $600, Pick 2 expires worthless → +$534 net on the book.
- Best case: ≥17 launches → both pay → $2,600 gross, +$2,414 net (≈13× the stake).
- Concentration: 100% of deployed risk is a single underlying thesis — SpaceX's June Falcon 9 cadence. There is no diversifying second thesis in this category that survives the liquidity screen. The cash reserve is the risk control.
Execution notes
- Limit orders only. Post 600 @ 0.11 on -15 and 2,000 @ 0.06 on -16; do not chase. If -15 won't fill below 12¢, shrink it rather than pay up — the depth gap above 11¢ is real.
- Reserve ($814) is opportunistic, not idle. Two triggers to deploy it: (1) a second launch in any 48-hour stretch before June 30 (back-to-back days) re-rates the whole tail — add to -15/-16 on the move; (2) if -15 ever dips back to ≤6¢ on no news, scale in toward the depth cap.
- Watchlist (no capital today): -13 (>13) — revisit only if it falls back under 72¢; -12 (>12) as a near-riskless park if it ever offers >4¢ over fair.
- Hard invalidation — close both immediately if any of: an FAA Falcon 9 stand-down / mishap is announced; SpaceX has <3 launches between June 20–26 (cadence broke); or the count is mathematically capped (e.g., 13 flown by June 27 with no manifest left).
- No hedge taken. The natural hedge (the existing >14 YES pick already in the feed) is the lower rung of this same ladder and is not re-bought here.
5 · What I rejected and why
- KXSPACEXCOUNT-26JUN-14 (>14 / ≥15) — excluded to avoid a duplicate. This is already an active pick in the feed at 13/14¢. On my numbers it is the most under-priced rung of all (true P(≥15)≈33% vs market 13.5%), which corroborates this whole thesis — but re-emitting it would duplicate a live pick, so it stays out of the numbered list and out of picks.json.
- KXSPACEXCOUNT-26JUN-13 (>13) — edge eaten. Sat at 69¢ at noon; rallied to 81/84¢ this afternoon on the NROL-179 launch. At 84¢ vs my ≈78–80% true, it is now fair-to-rich. Classic "the price already moved on the thesis in the last 48h."
- KXSPACEXCOUNT-26JUN-12 (>12) — fairly priced. True P(≥13)≈93–95%; ask 93¢ leaves ≈1¢, below the fee/slippage floor. A safe park, not an edge.
- KXGPT-OPENB-26JUN26 (GPT-5.6 before Jun 26) — no defensible edge. Whipsawed 82→31→73→64¢ in 48h on an unconfirmed release rumor; OpenAI has published no GPT-5.6 announcement (last confirmed model is GPT-5.5, 23 Apr 2026). Pure release-timing speculation with no public-information edge, and the price is unstable — exactly what the screen is meant to skip.
- KXCRITICALITY-26AUG-VALAR — edge already realised. Ran from ~75¢ to 93/97¢ over the last 48 hours on (apparent) Valar Atomics progress news. Buying YES at a 97¢ ask leaves ≤3¢ of room for a 6-week binary on a startup reactor — the move happened without me.
- KXNEWGLENN-262-JUL / -AUG — priced to the floor. ~1¢ (July) and 1/3¢ (Aug); the market already treats a near-term New Glenn flight as a near-certain NO and I found no scheduled launch to argue otherwise. Nothing to add.
- FDA-approval ladder, other nuclear-criticality legs, GPU compute-price ladders, AI leaderboard longshots — illiquid. Almost all carry vol24h ≈ 0 and 10–90¢ spreads (e.g. the H100/H200/B200/A100 "compute price above $X" rungs, KXFDAAPPROVE-*, KXCRITICALITY-* minor legs). Even where a price looks off, you cannot size without paying double-digit cents of slippage. Watchlist at best.
- Everything already in the feed (KXAISPIKE-*, KXFRONTIER-FRON-*, KXGROK-GROK5-*, KXLLM1-26JUN30-A, KXCLAUDE-MYTH-*, KXSPACEXSTARSHIP-13-*, KXCODEAI-*) — excluded up front to avoid duplicate feed entries.
6 · Sources