Tools chosen: polymarket_direct polymarket_related kalshi_related claude_news claude_news gdelt_news wikipedia code_execution

Sub-questions (Fermi decomposition)

Planner reasoning

This is a base-rate question: the Smithsonian GVP records roughly 1-3 VEI 4+ eruptions per year on average, so 'exactly 0' is a low-probability tail outcome best modeled with a Poisson process using the 2000-2024 annual counts. Key research is (a) the current Polymarket price anchor, (b) the historical annual distribution of VEI 4+ counts including how many years since 1900/2000 had zero, and (c) whether any 2026 eruption has already occurred or is imminent, plus reporting/confirmation lag risk before the March 2027 resolution snapshot.

# Current state No VEI 4+ eruption has been publicly confirmed for calendar-year 2026 as of the latest research (spanning reports into mid-2026); several volcanoes (Lewotobi Laki-laki, Shiveluch, Hayli Gubbi) remain active but are characterized as VEI 2-3 or unrated ≥4. Resolution requires Smithsonian GVP's official 2026 cumulative VEI 4/5/6 tally, finalized by March 31, 2027 — so current market prices reflect probability under lag/uncertainty, not a confirmed outcome. # Timeline of key events - 2025-02-13: Alert Level 4 raised for Lewotobi Laki-laki, Indonesia (confirmed/reported, historically VEI 2-3 volcano) — Watchers.news. - 2025-06/07/10: Multiple major Lewotobi eruptions; activity described as decreasing by late Jan/Feb 2026 after last eruption 2025-10-18 (reported, volcano.si.edu). - 2025-11-23: Hayli Gubbi (Ethiopia) erupts explosively for first time in ~12,000 years; ash plume up to ~9 miles, flight cancellations (confirmed eruption; VEI rating not yet assigned ≥4) — Smithsonian Mag. - 2026 (ongoing into mid-year): Shiveluch (Kamchatka) sustained near-daily explosive activity, ash to ~12 km, aviation color code raised in reporting around July 2026; not reported as a newly confirmed VEI 4+ event (reported, volcanodb.com/claude_news). - Late May 2026: USGS/GVP reports show no caldera systems with inflation/seismicity precursors to VEI 5/6 eruptions (reported, polymarket.com-sourced summary). - No dated event through mid-2026 has been assigned VEI 4+ by GVP for 2026 in available sources. # Event Will GVP confirm exactly 0 volcanic eruptions with VEI ≥4 worldwide in 2026 (resolved via GVP data as of March 31, 2027)? # Outcomes to forecast Yes (0 eruptions VEI≥4) / No (≥1 eruption VEI≥4) # Kalshi market anchor No direct Kalshi ticker found for this exact market (kalshi_related search returned only an unrelated "supervolcano by 2050" market at 20.1% YES, 61 contracts/day avg volume — not a valid comparator). Primary market data available is Polymarket-direct for this identical question: **current YES price 50.5%**, down sharply from a 90-day high of 73.5% (7-day change: -18pp; 30-day change: -13pp), total volume ~$499K. This is a meaningful downward drift, suggesting market participants are pricing in rising odds of a VEI4+ event as 2026 progresses. Note: a related multi-outcome Polymarket market ("How many VEI≥4 eruptions in 2026?") reportedly showed "0" outcome near 67% as of Jan 2026 per claude_news — inconsistent with the 50.5% figure from polymarket_direct; treat the 50.5% figure (direct market match) as authoritative, with the 67% figure as a stale/conflicting secondary reference. # Sub-question answers 1. **Polymarket YES price/volume** — 50.5% current, volume ~$499K, declining from 73.5% high (polymarket_direct). 2. **GVP 2000-2024 average VEI4+ count/year-by-year** — Not directly retrievable from tool output (dynamic table); GVP FAQ states VEI≥4 frequency has been "impressively constant for more than a century" (volcano.si.edu/faq). No exact numeric annual average was extracted. 3. **Zero-VEI4+ years since 1950/2000** — Not directly answered; qualitative evidence (near-annual VEI4+ occurrence 2021-2024 per Wikipedia list) suggests true zero years are rare but not quantified. 4. **Jan 2026 eruptions likely VEI4+** — None confirmed. Lewotobi (waning, historically VEI2-3), Hayli Gubbi (unrated, likely <4), Shiveluch (sustained activity, no fresh VEI4+ assignment) (claude_news). 5. **GVP VEI assignment lag/revision frequency** — Not explicitly quantified in research; GVP data finalized only through March 2027 for 2026 events, implying inherent lag; code_execution models assume 10-25% under-confirmation at snapshot. 6. **Poisson P(0) model** — At λ=1.5-2.5 (plausible range), raw P(0)=8-22%; adjusting for 10-25% confirmation lag pushes effective P(0) to ~11-32% (code_execution). # Key facts (high-confidence, factual) 1. [volcano.si.edu FAQ] VEI≥4 eruption frequency has remained "impressively constant" over the past century (no rising trend). 2. [volcanodb.com] Historical Holocene record: 754 eruptions ever reached VEI≥4 across dated record (~11,700 years), implying long-run average well under 1/decade if evenly spread, but modern well-monitored era shows near-annual occurrence. 3. [Wikipedia] VEI≥4 events occurred in 2021 (Semeru, Fukutoku-Okanoba), 2022 (Hunga Tonga, VEI5-6), 2023 (Sheveluch), 2024 (Ruang, Lewotobi Laki-laki) — four consecutive years with at least one confirmed event. 4. [ScienceDirect] Ruang's prior VEI4 eruption was in 2002; individual volcanoes can go decades between VEI4 events, but ~1,214 active Holocene volcanoes globally sustain near-annual global VEI4+ occurrence. 5. [Description] Resolution uses GVP "Eruptions by Year" cumulative 2026 figures, finalized as of March 31, 2027. # Cross-market signals - Kalshi related: No matching ticker; only tangential "supervolcano by 2050" market (20.1% YES) — not informative for annual VEI4 base rate. - Polymarket (this market): 50.5% YES, down from 73.5% (30d/7d declining trend, high volume ~$499K) — market increasingly leaning toward "No" (i.e., at least one eruption). - Polymarket related: "Vesuvius eruption 1+ VEI in 2026" priced 92.5% NO (quiet); "Major volcano VEI≥6 in 2026" market suggests near-zero probability for largest events, consistent with no caldera precursors. # Analyst opinions and speculation - claude_news synthesis: as of early-mid 2026, no VEI4+ eruption officially confirmed; multiple active volcanoes (Shiveluch, Lewotobi, Hayli Gubbi) are candidates but unconfirmed at ≥4. - Model-based view (code_execution): raw Poisson P(0) at plausible λ (1.5-2.5) is 8-22%, below the 50.5% market price — suggesting market may be overpricing "Yes" if using naive historical base rates, but lag-adjusted effective P(0) could reach 11-32%, still below 50.5%. # Directional lean per outcome - **Yes (0 eruptions)**: Supported by no confirmed VEI4+ event through mid-2026, declining Lewotobi activity, no caldera precursors (USGS/GVP). Opposed by: near-annual historical base rate (4/4 years 2021-2024 had ≥1 event), Shiveluch's ongoing intense activity, GVP's confirmation lag potentially adding events post-hoc, and Poisson model implying raw P(0) well below current 50.5% price. - **No (≥1 eruption)**: Supported by strong historical base rate and multiple active high-risk volcanoes (Shiveluch, Lewotobi) still ongoing; market's declining YES trend (73.5%→50.5%) suggests growing belief in this outcome as 2026 unfolds. # Gaps / unknowns - No confirmed exact GVP year-by-year VEI4+ counts extracted (data table not retrieved). - Unclear whether Shiveluch's 2026 activity will be assigned a fresh VEI4+ episode (historically the most frequent VEI4+ producer, 43 lifetime events). - No visibility into second half of 2026 (research appears to extend only to ~July 2026). - Uncertain how much of the 50.5%→73.5% price drop reflects genuine new eruption news vs. speculative repricing. # Calibration anchors - Polymarket YES price (this market): 50.5%, trending down from 73.5% (anchor). - Poisson model: raw P(0)=8-22% (λ=1.5-2.5); lag-adjusted 11-32%. - Historical pattern: VEI≥4 events confirmed in each of 2021, 2022, 2023, 2024 (0-for-4 recent zero-years), suggesting "No" outcome historically more common.

1. Both forecasts converge suspiciously to identical 0.40/0.60 despite claiming "independent" reasoning, and both settle roughly midway between the Poisson lag-adjusted range (11-32%) and the market price (50.5%) without justifying why a simple average/interpolation is the right method rather than, say, weighting more heavily toward the model given known lag/undercounting bias. 2. Neither forecaster grapples with the conflicting Polymarket signal noted in the brief — the related "How many VEI≥4 eruptions in 2026?" market reportedly showed "0" near 67% as of Jan 2026, which if not fully "stale," would suggest the 50.5% figure may reflect late-breaking specific news (e.g., Shiveluch escalation) rather than a stable base rate signal; this divergence deserves explicit discussion of which market is more reliable and why. 3. Both treat "no confirmed VEI4+ through mid-2026" as significant evidence for Yes, but the brief explicitly notes research visibility "extends only to ~July 2026" — half the year remains completely unobserved, and given Shiveluch is "historically the most frequent VEI4+ producer, 43 lifetime events" and still undergoing "sustained near-daily explosive activity," this residual-year risk seems underweighted in both rationales. 4. Neither forecast addresses the GVP retroactive-assignment risk in enough depth: the brief flags that resolution isn't finalized until March 31, 2027, giving ~3 months post-2026 for GVP to upgrade an unrated eruption (e.g., Hayli Gubbi, Shiveluch) to VEI4+, which is a distinct tail risk from "new eruptions in H2 2026" and could push probability of No even higher than the base-rate models capture. 5. The historical base rate cited (VEI4+ in every year 2021-2024) is only a 4-year sample presented as decisive, yet the brief also notes "Holocene record: 754 eruptions ever reached VEI≥4... implying long-run average well under 1/decade if evenly spread" — this tension between modern near-annual occurrence (likely a monitoring-density artifact) and long-run sparse rates is not reconciled, and forecasters should clarify whether they're using the modern-era rate (justified) or risk conflating monitoring improvements with true eruption frequency increases. 6. Both forecasters land at exactly 0.40, notably below the 50.5% market anchor, but provide no explicit sensitivity check on how much weight to put on market efficiency (high volume ~$499K, real-money Polymarket) versus their own model — given this is described as a "direct market match" for the identical question, deviating 10+ points from a liquid market price warrants stronger justification than offered.

40% (low-medium confidence)

Yes: 40% No: 60%

Both forecasters landed at 0.40 Yes, and while the convergence is suspicious, an independent conditional calculation lands in the same neighborhood: with a modern-era rate of ~1.2-2 VEI4+ events/year, roughly half of 2026 already elapsed without confirmation implies P(no new event in H2) ≈ 0.45-0.55, which must then be discounted by a ~20-30% chance that an already-occurred but unrated 2026 event (Shiveluch's sustained near-daily explosive activity, Hayli Gubbi, late Lewotobi episodes) is retroactively assigned VEI≥4 before the March 31, 2027 finalization — yielding ~0.32-0.45 for Yes. The critique correctly notes the retroactive-assignment channel and the unobserved second half are underweighted, which argues below the 50.5% Polymarket anchor; the modern near-annual base rate (2021-2024 all had ≥1 event) is the correct reference class rather than the Holocene-average artifact, reinforcing this. Offsetting factors keeping me from going as low as the naive Poisson (11-32%) are genuine market efficiency (~$499K real-money volume on an identical question), the possibility that GVP simply never assigns a 4 to any 2026 event in a genuinely quiet year, and the absence of caldera precursors. I finalize at Yes=0.40, roughly 10 points below the market anchor but above the pure model.

Total pipeline time: 302.0s