Scenario wins: SynapseSeer (58) Panshul42 (52) smingers-bot (30) AtlasForecasting-bot (20) preseen (16) cassi (9)
| Figure/Metric | Value | Source | Significance |
|---|---|---|---|
| Historical Crimea Conflict Events (2022-2024) | 227 total events | UK Home Office / ACLED | Establishes the low historical base rate (7.1/month). |
| Recent Drone Strike Frequency (May-June 2026) | ~40 per month | country_data_summary | Indicates a major regime shift in strike intensity. |
| Ukraine Middle-Strike Drone Procurement (2026) | 300% increase | Ukraine Defense Minister | Shows massive scaling of the primary strike asset. |
| June 26, 2026 Drone Wave | 660 drones | Russian Defense Ministry | Demonstrates the extreme volume of single-night assaults. |
| Daily Kerch Bridge Train Routes | 11 reduced to 7 | Wikipedia (June 2026) | Proxy for logistical degradation and security concerns. |
My analysis for the July 1 to August 1, 2026, forecast period is based on a fundamental shift in the operational environment in Crimea compared to historical baselines. Historical data from the UK Home Office and ACLED indicates that during 2022–2024, conflict events in Crimea and Sevastopol were relatively infrequent, peaking at approximately 12.8 events per month in late 2023 and dropping to roughly 5.5 per month by late 2024. However, indicators from the first half of 2026 suggest a massive escalation in both Ukrainian strike capability and intent.
Specifically, ACLED-cited reports for May and June 2026 already show a jump to nearly 40 drone strike events per month in Russian-controlled regions. The period immediately preceding our target window (June 21–26, 2026) saw a rapid-fire sequence of strikes against critical infrastructure, including multiple strategic bridges (Chonhar, Rozdolne), power substations in Sevastopol, the Tavriiska Thermal Power Plant, fuel depots in Kerch and Feodosia, and numerous air defense systems. This surge led to a state of emergency declaration on June 26.
The primary driver for the July 2026 window is the SBU’s announcement of a “40-day special operation” launched on June 25, 2026. This operation is explicitly designed to isolate the peninsula and destroy its logistics, and its timeline perfectly encompasses our target window. Because ACLED records individual rows for discrete locations and tactical incidents, a multi-target campaign involving daily or near-daily drone and missile launches will generate a significantly higher record count than seen in previous years.
I have modeled this using a mixture of scenarios. The central scenario (representing approximately 70% of the probability mass) assumes a sustained high-intensity campaign where Ukraine conducts multiple successful strikes across different Crimean municipalities, resulting in roughly 45–60 ACLED records over the 32-day period. I also account for an escalation scenario where the 40-day operation reaches a peak of activity, with highly distributed strikes (targeting rail, road, and air defense simultaneously) creating a density of 70–110 records. Conversely, a small probability is assigned to a lower-activity scenario (20–30 events) characterized by successful Russian air-defense adaptations or Ukrainian supply constraints. Overall, the evidence strongly points toward a month of unprecedented strike frequency for the Crimean peninsula.
Question: how many ACLED rows for Air/drone strikes + Shelling attacks + Other explosive in admin1 Crimea + Sevastopol will be recorded for 2026-07-01 through 2026-08-01 (ACLED UI/dyadic rows)? Forecasters agree the target is an ACLED row-count (not weapons/claims) and that ACLED’s aggregation rules, possible “disrupted weapons use” codes, and lack of live API access in-session are key caveats. All teams weight a strong late‑June 2026 campaign signal and a broader 2026 uptick, so they model a bursty, overdispersed count process rather than a smooth Poisson rate: typical posteriors center in the mid‑60s to mid‑70s (medians ~64–75), with wide 90% intervals roughly 25–136 and a small nonzero tail above ~199. Differences are mainly in tail mass and regime-mixture weightings (some forecasters give more probability to a quick lull after June, others to a sustained campaign or escalation). Practical guidance common to all: prioritize pulling the exact ACLED rows for Crimea+Sevastopol if possible, model distinct date-location rows (not claimed drone numbers), and account for negative‑binomial/bursty variance and possible classification leakage when assessing the final resolution.
Forecast: Probability distribution:
We are forecasting the total ACLED Ukraine Conflict Monitor UI counts (or equivalently raw-row counts) for Crimea + Sevastopol, summing Air/drone strikes + Shelling attacks + Other explosive, for 2026-07-01 through 2026-08-01 inclusive (32 days).
Synthesis of analyst views:
My take:
This yields a distribution centered in the ~70–100 range, but with tails that allow outcomes from a low-activity month (~10–30) up to very high activity (~160–200+) and a small chance of exceeding the 199 open bound.
An amateur forecast would commonly make one of two mistakes: (1) anchor on dramatic drone/interception headlines and overcount by treating “drones” or “targets” as ACLED events (inflating totals into the several-hundreds), or (2) anchor on older, lower-intensity Crimea baselines and miss the current elevated regime (underpredicting into the teens). This forecast improves by explicitly modeling the ACLED-record granularity/aggregation and by using a regime-mixture view (possible persistence vs. tapering) to keep both a realistic center and a sufficiently wide uncertainty band, including a modest open-upper tail.
Forecast rationale (numeric):
— Iteration 1 — The forecasts broadly agree that Crimea is likely to see elevated ACLED-recorded explosive/strike activity during the July 1–August 1, 2026 window, driven by the reported late-June “drone blockade” and the strategic importance of the peninsula.
The collective view is that Crimea is expected to remain a high-activity strike environment in July 2026, with a plausible central range roughly in the tens to low hundreds, and a skew toward higher values if the drone campaign intensifies further.
— Iteration 2 — Across the forecasts, the core view is that Crimea is likely to see an elevated number of ACLED-recorded explosive/strike events during July 1–August 1, 2026, driven by a recent intensification in strikes.
The collective reasoning suggests an elevated but highly uncertain July count, with the most likely outcome somewhere in the mid-double digits to around 100 ACLED explosive/strike events, shaped by recent escalation, ACLED’s event-count methodology, and the possibility of bursty strike waves.
Based on the rationales provided, forecasters agree that the reporting period will feature an exceptionally high operational tempo. However, the final event count will be heavily shaped by ACLED’s strict coding methodologies and the restrictive information environment in occupied territories.
Strategic Context and Military Capabilities
Forecasters consistently highlight the “40-day intermediate- and long-range strike campaign” authorized by Ukrainian President Zelensky in late June 2026. Explicitly designed to impose a “logistics lockdown” on Crimea, this offensive perfectly encompasses the July forecasting window. Ukraine’s ability to sustain this barrage is driven by a massive scale-up in domestic drone and missile production, compensating for dwindling Western supplies. Furthermore, previous Ukrainian strikes have severely degraded Russian S-300 and S-400 air defense networks on the peninsula, significantly increasing the probability of projectiles breaching defenses. Forecasters also point to symbolic July dates—such as the NATO Summit and Russian Navy Day—alongside favorable summer weather as historical catalysts for intense, multi-domain swarm attacks.
ACLED Methodology Constraints
All rationales emphasize that ACLED’s specific data aggregation rules will heavily modulate the raw number of munitions fired into a much smaller recorded event count.
Censorship and Reporting Lags
Finally, the forecasters account for Russia’s aggressive digital censorship and state-of-emergency declarations in Crimea. Severe penalties for civilians filming strikes and the throttling of local social media channels suppress open-source reporting. This forces ACLED to rely on broader state claims, partisan networks, and satellite imagery (like NASA FIRMS). Forecasters agree that this lack of localized visual evidence—combined with standard processing lags before the final dataset resolution—will likely obscure smaller sabotage events and place a ceiling on the overall recorded count.
Question: How many ACLED-recorded explosive/strike events will occur in Crimea between July 1, 2026 and August 1, 2026? Final Prediction: Probability distribution:
Total Cost: extra_metadata_in_explanation is disabled Time Spent: extra_metadata_in_explanation is disabled LLMs: extra_metadata_in_explanation is disabled Bot Name: extra_metadata_in_explanation is disabled
Forecaster 1: Probability distribution:
Forecaster 2: Probability distribution:
The research summarizes that Ukraine substantially intensified strikes on Crimea through June 2026, with near-daily attacks on ports, fuel depots, power plants, rail infrastructure, airbases, and air‑defense systems. It lists many June incidents (e.g., attacks on Feodosia, Simferopol TPP, Saki, Kerch port, multiple S-400 and Pantsir systems, a deadly Kerch fuel depot strike) and reports a broader escalation trend: drone swarm and long‑range strike rates increased from late 2025 into 2026, Ukrainian drone and strike system production rose markedly, and Ukraine adopted sustained operations to degrade Russian air defenses and logistics hubs in Crimea. ACLED expanded its event coverage in 2024–25 to include Crimea and the research notes ACLED counts are event‑based (one record per reported event), which can differ from raw incident counts reported in news.
Using ACLED historical counts and the June 2026 surge as reference, the research offers rough quantitative scenarios for July 1–August 1, 2026: a conservative 24–36 ACLED-recorded events (if Crimea is 10–15% of multi-region weekly strike totals), a moderate 64–96 events (2–3 ACLED-recordable events/day), and a high 96–160 events (3–5/day). It emphasizes major uncertainties—ACLED methodology, possible clustering of strikes, Russian countermeasures, and reporting delays—and notes gaps in Crimea-specific monthly ACLED data. It recommends monitoring daily ACLED updates, Ukrainian military announcements, and changes in air‑defense effectiveness or strategic priorities.
Sources used (as cited in the provided research): news reporting cited throughout (references [1]–[10], [17], [2], [4], [5], [6], [7], [8], [9], [10]); ACLED Data Update (April 2025) [11]; ACLED weekly counts and Ukraine event summaries [13], [21]; capability and escalation assessments cited in the research [15], [18], [22]; Polymarket ceasefire market reference [19]. The provided research did not include verbatim URLs for these citations.
Based on extensive news reporting, Ukraine has dramatically escalated its strike campaign against Crimea throughout June 2026, with near-daily attacks documented:
1. ACLED Data Update (April 2025) [11]:
2. Weekly ACLED Counts (March 14-20, 2026) [13]:
3. Overall Ukraine Weekly Data (May 31-June 6, 2025) [21]:
Ukrainian Strike Campaign Intensification [18][22]:
Ukrainian Production Capacity (Early 2026) [22]:
Why Crimea is Heavily Targeted:
Russian Air Defense Degradation:
No specific markets found for this exact question (ACLED Crimea events July 2026). Related markets found:
The absence of specific prediction markets for ACLED counts suggests this is a specialized forecasting question without established public betting markets.
Based on available data, rough estimates for the 32-day period (July 1-August 1, 2026):
Conservative estimate: If Crimea represents 10-15% of the 57 strikes across 5 regions in March week [13], that’s ~6-9 events/week or 24-36 events for 32 days
Moderate estimate: June 2026 news shows attacks on at least 9 distinct days with multiple locations/types. If pattern continues at 2-3 ACLED-recordable events per day average: 64-96 events for 32 days
High estimate: If June’s intense campaign continues/escalates with 3-5 events daily: 96-160 events for 32 days
Key uncertainty: Actual ACLED event counts may differ significantly from news-reported strikes due to methodology (each database record = 1 event, potential for multiple records per tactical incident, or multiple explosions counted as single event).
The superforecaster should monitor: (1) daily ACLED Ukraine Conflict Monitor updates, (2) Ukrainian military announcements, (3) any shifts in strategic priorities or ceasefire negotiations, and (4) Russian air defense effectiveness metrics.
Conditioned on non-annulment, I accept the critic’s main correction: the relevant variable is not “how intense are Ukraine’s Crimea-related operations?” but “how many ACLED rows or UI counts land specifically in admin1 Crimea/Sevastopol and in the qualifying explosive categories?” That is narrower. Some “isolate Crimea” activity is on the land corridor, Krasnodar side, Black Sea, or Russia proper, and ACLED does not convert drone counts or explosions into one-for-one events.
The most important methodological update is ACLED’s own air/drone-strike FAQ: it records one strike event for same date / same location / same actors, while separate locations can produce separate events; intercepted strikes that are prevented are instead “Disrupted weapons use,” though strikes causing damage/casualties still count as air/drone strikes. This supports lowering the center from my original 70, but not collapsing it into the low tens because Crimea attacks are spread across many named localities and target types. (acleddata.com)
I still think the critic may be a little too bearish. ACLED’s recent references show the baseline has shifted materially upward: in March, ACLED recorded 57 Ukrainian air/drone strikes across the four occupied mainland regions plus Crimea in one week, and noted Ukraine’s campaign to degrade Russian air defenses; in April, it said strikes on occupied Crimea more than quadrupled versus the 2025 monthly average. (acleddata.com) That does not prove July will be 70+, but it makes a median near ordinary 2025 levels inappropriate.
The late-June situation also remains real evidence for a high July count. Crimea’s authorities declared an emergency after days/weeks of Ukrainian attacks disrupting power and fuel, and Zelenskyy approved a 40-day operation overlapping almost the entire July 1–August 1 window. (themoscowtimes.com) But I now put more weight on (1) possible operational pauses after late-June waves, (2) target shifting outside the resolving geography, and (3) ACLED aggregation suppressing headline-implied counts.
So I’m moving the median down from ~70 to the mid-60s, with a broad right tail. My modal world is roughly: July continues at an elevated but clustered tempo, yielding around 1.5–2.5 qualifying ACLED events/day on average. The 100–150 range remains quite live if ACLED codes multiple sites per wave and the 40-day campaign stays Crimea-heavy. Above 199 is possible but still a small tail because that would require sustained >6 qualifying records/day in a rear-area region.
Updated prediction:
Percentile 0.1: 0.49
Percentile 1: 4.49
Percentile 5: 11.49
Percentile 10: 18.49
Percentile 20: 29.49
Percentile 30: 40.49
Percentile 40: 52.49
Percentile 50: 64.49
Percentile 60: 78.49
Percentile 70: 95.49
Percentile 80: 118.49
Percentile 90: 151.49
Percentile 95: 181.49
Percentile 99: 240.49
Percentile 99.9: 350.49
Conditioning on the question not being annulled and resolving normally via the ACLED monitor or raw ACLED rows.
I accept the critic’s strongest methodological points. My earlier March anchor was too easy to overread: ACLED’s March update says 57 Ukrainian air/drone strike events across Donetsk, Kherson, Luhansk, Zaporizhia, and Crimea combined, not Crimea alone, so Crimea’s ordinary share could be materially below my assumed 10–20%. The cleaner Crimea-specific anchor is ACLED’s September 2025 update: in an active Crimea week Ukraine carried out 13 strikes, which annualizes/extrapolates to roughly the high-50s over this 32-day window. That same update says over 40% of nearly 800 Crimea events since February 2022 were Russian interceptions, which is a major reason not to map “hits on targets” or “drones launched” directly into rows. (acleddata.com)
I also agree that “Crimea isolation campaign” is broader than the resolution geography. Port Kavkaz, Krasnodar-side bridge logistics, Black Sea events, and some land-corridor attacks can be strategically central while resolving outside admin1 = Crimea/Sevastopol. This should pull the median and especially the low/mid outcomes downward versus a narrative-based count.
Where I still disagree with the critic is that the latest June evidence is not just “more news volume”; it is a genuine regime shift. ACLED had already said Crimea strikes more than quadrupled in April compared with the 2025 monthly average, and by late June independent reporting described strikes on Kerch/oil facilities, gas compressors, radar systems, fuel restrictions, and the peninsula becoming a primary focus of Ukraine’s mid-range strike campaign. (acleddata.com) Euronews and The Guardian also report a newly announced 40-day Ukrainian operation, a regional emergency in occupied Crimea, fuel shortages, and power cuts tied to Ukrainian attacks, which overlaps almost exactly with the forecast window. (euronews.com)
So I am moving down from my original median of ~69.5, but not all the way to a strongly conservative view. My current center is about 66 events. I fatten the left tail: a post-June pause, geographic leakage, ACLED aggregation, or shift toward Moscow/Krasnodar could plausibly leave the count in the 20s–40s. I also keep a substantial right tail: ACLED’s Russia-wide May counts show that, when Ukrainian strike campaigns are frequent and granularly reported, event counts can scale very quickly; repeated multi-location Crimea nights plus interception/debris records could still push the outcome above 150 and occasionally above 199. (acleddata.com)
Updated prediction:
Percentile 0.1: 0.49
Percentile 1: 3.49
Percentile 5: 12.49
Percentile 10: 20.49
Percentile 20: 32.49
Percentile 30: 43.49
Percentile 40: 54.49
Percentile 50: 66.49
Percentile 60: 79.49
Percentile 70: 95.49
Percentile 80: 116.49
Percentile 90: 147.49
Percentile 95: 175.49
Percentile 99: 225.49
Percentile 99.9: 315.49
Modal: 25-50 (27.7%) • frontier aggregate • 4m18s
Resolution sources/checks:
Edge cases:
frontier_1 - Modal: 50-75 (34.0%)
frontier_2 - Modal: 50-75 (18.0%)
frontier_3 - Modal: 0-25 (38.0%)
frontier_4 - Modal: 25-50 (52.0%)
Material notes
Guidance
Revision
| Bin | Probability |
|---|---|
| 0-25 | 16.5% |
| 25-50 | 27.7% |
| 50-75 | 22.9% |
| 75-100 | 13.2% |
| 100-124 | 7.4% |
| 124-149 | 4.8% |
| 149-174 | 3.0% |
| 174-199 | 1.8% |
| Above 199 | 2.7% |
wooley_rhino • run 5003ae1c • $2.92 • 4m18s
Extracted Percentile Values: {0.1: 1.2, 1.0: 2.5, 5.0: 6, 10.0: 9, 15.0: 12, 20.0: 15, 25.0: 18, 30.0: 21, 35.0: 24, 40.0: 27, 45.0: 30, 46.0: 31, 47.0: 32, 48.0: 33, 49.0: 34, 50.0: 35, 51.0: 36, 52.0: 37, 53.0: 38, 54.0: 39, 55.0: 40, 60.0: 45, 65.0: 50, 70.0: 57, 75.0: 65, 80.0: 75, 85.0: 88, 90.0: 105, 95.0: 130, 99.0: 180, 99.9: 235}
Model’s Answer: Step one is to translate the battlefield reporting into the exact ACLED metric. This question is not asking how many drones or missiles are launched, nor how many targets are claimed. It is asking how many ACLED records end up appearing in three specific sub-event buckets, summed across Crimea and Sevastopol over 32 days. That matters a lot, because ACLED often collapses a large salvo into one event when it is the same date, same location, same actors. So the recent reports of very large Ukrainian drone attacks do push the forecast upward, but much less than a naive “hundreds of drones means hundreds of events” reading would imply.
Step two is to set the base rate. I do not have direct historical ACLED region-level counts for exactly these categories in Crimea and Sevastopol, which materially lowers confidence. So I am leaning on the most recent and most relevant evidence: late June reporting shows an unusually active Ukrainian strike campaign against Crimean logistics, fuel, power, bridges, air defense, and naval infrastructure. Multiple reputable sources describe repeated attacks in the week immediately before the forecast window, including strikes around Kerch, energy facilities, transport links, and Sevastopol-area infrastructure. That makes a very low total less likely than it would have been a month or two earlier.
Step three is to discount the raw strike rhetoric into plausible ACLED rows. The best way to think about this is “distinct locality-day incidents,” not munitions. Over a 32-day period, if Ukraine maintains a moderate campaign focused on Crimea, one should expect many attack nights but not necessarily attacks every day, and on some active nights several Crimean locations may each generate separate ACLED records. Shelling attacks are likely a minority of the total. “Other explosive” probably contributes only a few unless there is a notable sabotage wave. So the count should be dominated by air/drone strike records.
My central case is that the late-June surge partially persists through July, but not at peak intensity every night. That points me to a median in the low-to-mid 30s. Put differently, I think the most likely regime is roughly around one qualifying ACLED event per day on average across the two admin1s, plus some clustering and multi-location nights.
A low outcome would happen if the late-June burst was mostly a short campaign peak, Russian air defenses adapt quickly, weather reduces drone operations, Ukraine shifts scarce long-range assets toward mainland Russia or frontline support, or ACLED aggregates a surprising number of incidents into fewer rows. In that world, totals can fall into single digits or low teens. I put the chance of a genuinely low outcome, meaning under 10, at around 10%.
A high outcome would happen if Ukraine sustains a concentrated “isolate Crimea” effort for most of July, repeatedly hitting multiple distinct locations per night, especially Kerch, Sevastopol, western Crimea, airfields, depots, radar sites, and transport nodes. ACLED coding would then split many of those by locality/date, producing a much fatter upper tail than casual observers might expect. I put the chance of a high outcome above 80 at around 15%, and above 120 at around 5%.
The assumptions doing the most work here are fairly clear. The first is that the late-June strike tempo is not just a brief spike; I think there is about a 35% chance this assumption is too strong, which would pull the forecast down materially. The second is that ACLED coding behaves in the usual aggregated way rather than splitting incidents unusually finely; I assign about a 25% chance this assumption is materially wrong, which would mostly raise the upper tail. The third is that Crimea remains a priority target set relative to mainland Russian energy and logistics targets; I assign about a 30% chance that this is wrong enough to matter, which would reduce the middle of the distribution.
Because I lack authoritative historical month-by-month ACLED counts for this exact regional slice, I am keeping the distribution fairly wide. My 90% interval is intentionally broad, and I am leaving a meaningful but small extreme right tail because resolution is by finalized ACLED data, where coding choices can produce more variation than the news flow alone suggests.
Percentile .1: 1.2 Percentile 1: 2.5 Percentile 5: 6 Percentile 10: 9 Percentile 15: 12 Percentile 20: 15 Percentile 25: 18 Percentile 30: 21 Percentile 35: 24 Percentile 40: 27 Percentile 45: 30 Percentile 46: 31 Percentile 47: 32 Percentile 48: 33 Percentile 49: 34 Percentile 50: 35 Percentile 51: 36 Percentile 52: 37 Percentile 53: 38 Percentile 54: 39 Percentile 55: 40 Percentile 60: 45 Percentile 65: 50 Percentile 70: 57 Percentile 75: 65 Percentile 80: 75 Percentile 85: 88 Percentile 90: 105 Percentile 95: 130 Percentile 99: 180 Percentile 99.9: 235
Crimea’s fixed geography and concentrated military-logistics infrastructure create persistent strike incentives.
Limited distinct targets and seasonal logistics cycles constrain the number of auditable incidents per month.
Recent interdiction efforts have elevated pressure on fuel, power, transport, air-defense, and port nodes.
Russian air defenses, repair capacity, ammunition availability, and prioritization of fronts actively modulate strike outcomes.
A sustained, focused Ukrainian campaign against logistics and ports would cluster recorded incidents across multiple locations.
Conversely, improved concealment, rapid repairs, or defensive suppression can decouple operations from recorded event counts.
Event-recording and coding conventions systematically compress many sorties and impacts into fewer auditable events, widening right-tail uncertainty.
Political suppression of reporting, post-hoc event revision, and episodic escalation spikes are the dominant sensitivities to the projected monthly count.
Forecast: median = 57.05 ACLED-recorded explosive/strike events
I interpret this market exactly as a count of ACLED rows in Crimea plus Sevastopol, restricted to the sub-event types named in the prompt, for 2026-07-01 through 2026-08-01 inclusive. My best single-number forecast is 26, and my modeled mean is 28.
The baseline is not a quiet Crimea. ACLED said that by 25 March 2026 it had already recorded 3,364 Ukrainian strike events in Russia since the start of 2026, and in the week of 14-20 March it recorded 57 Ukrainian air/drone strikes across occupied Donetsk, Kherson, Luhansk, Zaporizhia, and Crimea, including a hit on a military electronics facility in Sevastopol. That means Crimea entered summer from an already elevated strike environment, not from a low base. (acleddata.com)
The campaign then shifted harder toward isolating Crimea. The Kyiv Independent reported that since April Ukraine has systematically targeted ferries, bridges, fuel depots, gas facilities, and electrical infrastructure tied to the peninsula, while Le Monde described a focused campaign since May against military vehicles, fuel tankers, the Dzhankoi rail hub, and bridges connecting Crimea to the mainland. ISW likewise assessed in late June that Ukrainian strikes on transport and oil infrastructure were already disrupting Russian ground lines of communication and worsening logistics in Crimea. (kyivindependent.com)
Late June shows the current operational tempo clearly. On 20 June, Ukrainian forces reportedly hit four gas compressors in Crimea and the Hlibivka underground gas storage facility. On 21 June, reports described strikes on oil terminals and radar systems around Kerch and on both sides of the Kerch Strait. On 23-24 June, reports described hits on a railway bridge, the Kerch thermal power plant oil storage, a western Crimean substation, an LNG distribution site in Simferopol, Saky and Hvardiiske airfields, Kerch-area air defenses, Simferopol Power Station, and Sevastopol’s main electric substation. On 26 June, Ukraine said it hit military supply vessels, a ferry, and air-defense systems in Kerch. These reports do not map one-for-one to ACLED rows, but they strongly suggest that multi-record strike days are now common. (kyivindependent.com)
The sustainability signal is also strong. The Kyiv Independent quoted RAND’s Michael Bohnert saying the Ukrainian campaign looked sustainable because the drones used are cheap and domestically producible, while ISW assessed that the range, volume, and intensity of Ukrainian long-range strikes have increased significantly since March 2026. Russian-installed authorities have responded with fuel-sale suspensions, major power outages, and eventually a regional state of emergency in Crimea, which is consistent with a campaign that is biting and likely to continue. (kyivindependent.com)
The strongest near-term reason not to fade the pace is political: Zelensky approved a 40-day campaign on 25 June aimed at pressuring Russia, and that window overlaps almost the entire resolution period from 1 July through 1 August. (kelofm.com)
My quantitative translation is an inference from that evidence. I expect roughly 10-14 active strike days in the 32-day window, with about 2.0-2.5 ACLED-countable rows on an average active day in Crimea and Sevastopol, implying a naive range around 20-35. I keep the center near the upper half of that range because of the 40-day campaign and the late-June clustering, but not higher still because some of the broader interdiction campaign lands outside the resolution geography and because ACLED can aggregate tactical activity into fewer event records than the raw number of munitions or aimpoints. That leads me to a mean forecast of 28 and a median around 26. I encode uncertainty with a negative binomial distribution with mean 28 and dispersion parameter r = 7, which allows for clustered bursts and quieter spells. Main downside risks are Russian adaptation, weather, or a shift of Ukrainian effort toward deep Russia rather than Crimea; main upside risk is that late-June tempo becomes the new normal and ACLED splits multi-target waves into several rows. This paragraph is my inference, not a directly reported figure. (kyivindependent.com)
This question resolves to ACLED event rows, not to the number of drones or missiles used. That matters a lot: even a very large one-night swarm can still resolve to only a modest number of ACLED records if ACLED codes several locations/incidents rather than each munition. So I am forecasting ACLED-coded explosive/strike events in Crimea and Sevastopol between July 1, 2026 and August 1, 2026, inclusive, not raw strike volume.
Recent ACLED evidence points upward. ACLED says Ukrainian attacks in Russia and Russia-occupied regions have been rising since February, and that strikes on occupied Crimea more than quadrupled in April compared with the monthly average in 2025. In its 14-20 March 2026 update, ACLED recorded 57 Ukrainian air/drone strike events across the occupied Donetsk, Kherson, Luhansk, and Zaporizhia regions plus Crimea in a single week. ACLED’s 13-19 June 2026 monitor update also says it recorded nearly 40 drone-strike events in Moscow city and region between May 1 and June 19, versus an average of four per month in 2025, which I use as evidence that Ukraine’s long-range strike capacity and tempo are materially higher than last year. (acleddata.com)
Recent reporting from Crimea is consistent with that trend. Reuters reported a Sevastopol strike on May 27 involving more than 20 drones and Storm Shadow missiles. Reuters then reported strikes on Simferopol and Sevastopol on June 4, a train strike in Crimea and continuing fuel rationing on June 8, and a Sevastopol blackout after strikes on June 24. AP reported that public gasoline sales in Russian-held Crimea were halted on June 21 after further Ukrainian attacks, and AP described June 26 as one of Ukraine’s heaviest drone assaults of the war, including Russian-held Crimea. Taken together, I read these as evidence of a sustained campaign against Crimea’s air defenses, fuel network, transport links, and power infrastructure rather than isolated one-offs. (streetinsider.com)
On the downside, some lull is possible if Ukraine shifts assets toward deeper-Russia strikes or if diplomacy temporarily suppresses operations. But the diplomatic signal is weak. AP reported that Putin rejected Zelenskyy’s proposal for a face-to-face meeting on June 5. Reuters reported on June 18 that several EU leaders saw little point in rushing into talks with Russia. Reuters also reported on June 22 that Ukraine might revise its ceasefire offer if the UN route failed. That keeps a low-activity scenario alive, but it is not my base case. (apnews.com)
I therefore use a three-scenario mixture. Scenario 1 is a temporary lull/focus shift, with 20% weight and mean 8 ACLED events. Scenario 2 is continuation of roughly current Crimean tempo, with 55% weight and mean 19. Scenario 3 is an intensified isolation campaign aimed at Kerch, Sevastopol, and supporting logistics/power nodes, with 25% weight and mean 34. I model each scenario with a negative binomial distribution because event rows should be overdispersed: strikes come in waves, and one wave can create multiple ACLED rows across different locations. The weighted mean is 20.55, so my rounded point forecast is 21.
My resulting distribution is centered in the low 20s, with meaningful downside into single digits if the campaign pauses and a meaningful right tail into the 30s or 40s if Ukraine keeps pressing Crimea at late-June tempo. I do not discount the count heavily for Crimea-specific undercoverage because ACLED expanded its Russia-Ukraine sourcing on April 2, 2025, specifically improving coverage in eastern and southern occupied areas including Crimea from August 7, 2024 onward. Since this market resolves from ACLED data as displayed on or after September 1, 2026, there is also some time for normal backfill before resolution. (acleddata.com)
I treat this as a forecast of ACLED-coded events, not of individual drones, missiles, or explosions. That distinction matters because ACLED records a single Air/drone strike event for all same-date, same-location, same-actor air/drone strikes, and the Ukraine Conflict Monitor’s summary boxes are filtered event totals. ACLED’s accessible monitor explains the date/subset filters and notes that the downloadable Ukraine data file is restricted, so I could not directly pull a custom historical Crimea-plus-Sevastopol monthly series from the public UI. (acleddata.com)
The current operational picture as of Friday, June 27, 2026 is clearly one of escalation around Crimea. ACLED says Ukraine’s offensive reach inside Russia has been increasing in 2026 and noted nearly 40 drone-strike events in Moscow city and region between May 1 and June 19, versus an average of four per month in 2025. AP and Reuters reported late-June strikes in Crimea hitting an oil storage depot at the Kerch thermal power plant, an electrical substation in western Crimea, a gas distribution station in Simferopol, a rail bridge near Rozdolne, and Sevastopol’s main power substation; Reuters and The Washington Post also reported fuel-sale suspensions, public-life restrictions, rolling outages, and an emergency declaration in occupied Crimea after weeks of strikes. (acleddata.com)
The cadence is not just a one-night spike. Reuters reported Crimea-focused pressure on June 4, June 10, June 11, June 21, June 24, and June 26, including strikes or strike effects in Dzhankoi, Sevastopol, Armiansk, Kerch, Simferopol, and logistics routes feeding the peninsula. Reuters also quoted Ukraine’s drone commander on June 11 saying the campaign goal was to cut Crimea off from Russia, while AP reported on June 23 that Ukraine was seeking to isolate the peninsula and had already been intensifying the campaign. (marketscreener.com)
For base rates, the most useful ACLED figure I found was that in the week of March 14-20, 2026, ACLED recorded 57 Ukrainian air/drone strikes across the occupied Donetsk, Kherson, Luhansk, and Zaporizhia regions plus Crimea. Crimea is only one part of that geography, but unlike a normal week it is now a premier target set. That is reinforced by Zelensky’s approval on June 25 of a 40-day strike campaign intended to pressure Russia, which covers nearly the entire July 1-August 1 resolution window. In other words, I expect Crimea’s share of occupied-territory strike records in July to be materially above its ordinary share. (acleddata.com)
The main downward adjustments are: first, some of the broader ‘Crimea isolation’ effort is actually aimed at land-corridor routes in occupied Kherson or Zaporizhzhia and so will be out of scope here; second, improved Russian air defenses may turn some attempts into non-counting interceptions; and third, ACLED aggregation compresses large drone swarms at the same place on the same day into one event. The main upward adjustments are: multiple distinct Crimea locations can be hit on the same day and each can generate a separate ACLED record, and current reporting suggests logistics, power, air-defense, port, bridge, and vessel targets are all in play at once. (acleddata.com)
Putting that together, I center the forecast at about 29 total ACLED-recorded explosive/strike events in Crimea plus Sevastopol for July 1, 2026 through August 1, 2026 inclusive. I represent uncertainty with a three-regime negative-binomial mixture: a de-escalation regime, a sustained-current-campaign regime, and an intensified-campaign regime. That keeps meaningful probability on low-teens outcomes if the campaign is blunted or partly shifts out of scope, but also leaves a real upper tail if the late-June pace persists through most of July under the newly announced 40-day campaign. Because I could not directly extract a public historical custom-window Crimea-plus-Sevastopol series from the monitor UI, that missing time series is the biggest unresolved uncertainty in the estimate. (acleddata.com)
I treat this as a forecast of ACLED event rows, not raw missiles or drones. ACLED’s Ukraine Conflict Monitor is near real time and its count box changes with date/subset filters; importantly, ACLED says one Air/drone strike event can aggregate all air-to-ground strikes on the same date, in the same location, involving the same actors. So very large strike waves do not translate one-for-one into ACLED events, which keeps the expected count materially below headline attack totals. (acleddata.com)
For historical anchoring, I could not directly access the gated ACLED export, so I used a third-party ACLED mirror only as a rough proxy, not as the resolution source. That mirror was last updated on March 11, 2026, and its Crimea pages treat Sevastopol as a location within the Crimea region pages, so I use it only to approximate combined peninsula activity. In that proxy, Explosions/Remote violence counts for Crimea were 11 in July 2024, 17 in August 2024, 15 in May 2025, 14 in June 2025, 10 in July 2025, 13 in August 2025, 11 in January 2026, and 3 in February 2026. That gives a pre-June-2026 baseline in roughly the low teens per month, with occasional movement toward the mid-to-high teens. (ukraine-war-analytics.online)
I then adjusted that baseline upward for the current operational picture. ACLED’s June 2026 monitor says Ukraine’s deep-strike reach in 2026 has been increasing, and ACLED’s 14–20 March update already described a large Ukrainian strike week that included Crimea. In late June 2026, AP reported Ukrainian strikes on an oil depot at the Kerch thermal power plant, an electrical substation, an LNG distribution station in Simferopol, and a rail bridge near Rozdolne. Reuters reported fuel shortages on the peninsula, a Sevastopol power outage after attacks on energy facilities, and then a Russian-installed state of emergency in Crimea. The Washington Post also reported that this followed weeks of punishing strikes and that Volodymyr Zelensky had authorized a 40-day strike campaign. That combination makes July 1 through August 1, 2026 more likely to land above the historical low-teens baseline. (acleddata.com)
My quantitative forecast is a mixture of negative-binomial scenarios: a small de-escalation case, a moderate-continuation case, a sustained-campaign case, and an escalation case. The scenario means are 7, 16, 26, and 40 events with weights 0.05, 0.35, 0.45, and 0.15, which gives a weighted mean of about 23.55 events. I like this as a balance between two facts: first, ACLED aggregation compresses even heavy strike packages into fewer rows; second, the late-June evidence strongly suggests Crimea and Sevastopol are in the middle of an intensified Ukrainian logistics-and-energy interdiction campaign. My point estimate is 24 events, with most of my probability mass in roughly the mid-teens through mid-30s and meaningful but smaller tail risk into the 40s if repeated multi-location strike waves continue through July. (acleddata.com)
I am forecasting the ACLED count defined in the prompt: the sum of Air/drone strikes, Shelling attacks, and Other explosive for admin1 Crimea and Sevastopol from 2026-07-01 through 2026-08-01 inclusive. My outside view starts with ACLED’s own description of trend direction. ACLED wrote that Ukrainian attacks in Russia and Russia-occupied regions had been rising since February 2026, and that strikes on occupied Crimea specifically more than quadrupled in April 2026 relative to the monthly average in 2025. ACLED also noted a 2025 source expansion that added more than 1,110 new events focused particularly on Crimea and other front/occupied regions, with 89% of those additions being Explosions/remote violence events. That matters because this question resolves on ACLED records, not on a generic press count, so older baselines likely understate what a fully covered Crimea strike month now looks like. (acleddata.com)
The inside view for late June is clearly hot. Reuters and AP reported that Ukrainian strikes had produced fuel rationing and then a suspension of civilian gasoline sales in Crimea, with AP calling it the worst fuel crisis on the peninsula since 2014. Reuters also reported shortages in Sevastopol and Yevpatoriya and said recent strikes disrupted the road-and-rail supply routes into Crimea. These are exactly the kinds of effects I would expect when the target set is fuel, logistics, bridges, and energy nodes rather than isolated symbolic attacks. (internazionale.it)
Recent waves have also been multi-node, which is important because one operational wave can create several separate ACLED records if multiple distinct locations are struck. Reuters reported a 4 June wave hitting Simferopol while more than 20 drones were intercepted over Sevastopol. AP then reported that the 22-24 June wave included an oil storage depot at the Kerch thermal power plant, a western Crimea electrical substation, an LNG distribution station in Simferopol, and a rail bridge near Rozdolne, followed by another strike against repair equipment at the bridge. On 25 June, Zelensky said he approved a 40-day campaign to influence Russia to end the war, and on 26 June AP described one of Ukraine’s heaviest drone assaults yet as reaching Russian-held Crimea and surrounding seas. Since the resolution window begins on 1 July 2026, almost the entire question window sits inside that announced 40-day campaign. (investing.com)
The main reason not to push the mean much higher is that not every operation aimed at isolating Crimea will resolve inside the admin1 boundaries of Crimea or Sevastopol; some interdiction happens on approaches in other occupied regions, which would not count here. There is also a real adaptation effect: AP reported Ukraine’s claim that Russia is shifting more air defenses toward the Kerch Bridge and other key sites. So my central case is not daily successful multi-location strikes for 32 straight days; it is repeated clustered attacks, with some suppression by Russian defenses and some spillover of the broader campaign outside the scoring geography. (apnews.com)
I therefore used a three-regime mixture rather than a single thin-tailed count model. My regimes are: 25% de-escalation with mean 15, 55% continuation with mean 27, and 20% escalation with mean 42, each represented by a negative binomial distribution to allow clustering and overdispersion. This gives an overall mean of 27.0 events over the 32-day window and a standard deviation of about 13.6. In words, I expect the most likely outcome to be in the 20s, with meaningful probability in the teens and 30s, and a smaller but nontrivial tail into the 40s and 50s if late-June tempo persists or intensifies. My center of gravity is therefore in the high 20s: above a normal 2025 month because ACLED says Crimea activity already jumped sharply in April and because late-June evidence points to an active Crimean strike campaign continuing into July, but below an extreme scenario where successful multi-target waves occur almost every day. (acleddata.com)