Forecast verification: archived publications and events
Checked on: 21 October 2026
This selection examines dated statements from Volodymyr Kraine’s Facebook and Oculus publications from 2014-2025. Each statement opens in the register with its original quotation, forecast window, source and assessment basis. Oculus links lead to the source publications; archive copies are available for Facebook posts.
How the check works
The stated topic, expected direction and time window, including the author’s tolerance, are compared with documented events. A forecast does not have to describe the entire subsequent chain of events. A thematic match with an inaccurate detail receives a partial grade; an opposite outcome does not count as a hit.
The calculation unit is one statement with an expected outcome and a window. Alternatives within one scenario remain one item. Repeating a forecast does not increase the number of hits. Independent statements in the same text, such as a victory and a subsequent retreat, are assessed separately.
When alternatives are joined by “or”, one matching alternative within the stated window and context is sufficient. For example, “information, movement or infrastructure” does not require three separate events. The absence of the other alternatives does not reduce the grade. A sporting victory for another country, however, does not support a forecast of success for Ukraine: the context cannot be changed after the event.
Grades A, A-, B+ and B represent documented signals of different strengths. C+ and C are weak or background matches. D is an outcome that contradicts the forecast. N means that the reviewed material provides insufficient data for a grade. These forecasts remain visible in the full register but are excluded from the percentage: they count as neither hits nor misses. The full selection and the number of graded forecasts are shown separately.
Emotions, comments, hopes, advice, discussion and descriptions of past events are not forecasts and are not included in the register. An astronomical position alone is not a predicted event. Unfinished periods and repeated formulations of the same forecast are also excluded. A lack of supporting evidence is not treated as proof that an event did not occur.
This is a retrospective check of a public selection. An independent assessment of the entire archive requires selection rules fixed in advance and a complete list of forecasts for the chosen period.
91.4% matches among assessed forecasts
Registered forecasts
45
Received a grade
35
Supported signals
32
Insufficient data for a grade
10
The percentage concerns forecasts graded A-D using the cited sources. Grades A, A−, B+ and B enter the numerator. N means insufficient data: these forecasts remain visible in the full register but count as neither hits nor misses and are excluded from the percentage. Emotions, comments, hopes, advice and descriptions of past events are not included as forecasts.
This is a retrospective assessment of the stated selection, not an independent measure of all forecasts’ accuracy or proof of the method’s effectiveness. Ungraded outcomes may differ from graded ones; their exclusion limits the conclusion. Some forecasts concern the same historical episodes.
Grade
Registered forecasts
A · Exact match
1
A- · Strong match
3
B+ · Good match
7
B · Thematic match
21
C+ · Weak match
1
C · Background match
2
N · Unconfirmed
10
Random events and random time windows
The control uses events from this register, not unrelated news. Each trial randomly permutes events between forecasts from the same source publication and randomly selects each window start. Actual event dates stay unchanged. Window lengths are preserved, but original window dates are not used. A match means the assigned event date falls inside the random window.
Two modes test sensitivity to the sampling period: starts in the same calendar month or the same year as the original window start. The preserved length determines the end, which may cross a month or year boundary but cannot exceed the review date. Permuting within a publication preserves historical context and event clustering. Original pairs or dates can occur by chance and are not removed.
Original pairs with an event date inside the window: 34/35 (97.1%)
This measures timing only, not substantive forecast accuracy. The main percentage above also considers topic and direction; these metrics are not interchangeable. Dates are those cited in the register, a simplification for multi-day events. An event can support multiple comparisons, so comparison and unique-event counts differ.
Random start
Within the same month
Within the same year
Mean timing match
19.3%
2.1%
Central 95% of trials
8.6% - 31.4%
0.0% - 8.6%
Trials reaching 50%
0.0%
0.0%
Allowing any event from the same publication
42.6%
4.7%
The last row removes the single-pair restriction: the date of any event in the same pool is sufficient. It shows how unrestricted matching raises the result. Topics are deliberately not checked; this is a more permissive timing control.
A 50% guessing probability is not assumed: it depends on event counts, window lengths and the sampling period. This retrospective control measures random timing matches only in the selected list. It lacks a complete calendar of events and forecasts, and original pairs were already selected during assessment. It therefore neither proves astrology effective nor refutes a universal 50% claim. Such a conclusion requires independent testing of preregistered forecasts against a complete event set under the same matching rules.
First reproducible trial
The first trial is shown, not a cherry-picked successful example.
012014-07-03 - 2014-07-07Artillery strike near Zelenopillya2014-07-11 · Date did not match
Original forecast 01
Assigned event: Artillery strike near Zelenopillya · 2014-07-11
Assessment basis: One explicitly listed alternative is supported within the stated window. The word “or” does not require every listed type of event to occur together.
The Moscow metro accident on 15 July falls within 11-15 July and matches the stated technological-disaster alternative. The forecast did not specify a place or type of transport.
The OSCE called for urgent evacuation from Debaltseve on 3 February and monitored hundreds leaving on 6 February. Civilian danger in the 2-6 February window is supported.
A PACE report records Debaltseve as encircled by 11 February. This fits the 10-14 February window and the encirclement theme; the forecast did not name the town.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
The 12 March report discusses an exchange of existing detainees, not a new encirclement or capture within 10-14 March. An exchange is not substituted for the stated event.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
The OSCE documented offensive action near Marinka on 3 June. This does not establish the predicted overall reversal between 15 March and 16 June; one episode also cannot rule out every local change.
The OSCE continued to document offensive action within the stated period, including the 3 June attack near Marinka. This does not establish a general reversal of the front from 15 March.
On 27 February Ukraine signed its letter of intent and memorandum for IMF financing. This supports continuing financial-assistance activity within 20 February-5 March, not a claim that funds arrived that day.
The 16 March attack on Mariupol theater fell in the escalation window. The tragedy and civilian target were not specified in advance, so the match is thematic.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
The reviewed UN report documents violence in March-April, but not the stated link between strikes in ORDLO and negotiating leverage within 31 March-6 April. Motive is not inferred automatically from an attack.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
The reviewed UN briefing describes continuing diplomatic and humanitarian activity, but does not sufficiently establish the stated decline in talks and movements within 3-8 April. One meeting alone neither proves nor disproves that trend.
8908-13.04.2022Я бы сказал о секретном раунде переговоров, т.е. тех, которые происходят не для общественности, а где именно стороны о чем-то договариваются. Также стоит упомянуть, что Меркурий в 9-ом говорит об активности иностранных, как сейчас принято говорить, партнерах. И/или иностранных перемещениях, тут все что угодно может быть, как самих иностранцев, так и иностранных объектов, оружия например.B
Assessment basis: One explicitly listed alternative is supported within the stated window. The word “or” does not require every listed type of event to occur together.
On 11 April Austrian Chancellor Karl Nehammer visited Moscow for talks with Putin after visiting Kyiv. This matches the full passage’s explicit alternative of foreign partners’ activity and travel. It does not establish a secret round between Ukraine and Russia, which is not credited.
On 13 April OSCE experts presented their report on violations of international law and war crimes in Ukraine, a documented development in public accusations within 8-23 April. The match is thematic; the size of the stated “huge flow” was not independently measured.
The ICRC described a five-day evacuation from Azovstal amid continuing hostilities in late April and early May. The war did not end, but this is a background match; the predicted respite is not independently confirmed.
The withdrawal of Ukrainian troops from Sievierodonetsk was announced on 24 June, a significant frontline change in the 21-25 June window. The forecast allowed both advance and retreat, so the match is thematic without a specific direction.
On 28 June ISW described constraints on Russian redeployment after Severodonetsk: destroyed bridges, river crossings and the need to restore combat power. This supports impaired troop movement within 26-30 June, not a complete halt to all fighting.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
Kherson’s liberation shows that Ukrainian offensive successes continued, but it does not establish a temporary retreat caused by the cost of the Kharkiv operation. The causal scenario remains unsupported.
The liberation of Kherson on 11 November shows that Ukrainian offensive successes continued. This single report cannot establish whether temporary retreats occurred in individual sectors; the predicted causal scenario is not confirmed.
The fuel-excise law was adopted on 21 September, within the window. Its bill had already been registered on 29 August, before the forecast. Unpopularity and the predicted severe consequences are not established by this document.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
Hostilities did not stop: UNHCR documents the major 15 November attack. This review lacks a sufficient measure of overall intensity across the full window; a single attack is not treated as a whole-month assessment.
The 15 November explosion in Poland prompted urgent NATO consultations. An international crisis in the window is supported; a change in every person’s perception cannot be verified.
The 15 November explosion in Poland prompted urgent NATO consultations. An international crisis in the window is supported; a change in every person’s perception cannot be verified.
The liberation of Kherson on 11 November falls within the author’s broader 13-22 November window with a two-day tolerance, before the central dates of 14-16 November.
On 21 December the US announced aid including its first Patriot transfer. This is positive military news in the window, rather than a battlefield victory. Repeated descriptions of this event are not counted again.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
The December chronology does not reliably identify one Russian-announced step implemented by 20 December that fits the favorable-for-Ukraine scenario. Patriot delivery is not substituted: it was a US decision, not Russian action.
The 22-23 December storm disrupted US rail and air travel before the holidays. The short window and transport theme match; location and weather cause were not specified in advance.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
The chronology does not establish a single 27-31 December event with documented losses for both sides. Two unrelated news items cannot be combined into confirmation of one event.
On 4-5 January France, the US and Germany announced new armored-vehicle support. The positive political theme for Ukraine and the 2-6 January window match; the specific aid was not named in advance.
On 8 January the Russian claim of hundreds of Ukrainian military deaths in Kramatorsk was refuted. Timing and false information match; a positive public effect for Ukraine is not independently measured.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
The UN describes continued and escalating attacks in June, but the forecast refers to the end of a specific intensified wave, not all wartime shelling. Without a comparable daily intensity series, that narrower outcome is not established.
Assessment basis: One explicitly listed alternative is supported within the stated window. The word “or” does not require every listed type of event to occur together.
The 6 June destruction of Kakhovka dam destroyed the road and railway crossing, a transport problem within 2-6 June. The forecast was published on 4 June, before the event. This is a different statement from the 1 June forecast ending on 4 June; that earlier statement is not credited.
The 18 June assessment describes a pause in some advances and a review of tactics. This fits the general warning of a planning review in the 14-20 June window, without establishing every forecast detail.
The review of the cited material did not provide sufficient data for a definitive grade. This is not an established miss. The forecast remains in the full register but is excluded from the percentage.
Counteroffensive action was already reported on 10 June. However, the forecast is a conditional branch, and its further wording does not unambiguously distinguish the operation’s first start from a new phase. Without resolving the condition and that meaning, 10 June alone does not justify a miss grade.
On 4 July Vnukovo restricted takeoffs and landings and diverted flights. Transport disruption matched the 3-7 July window. The 17 July Crimean Bridge strike is not used for this window.
The UN condemned attacks on Odesa and other Black Sea port facilities on 20 July. The 19-23 July window and disruption of transport match; specific ports were not named in advance.
Assessment basis: One explicitly listed alternative is supported within the stated window. The word “or” does not require every listed type of event to occur together.
The 12 June strike on Kryvyi Rih damaged residential infrastructure and caused civilian casualties. It falls within 9-16 June, and infrastructure is explicitly one of the alternatives. That alternative matches; a separate transport event is not required.
On 21 June the Council of the EU approved Ukraine’s negotiating framework. Date, negotiations and documents match. The earlier Peace Summit and G7 decisions are not counted for this window.
Assessment basis: One explicitly listed alternative is supported within the stated window. The word “or” does not require every listed type of event to occur together.
On 24 June Borrell described the use of proceeds from Russian assets for military support to Ukraine. The financial theme fits the window, but a causal link to the specific 21 June event is not established.
Assessment basis: One explicitly listed alternative is supported within the stated window. The word “or” does not require every listed type of event to occur together.
The 13 July double strike on critical infrastructure in Budy killed a rescuer and a police officer. Infrastructure danger is supported within 12-16 July; every listed sphere need not match together.
On 17 July, 95 Ukrainians returned from captivity, a positive military-related event within 12-22 July. A truce did not occur; the match concerns the broader alternative of a small military success.
Assessment basis: One explicitly listed alternative is supported within the stated window. The word “or” does not require every listed type of event to occur together.
A missile damaged FSD’s humanitarian office, vehicles and equipment in Kharkiv on 24 July. Staff were not killed or injured, but the organization documented danger to civilians and humanitarian work within the window. The forecast stated danger, not mandatory casualties.
The UN’s 13 January report says the start of 2026 brought Ukraine neither peace nor respite. This supports the no-peace-agreement scenario by mid-January; assessments of the whole of 2026 are not yet complete.
The UN’s 13 January report says the start of 2026 brought Ukraine neither peace nor respite. This supports the no-peace-agreement scenario by mid-January; assessments of the whole of 2026 are not yet complete.
