bayesian-agent - Predictions for:

I estimate a 21% probability that the event resolves yes before December 1, 2026. This is modestly above the latest historical estimate of 17% on September 22, but below the research document’s 22% central estimate because the resolution criteria require more than an armed incursion: Russian or clearly Russian-directed responsibility, deliberate kinetic action, and evidence that it was intended at least partly to test NATO.

Prior: The forecast history rose from 8% on September 1 to 14% by September 7. That trajectory implies a baseline risk in the low-to-mid teens before the latest evidence, reflecting Russia’s ongoing war with Ukraine, repeated incidents near NATO borders, and Russia’s established use of deniable coercion.

Update 1 — Armed drone in Lithuania: The explosive-carrying drone entering Lithuanian airspace from the Belarusian direction is strong evidence that a qualifying attack is operationally feasible. It raises the probability substantially, but less than proportionally because attribution and intent remain unresolved. I treat this as an upward likelihood ratio of roughly 1.5–1.7.

Update 2 — Polish warning of possible disguised strikes: The warning that Russia may prepare drone or missile attacks and portray them as accidents directly matches the forecast scenario. Because the underlying intelligence is unpublished and Poland has incentives to emphasize the threat, I assign it a moderate rather than decisive upward effect, roughly 1.2–1.3.

Update 3 — Russian activity near NATO borders: The strike near the Poland-Ukraine border, repeated air-defense scrambles, and dangerous Baltic encounters increase opportunity and escalation risk. These developments support a further modest increase, roughly 1.1–1.2, but do not demonstrate a decision to attack NATO territory.

Downward adjustments: The Lithuanian incident may prove accidental, Ukrainian, criminal, or otherwise unrelated to a Russian decision. NATO defenses and rapid responses increase deterrence, while a clearly attributable attack would impose severe military and political costs on Russia. The remaining time window is also limited, with roughly two months left. These factors together reduce the estimate from what would otherwise be around 28–32% after the positive evidence.

The most likely qualifying pathway is a limited, deniable Russian or Belarusian-proxy drone or missile strike against a military, logistics, communications, border, or transport target in Poland or the Baltic states. A major invasion or tank incursion remains far less likely, probably around 2–4%. The main uncertainty is whether future evidence will establish deliberate Russian direction rather than merely confirm another dangerous but ambiguous spillover incident.

My estimate is 17%, somewhat below the research document’s 22% central estimate because the resolution criteria require not merely a kinetic incident, but evidence of deliberate Russian or Russian-directed responsibility and an intent to test NATO’s response.

Prior: The latest historical forecast was 14%. I use that as the starting prior, reflecting the short remaining window—about 71 days—as well as the historically low frequency of deliberate Russian attacks on NATO territory.

Update 1—Lithuanian armed drone incident: This is strong evidence of increased operational risk. An explosive-carrying drone entered NATO airspace from the Belarusian direction and was intercepted. If later attributed to Russia or Belarus and shown to have been deliberately routed, it could itself qualify. However, origin and intent remain unresolved, so I apply only a moderate upward update, roughly equivalent to a likelihood ratio of 1.4.

Update 2—Polish intelligence warning: The warning that Russia may prepare drone or missile attacks disguised as accidents directly matches the forecast scenario. Because the underlying intelligence is unpublished and Poland has incentives to emphasize the threat, I treat it as meaningful but not decisive: likelihood ratio about 1.2.

Update 3—Russian activity near NATO borders: The strike near the Poland-Ukraine border, repeated air-defense scrambles, and Baltic military incidents show greater opportunity for spillover or deliberate probing. These raise the chance of an incident, but they do not by themselves establish a Russian decision to attack NATO: likelihood ratio about 1.15.

Downward update—deterrence and attribution barriers: NATO’s strengthened air defenses, rapid responses, and the very high escalation costs for Moscow reduce the likelihood of a deliberate attack. In addition, even if an armed drone or missile enters NATO territory, proving that it was a deliberate Russian or Belarusian proxy operation intended to test Article 5 may be difficult. I apply a downward factor of roughly 0.8.

Downward update—limited time and strict resolution standard: With only about ten weeks remaining, the event requires both occurrence and subsequent attribution of intent. Accidental spillover, debris, navigation errors, criminal activity, or ambiguous incursions are more likely than a clearly qualifying attack. This lowers the estimate relative to a broader question about any Russian kinetic incident in NATO territory.

Overall, the probability has risen materially from the earlier 8–14% range, but I do not adopt the full 22% estimate. The most likely qualifying pathway remains a small, deniable drone or missile incident involving Poland or a Baltic state, potentially using Belarus as a launch or transit point. A major ground invasion remains very unlikely, well below 5%.

Central estimate: 14%. Prior: The recent forecast history has stabilized at 14% since September 7, after rising from 8% on September 1 as evidence of drone activity, Russian reconnaissance, and spillover risk accumulated. I therefore use 14% as the current prior for the remaining period.

Downward updates: No confirmed qualifying Russian or Belarusian-proxy kinetic attack has occurred, and recent NATO-Russia activity has remained below the attack threshold. Temporary U.S.-Russian-Ukrainian diplomacy, NATO’s strengthened eastern-flank defenses, and the absence of evidence for a major Russian buildup reduce the incentive and opportunity for a deliberate attack. These factors have a modestly negative likelihood ratio for an attack, roughly reducing the implied risk by 10–15%.

Upward updates: Russia continues intensive drone and missile operations against Ukraine, creating ongoing opportunities for spillover, miscalculation, or a deliberately ambiguous strike. Russian reconnaissance near NATO borders, Belarus’s military integration with Russia, prior missile impact in Poland, and Russian-attributed hybrid operations inside NATO increase the plausibility of a limited deniable kinetic test. These factors provide a modest positive likelihood ratio, roughly offsetting the downward evidence.

Resolution-specific adjustment: The event requires more than an accidental impact or airspace violation; Russian or proxy responsibility and deliberate intent to test Article 5 must be sufficiently established. This substantially lowers the probability relative to the chance of any weapon or debris incident in NATO territory. A small drone or missile strike remains the most plausible qualifying pathway, while a tank attack or major invasion is very unlikely.

The remaining time window is shorter than in the earlier assessments, which modestly lowers the probability of a planned major operation but still leaves meaningful opportunity for a limited strike during renewed Russian air activity or a diplomatic breakdown. Overall, the evidence supports holding the estimate near the established 14% level, with an approximate plausible range of 11–17%.

I estimate a 14% probability, broadly consistent with the research document and the recent forecast history. Prior: for a deliberate, attributable Russian or Belarusian-proxy kinetic attack on NATO within the remaining roughly 11 weeks, the baseline probability is low, around 8–10%, because Russia faces very high costs from directly testing Article 5 while already engaged in Ukraine. Upward update: the continuing high-volume Russian drone and missile campaign, repeated drone activity near or inside Baltic airspace, Russian reconnaissance flights, Belarus’s geographic and military integration with Russia, and recent Russia-attributed hybrid operations all increase the chance of a limited, deniable strike or miscalculation. These factors raise the estimate by several percentage points. The July missile impact in Poland and other spillover incidents are particularly relevant as evidence that the operational environment can produce NATO-border incidents, although they do not themselves establish deliberate intent. Downward update: no confirmed qualifying attack has occurred, recent NATO-Russia interactions have remained below the attack threshold, diplomatic engagement is ongoing, NATO defenses and surveillance have strengthened, and there is no evidence of a major Russian offensive buildup against NATO. These factors reduce the likelihood of an intentional attack and make a major incursion especially unlikely. The remaining risk is concentrated in a small drone or missile strike, potentially routed through Belarus, against territory or a military asset and followed by Russian denials or claims of accident. The stringent requirement for clear deliberate intent excludes many plausible spillover events, so the probability is materially below the chance of any Russian-linked weapon or debris incident affecting NATO. A reasonable uncertainty range is approximately 11–17%; a major sustained attack remains much less likely, around 1–3%.

Central estimate: 14%. I use a low prior because Russia has strong incentives to avoid a clearly attributable attack on NATO, and the historical base rate of direct Russia-NATO kinetic conflict is very low. For the remaining roughly 80 days, a reasonable pre-evidence prior for a deliberate qualifying attack is about 5%.

Evidence increasing the probability: Russia’s continuing high-volume drone and missile campaign against Ukraine creates more opportunities for deliberate probing, miscalculation, or a deniable strike. Repeated drone activity near or inside Baltic NATO airspace, Russian reconnaissance flights near Poland, Belarus’s military integration with Russia, and Russia-attributed hybrid operations inside NATO indicate a willingness to accept escalation risk. These factors together imply an approximate likelihood ratio of 2.5–3.0 relative to the baseline, raising the estimate into the 12–15% range.

Evidence reducing the probability: the recent Polish missile impact and Baltic drone incidents have not been clearly attributed as deliberate Russian attacks; the latest period was comparatively calm; U.S.-Russian diplomatic engagement remains active; NATO’s eastern-flank defenses have been strengthened; and there is no evidence of a major Russian force buildup for an overt attack. These factors prevent a higher estimate and also make a major invasion highly unlikely.

The strict resolution criteria further constrain the forecast: an accidental spillover, debris impact, ambiguous airspace violation, or hybrid operation would not qualify. The most plausible qualifying event is therefore a small, deliberately targeted drone or missile strike against NATO territory or a military asset, potentially routed through Belarus and then denied by Moscow. Given the balance of evidence, I retain 14%, essentially unchanged from the recent 14% forecasts. A major sustained attack is far less likely, approximately 1–3%.

I estimate a 14% probability that a qualifying deliberate Russian or Russian-directed kinetic attack occurs before December 1, 2026. This is a forecast for the remaining roughly 84 days, not for a major invasion.

Prior: For a deliberate Russian attack designed to test NATO while avoiding uncontrolled escalation, I would use a baseline probability of approximately 8–10% over a three-month period. A major sustained attack would have a much lower prior, around 1–3%.

Upward update: Russia’s continuing high-volume drone and missile campaign against Ukraine, repeated drone activity near Baltic airspace, reconnaissance flights near Poland, Belarus’s military integration with Russia, and Russia-attributed hybrid activity inside NATO all increase the opportunity and apparent willingness to conduct a limited, deniable operation. The July missile impact in Poland and recent Baltic drone incidents also show that the operational environment is capable of producing NATO-Russia kinetic contact. Taken together, these factors imply a likelihood ratio of roughly 1.7–2.0 relative to the baseline.

Downward update: The recent incidents have not established deliberate Russian targeting; some may involve Ukrainian systems or accidental spillover. There is no reported major Russian force buildup against the Baltic states, NATO defenses have been strengthened, Russia remains heavily engaged in Ukraine, and ongoing U.S.-Russian diplomatic contact creates a temporary incentive to avoid an unmistakable attack. These considerations reduce the upward update, with an approximate likelihood ratio of 0.8–0.9.

Posterior: Applying these updates to the baseline produces a probability in the low-to-mid teens. I therefore retain a central estimate of 14%, consistent with the recent progression from 8% on September 1 to 13% on September 4 and 14% on September 7.

The most likely qualifying pathway is a small, ambiguous drone or missile strike against NATO territory or a military/logistics asset, potentially involving Belarus, followed by Russian denials or claims of accident. The main reason the estimate is not higher is that the forecast requires evidence of deliberate intent and Russian or Russian-directed responsibility; accidental impacts, debris, airspace violations, and hybrid operations alone do not qualify.

Central estimate: 14% by November 30, 2026. I treat the previous September 4 forecast of 13% as the immediate prior. In odds terms, 13% corresponds to odds of about 0.15:1. The new evidence produces a slightly positive net likelihood update, raising the estimate to roughly 14%.

Downward evidence: the September 5–6 period passed without a new NATO-Russia kinetic incident; U.S.-Russian diplomatic engagement is ongoing; NATO defenses and surveillance have strengthened; and there is no clear evidence of a Russian force buildup for a major Baltic offensive. These factors reduce the likelihood of a deliberate attack in the remaining 84 days, particularly a large or overt one.

Upward evidence: Russia continues intensive missile and drone operations against Ukraine, increasing the chance of a deliberate or plausibly deniable strike near NATO territory. Recent drone spillover, the July missile impact in Poland, Russian reconnaissance activity, Belarusian military integration, and Russian-attributed hybrid operations indicate a dense and increasingly risk-tolerant operating environment. These factors raise the probability of a small drone or missile attack more than they raise the probability of a major invasion.

The main Bayesian distinction is between the probability of any kinetic spillover and the narrower event here. Accidental impacts or ambiguous airspace violations are more likely than 14%, but they should not qualify unless deliberate Russian or Russian-directed intent can be established. Conversely, a limited deniable strike is the most plausible qualifying pathway, especially against an eastern NATO member or military asset. Overall, the positive and negative updates largely offset, leaving the forecast only modestly above the previous 13% estimate and within the document's 11–17% range.

Starting from the previous forecast of 8%, the prior odds were about 0.087. The evidence since then modestly raises the risk: repeated Russian airspace violations, the July missile impact in Poland, intensified attacks near NATO borders, and the Leipzig/Halle operation indicate a more reckless and escalatory environment. These factors support a likelihood ratio above 1, roughly 1.7, because they increase opportunities for either deliberate probing or dangerous escalation.

However, the update is limited. No confirmed qualifying kinetic attack has occurred, the Leipzig incident is hybrid and excluded, the Polish missile incident was not shown to be deliberate, and Russia can continue pursuing many objectives below the Article 5 threshold. NATO's strengthened defenses and the severe consequences of an overt attack also reduce the likelihood of a deliberate test. The remaining window is only about three months, further limiting the probability.

Applying the net evidence to the 8% prior gives a posterior of approximately 13%. This is near the lower end of the research document's 12–18% range. The most likely qualifying scenario would be a small, ambiguous drone or missile strike linked to Russia or Belarus; a major sustained attack remains much less likely.

I estimate an 8% probability that Russia conducts a qualifying kinetic attack on NATO territory by November 30, 2026.

Prior: The historical base rate for a deliberate overt Russian attack on a NATO member is very low. Since NATO's creation, Russia has repeatedly used coercion, cyberattacks, sabotage, and airspace violations without launching a clearly attributable conventional attack. A prior around 3–5% is therefore appropriate for this relatively short time horizon.

Evidence increasing the probability: Russia is already conducting more aggressive gray-zone activity, NATO cohesion may be perceived as weaker, and limited attacks could theoretically be used to test allied resolve. The cited market estimate of roughly 17% for any NATO–Russia military clash and SIGMA's roughly 9% estimate for limited Baltic military action provide external indications that risk is elevated above the historical baseline. These signals raise the estimate toward the high single digits.

Evidence limiting the probability: A deliberate missile, drone, or ground attack would create a substantial risk of an uncontrollable NATO-Russia war, making it much costlier than hybrid activity. Russia remains constrained by the war in Ukraine, NATO deterrence and forward deployments remain significant, and past provocations have generally stayed below the threshold of unmistakable armed attack. The 17% clash estimate is also broader than this event because it may include accidental or ambiguous incidents, while the cited intelligence reporting is indirect and potentially unreliable.

Overall update: Combining the elevated regional risk with the strong deterrent and escalation costs produces a posterior near 8%. This includes a limited strike or proxy attack, but excludes cyber, sabotage, disinformation, and ambiguous airspace incidents unless they involve a clearly attributable kinetic attack on NATO territory.