Who killed people on Kyiv’s Maidan on February 20, 2014: a dossier on each of the 48 deaths — Claude’s analysis

Most of the 48 dead, in Claude’s estimate, were killed by fire from the security forces, mainly the Berkut special company. Fire from the Maidan side probably also took protesters’ lives: about 8 deaths by the model’s median, about 9 on average, within a range of roughly 5 to 16 depending on how reliable the court’s evidentiary base is. Taking each person separately, Claude assigns 40 deaths to the security forces (34 to the special company) and 4 to the Maidan side, while in four cases no side reaches even 56%. Who exactly fired from the Maidan side, and whether the order to fire at people came from above the regiment and company level, has not been established.

The security forces account for about 39 deaths on average: about 28 for the Berkut special company and about 11 for other security forces. Person by person, Claude assigns 40 people to the security forces, 36 of them “almost certainly.” The evidence: bullets from special company rifles (in 12 of the 48 deaths), videos in which the special company fires live rounds at people in the same places and at the same time, and the time-aligned recordings of Vysota (“The Height”) and the National Bank (NBU) cameras (Parts 4–6).

Maidan-side positions account for about 9 deaths on average; in 90% of the model’s runs, between 5 and 16. About 5.4 of these are backed by a direct indicator; the rest consists of small shares across many episodes, including a residual for possible errors in expert examinations. Person by person: Volodymyr Zherebnyi—“almost certainly” (85%, right at the threshold), Oleh Ushnevych and Ihor Kostenko—“probably,” Vitalii Kotsiuba—“more likely”; four more are “too close to call.”

The court explicitly pointed to the Hotel Ukraina sector in two fatal episodes, Volodymyr Zherebnyi and Oleh Ushnevych (for Ushnevych, the hotel or the area in front of it), but these findings rest on different grounds. For Zherebnyi, his posture at the moment he was hit was established from two videos, and with that posture an entry wound on the left rules out all security-force positions: they were to his right and front-right. There is no bullet and no caliber, and the size of the entry wound leaves a small probability that entry and exit were mixed up—in which case the shooter would have been at the barricade. This is the most solid of the episodes in which the shot came from the Maidan side; Claude puts it at 85%, right at the threshold of “almost certainly.”

For Ushnevych, the finding rests on a posture described by a single eyewitness, and the case file contains two different bullets. The court accepted a bullet with 4 grooves (“probably a Makarov”), which does not match the Omega officer’s Fort-14TP; the other, six-groove bullet, which the court rejected, is categorically linked to the Fort by expert examinations of 2019–2020 (ordered by the Prosecutor General’s Office (PGO) and used by the SBI as the basis for the notice of suspicion). A bullet from the same Fort (according to the same examinations) hit the shield of a protester behind the same parapet, and one of the men next to him at the parapet was wounded from the security-force area in the same seconds. But both fatal shots came from a pistol, the second eyewitness points toward the October Palace (Zhovtnevyi Palace), “not from the top of Instytutska,” and another man next to him was wounded in the back from the direction of the hotel in the same minute. Here Claude gives the Maidan side 72% and the security forces 28% combined—noticeably less solid than for Zherebnyi.

In one more episode (Ihor Pekhenko) the court allows for the hotel. In eight episodes the court has no direct data on the shooter (in two of them, Voitovych and Opanasiuk, it nevertheless attributed the death to the security forces, relying on neighboring episodes).

The court also acknowledged that in the morning there were people in the hotel with weapons that were fired, while stipulating that this does not refute fire by the special company and other security forces at activists. The court also found fire from the Maidan side at the security forces proven: from about 05:30, three security officers were killed and 39 wounded, and at about 09:16 another special company officer was killed near the October Palace. The court describes the Maidan-side shooters only as groups and roles; it did not identify a single one by name (Part 7).

The shooter’s intent has not been established in any episode. In Claude’s estimate, of the roughly nine deaths from the Maidan side, about 3.5 were accidental hits (a miss, a ricochet, crossfire), about 1.9 were mistaken identity, and about 3.6 were a shot aimed deliberately at a protester. In no episode does the last grade reach 56% (for Zherebnyi it is 46%). The probability that there was at least one such killing ranges from ≈46% to ≈91%; the uncertainty model gives ≈97%, but for this question it comes close to a count that treats episodes as independent and therefore overstates it (Part 7).

The most likely picture is a small independent group or individuals (≈60% of the deliberate share). A planned provocation with the knowledge of part of the Maidan leadership or field commanders: ≈20%; a provocation from within the security forces: ≈10%; Russian involvement: ≈3%. None of the versions rests on an identified shooter (Part 7).

The court found it proven that on February 18 the Minister of Internal Affairs, the head of the SBU (Security Service of Ukraine) and other officials decided to use the powers of the security forces, including the use of weapons, and that on the same day the special company received weapons on the instructions of the regiment commander. As for who ordered shooting on February 20, the court established this at the regiment and company level: at about 09:05 the deputy regiment commander ordered indiscriminate fire to get the security forces out of the palace, and from about 09:17 he, together with the special company commander, ordered them to shoot to kill in order to stop people. The court did not establish that the order to open fire came from above. The in absentia verdicts against the regiment and special company commanders were handed down without the parties contesting the case and carry less evidentiary weight (Part 8).

He is right that there was fire from the Maidan side and that the court itself pointed to the hotel; in six episodes the Maidan side, in Claude’s estimate, is no less likely than the security forces. His criticisms of the investigation are also valid: there is almost no directional ballistics, some physical evidence was lost or excluded, and the rifles were sawn apart. But his thesis is much broader: that most or nearly all of the dead were killed by organized Maidan shooters in a false-flag operation carried out by part of the Maidan leadership. In Claude’s estimate, such a thesis has a probability of less than about 3%: even if the court’s entire evidentiary base is wrong, the Maidan side accounts for about 21 of the 48 deaths (Parts 7 and 9).

His main argument—that the synchronized videos show no Berkut shot at the moment of a specific death (for 39 of 49; for 10 there is no video) Katchanovski, ch. 3, p. 64—is an argument from absence. Synchronization does exist, however: the court relied on the recordings of Vysota and the National Bank cameras aligned to astronomical time and used them, for example, to establish that in the second when Melnychuk was shot (16:50:16 by Vysota) a special company officer behind the concrete barricade fires an AKMS—this is visible on the National Bank camera 2023 verdict, para. 5.4.125.

Katchanovski himself acknowledges this de facto: in the same paragraph he writes that in several cases the time of death coincides with Berkut shots, explains this by the different sound of the shots and by the officers not aiming at these people, and allows for accidental ricochet hits Katchanovski, ch. 3, p. 64; the death of Solchanyk at 09:08, by his own account, coincides with a Berkut officer’s shot into the ground a few meters in front of him Katchanovski, ch. 3, p. 56.

The argument from absence therefore outweighs neither what the court saw in the recordings nor the direct ballistics (in 12 deaths, a bullet from a special company rifle). The first protester killed, Heorhii Arutiunian (8:59:34), was linked by the court, on the basis of video, to Berkut officers in blue camouflage who fired hunting buckshot. Khrapachenko (11:28) was linked by the court to security-force positions, not to the hotel; so was Melnychuk (16:50), although the court at the same time acknowledged that an aimed shot was possible from the hotel but not from the barricade.

The verdict has not entered into force: the appeal has not concluded, and there was no decision as of October 6, 2026. Most deaths were attributed to the defendants through group findings, without a bullet tied to a specific body, and the chain of custody of a number of bullets is questionable; therefore even for episodes with direct ballistics Claude puts the special company no higher than 92%. The spread of the results is driven mainly by two common risks: that the entire ballistic base is wrong at once, and that all the court’s group findings are wrong at once. You can enter your own values for these risks in the calculator in Section 6.1, “Probabilistic result.”

The jury court answered the question of the guilt of five specific defendants, not the general question of who fired. That is why the verdict contains three levels of wording, and they must not be conflated:

• “Доведено поза розумним сумнівом” (Proven beyond reasonable doubt)—the involvement of a specific special company officer (a bullet from his assigned rifle and/or a visible shot at the moment of wounding) 2023 verdict, para. 5.4.99. This is the strongest level: it was tested by both parties to the trial.
• “Безспірно встановлено” (Established beyond dispute)—a shot from territory under the exclusive control of security forces, without a specific shooter 2023 verdict, para. 5.4.105. These findings do not concern the defendants’ guilt (their involvement in these episodes has not been proven), and the parties contested them less; in Claude’s estimate, in some episodes the word “безспірно” (beyond dispute) is stronger than its justification.
• Not established / “не виключається” (cannot be ruled out)—the source has not been established; the involvement of security forces has not been proven, and that of other persons has not been ruled out 2023 verdict, para. 5.7.

The hotel as the source of a shot is a separate category. For Zherebnyi and Ushnevych it is a finding of the court 2023 verdict, paras. 5.4.74, 5.4.83, and likewise for a protester wounded nearby 2023 verdict, para. 5.4.75; for Pekhenko it is only “не виключається” (cannot be ruled out) 2023 verdict, para. 5.4.27. Where the court established something, where it merely did not rule it out, and where it acquitted on grounds of doubt is always stated separately in the dossier; episode cards are marked with the level of the court’s wording.

For each death, 100% is distributed among three versions: the Berkut special company, other security forces, and shooters on the Maidan side. First, a baseline weight is set according to what the court established:

Table 2.1

The baseline already includes the risk that the court’s finding is wrong. For ballistics, this is the probability of a false match of a bullet to a rifle—from 4.5 to 18% per episode; for findings based on video, group and territory, about 8% for the whole evidentiary base. If a finding is wrong, the episode is assessed as if that finding did not exist. Hence the residual of 5–7% for the Maidan side even with direct ballistics, and 13–16% with findings based on video and territory.

Then the weight shifts according to the details of the specific episode. Each shift is recorded in the episode card (“Why these percentages”): how many percentage points, from where to where and on what fact, so the arithmetic can be checked. The strength of a shift is an odds ratio: with a different baseline, the same fact changes the share in the same proportion, not by the same number of points. The main kinds of shifts:

• Time. According to the court, the special company came out and opened fire with rifles at about 9:05 (from 9:03:05, shots with non-lethal means from Fort-500s). Deaths before then almost rule out the special company.
• Caliber and ammunition. Bullets not of 7.62×39 caliber (hunting .308; a fragment that according to the experts “approximately” corresponds to .30-06—the court did not consider the caliber established; sniper 7.62×54R) move weight away from the special company and distribute it between other security forces and the Maidan side.
• Ricochet. If the court allows for a ricochet, the direction of the wound says little about the shooter, and the uncertainty remains greater.
• Witnesses and counterarguments. Testimony of those standing nearby, Katchanovski’s arguments (steep wound angles, moments when Berkut is not firing on video), investigative experiments.

The expected number of deaths on each side is the sum of the corresponding percentages across all 48 episodes. An episode at 90% for the Maidan side adds 0.9 deaths; an episode at 12% adds 0.12. The simple sum gives 27.9 / 10.9 / 9.1; the model with common factors (step 4) gives ≈28.0 / 11.0 / 9.0.

The case-by-case result (Overview; Section 6.2) answers a different question: which version is the most plausible for each person. Episodes where the Maidan side is more probable contribute less to the sum than their number, and the remainder accumulates from small shares across many episodes:

Table 2.2

That is why the two results do not contradict each other: the four people the case-by-case count assigns to the Maidan side and the roughly nine expected deaths from that side are honest answers to different questions.

Doubts about different episodes are not independent. If the court’s evidentiary base itself is wrong—the reference test-firings of the rifles, the interpretation of videos or of forensic medical examinations—this affects many episodes at once. That is why the spread is calculated with common factors:

• Per-episode error risk.

  For the 12 episodes with ballistics, it is the probability that the match of the bullet to a special company rifle is false: from 4.5 to 18% per episode; of this, about 4% is common to all, and the rest is specific to the episode. For the 22 episodes with group, video-based or “territorial” court findings, the risk is about 8%, and it is entirely common. If a finding is wrong, the episode is assessed as if the court had not identified the shooter, but with the same episode-specific arguments. The remaining 14 episodes have no such branch: their uncertainty is already contained in the percentages themselves.

• Common factors.

  An error in all reference samples at once (about 4%) makes all the ballistic and all the group findings wrong at once; an error in the interpretation of videos and expert examinations (about 4%) makes all 22 group findings wrong at once. Together they produce the roughly 8% for group episodes.

• Maidan-side fire after 09:20.

  Up to 09:31 the court records an armed group on the hotel side, shots and an exchange of fire; after 09:31, a general formula about “covert fire,” a single shot after which activists shout toward the hotel (up to 09:43), and armed people in the hotel at 10:12–10:31; the court does not establish any gunshot wounds to security officers after 09:20. Fire is considered “significant” with a probability of 85% before 09:31 and 50% after; if it was “sporadic,” then in episodes in the 09:20–10:31 window where the Maidan share is background or rests only on a weak indicator, accidental hits (M1) are reduced to 30% of their value, and the removed share passes to the security forces in proportion to their shares.

• Linked episodes.

  Episodes in one group (at Arkada, the late ones at 10:21–10:31, the footbridge and the palace terrace, and others) are simulated jointly, using a random variable common to the group (correlation 0.5). Match errors for bullets from the same rifle are partly common: each such bullet, with probability ½, takes the draw common to that rifle.

Then the computer “replays” the events 40,000 times: first the common factors, then the shooter according to each episode’s probabilities. These simulation runs yield the mean, the median and the 90% interval. Two bounding scenarios for reference: if all the court’s findings and expert examinations are correct, about 8 deaths fall to the Maidan side; if the entire evidentiary base is wrong at once, about 21. Even in the second case not all deaths fall to the Maidan side: even if the expert examinations are wrong, the videos show the special company firing live rounds at people in the same places and at the same time; the episode is then assessed as “shooter not established” with its own arguments.

According to official data cited in the Sviatoshynskyi court’s verdict, from January 19 to 26, 316 protesters and 191 members of the security forces were injured. Four protesters died, three of them from gunshot wounds 2023 verdict, para. 1. The court describes the appearance of firearms in those days as “точковою” (isolated), but notes that it strongly affected the course of the protest 2023 verdict, para. 5.6. The court lists the dead without naming them 2023 verdict, para. 5.6. Below are the four people who died in those days, according to media reports.

Table 3.1

Katchanovski believes the January killings were carried out from the Maidan side as a false-flag operation Katchanovski, ch. 7, pp. 204–205. According to the latest reports, the investigation favors the opposite version and suspects the security forces. No court has examined either version, and none of the shooters has been identified.

The court cites official data. On February 18, 771 protesters and 398 members of the security forces were injured; 19 protesters died, 12 of them from gunshot wounds, as did 8 members of the security forces, all from gunshot wounds. Gunshot wounds were sustained by 120 protesters and 133 members of the security forces. During the night and morning of February 19, another 156 protesters and 105 members of the security forces were injured, and another 6 protesters died, 3 of them from gunshot wounds, as did 1 member of the security forces 2023 verdict, paras. 1, 5.6.

Violence, in the court’s wording, was used “з обох боків, в тому числі з прихованим використанням вогнепальної зброї” (by both sides, including with the covert use of firearms) 2023 verdict, para. 1. In the court’s assessment, on February 18 more gunshot wounds were sustained by the security forces, but fewer of them died thanks to their equipment. From February 19, protesters had more gunshot wounds 2023 verdict, para. 5.6. Data on protesters were collected only from requests for medical help, so they are probably incomplete 2023 verdict, para. 5.6.

Table 3.2. Who is suspected of the killings of February 18–19.

The summary data examined by the court give the time and place of death for all eight who died on February 18. Two Internal Troops servicemen died at 16:40 near the Trade Union Building, and two more between 16:00 and 17:00 near the October Palace. An officer of the Crimean Berkut died at 21:00 near the flower clock, and an officer of the Kyiv regiment (PMOP) at 21:00–22:00 near the Globus shopping mall.

Another Internal Troops serviceman died at 22:30 near the palace, and another at 23:00 on the steps by the Stela 2023 verdict, para. 5.6. Contrary to a witness’s account, the court placed the deaths of the Internal Troops servicemen not in the clashes in the government quarter but in the period of the fighting for the Maidan 2023 verdict, para. 5.5.2. In none of these cases has it been established who fired. In the case concerning the wounding and deaths of security-force members over the entire period of the protests (13 dead, all on February 18–20), as of December 2018 two people had been served notices of suspicion 2023 verdict, para. 5.6. The defense argued at trial that the investigation’s attention was clearly skewed; the court found this argument “не безпідставними” (not unfounded) 2023 verdict, para. 5.6.

The individual leads are as follows. In 2015 investigators found the author of an online article who, under a pseudonym, confessed to killing Berkut officers on February 18. He took part in the protests and was wounded in the thigh on February 18, according to him by a Makarov pistol. The investigator sought access to his medical records to check these statements (Pecherskyi court ruling 757/26409/15-к). How the check ended is unknown. The Omega commander testified in court that on February 18 he saw a man in a balaclava wound an Omega officer with a pistol at the intersection of Shovkovychna and Instytutska Streets 2023 verdict, para. 5.5.1.

Katchanovski writes about several more public confessions to killing security-force members on February 18. He also cites rulings of Kyiv courts: the weapons used to kill two Internal Troops servicemen on February 18 allegedly matched the weapons of participants in an attack on a checkpoint near Sloviansk in April 2014 Katchanovski, ch. 4, p. 124; ch. 7, p. 200. These rulings have not been found, and the data have not been verified.

The chronology of February 18 was reconstructed by the Jus Talionis lab. According to its data, by 16:00, 11 people had been killed or fatally wounded (15 including the January deaths), and not a single member of the security forces. The first two security-force members were fatally wounded around 16:20–16:25 on Khreshchatyk. At the same time, the authors of the reconstruction note that there is no confirmation that it was the protesters who opened fire on the security forces. According to the summary cited by the court, two Internal Troops servicemen were fatally wounded at 16:40 near the Trade Union Building, that is, also on Khreshchatyk, and two more between 16:00 and 17:00 near the October Palace. The times differ by 15–20 minutes; the locations are close. That more than ten protesters had already died by that point is known only from the Jus Talionis reconstruction; the court does not give the times of the protesters’ deaths.

Black uniforms; from February 19, yellow armbands made of tape, “щоб відрізнятися від інших правоохоронців у схожій чорній формі” (in order to be distinguished from other law enforcement officers in similar black uniforms) (defendant’s testimony, 2023 verdict, para. 5.1); weapons: AKMS 7.62×39, Fort-500, pistols. Its movements as described by the court:

Table 3.4. Other security forces: summary table. Times are based on the court’s descriptions and on testimony; the Vysota video starts at 08:53–08:55, and everything earlier is based on other recordings.

Of the security forces, this is the only one for which the verdict describes shooting in the area of the deaths before 09:03:05: at 08:55–08:59 it covers the Internal Troops’ retreat from the Stela with Fort-500 shots 2023 verdict, para. 5.3.2; at 08:57–09:00:20, 17 of its officers fire from behind the parapet, two of them with pistols 2023 verdict, para. 5.4.4.

The court considered it established (“встановлено” (established)) that the group conducted “прицільний вогонь з гладкоствольної вогнепальної зброї” (aimed fire from a smoothbore firearm) and that Arutiunian was wounded “шляхом відстрілу з неї мисливського патрону, спорядженого свинцевою картеччю, у натовп” (by firing from it a hunting cartridge loaded with lead buckshot into the crowd) 2023 verdict, para. 5.4.4. The group then returns for a minute to the corner of the hotel and leaves uphill, past the NBU to Bankova Street 2023 verdict, para. 5.3.2; it does not appear in the episodes after 09:12. The group’s officers were neither questioned nor charged in the 2023 trial.

The court writes that it identifies the officers “як спецпідрозділу «Беркут» ГУМВС в АР Крим” (as the “Berkut” special unit of the Main Directorate of the Ministry of Internal Affairs in the Autonomous Republic of Crimea), and that their affiliation with the Kyiv PMOP is “беззаперечно спростована” (refuted beyond dispute) 2023 verdict, para. 5.3.2. The basis is the video “Exclusive: The Kyiv Maidan through the Eyes of the Sevastopol Berkut” (orig. in Russian), filmed among these officers 2023 verdict, para. 5.3.2. A witness from the Main Directorate of the MVS testified that “нетипова форма була у спецпідрозділу «Беркут» АР Крим” (the “Berkut” special unit of the AR of Crimea had an atypical uniform), while blue camouflage was the regular Berkut uniform 2023 verdict, para. 5.5.1.

According to the OPG/SBI indictment (before the court since October 15, 2024; secondary data), this was the Sevastopol Berkut company of the MVS Directorate: around 09:00, Fort-500, lead buckshot, 3 dead and 3 wounded—according to media reports, Arutiunian, Vaida, Baliuk. This is the prosecution’s position, not a verdict. Neither the video nor the prosecution supports the court’s wording “Autonomous Republic of Crimea”; Claude considers the affiliation stated in the verdict not established. For the episode estimates it makes no difference: in either case these are visiting Berkut officers, not the special company and not the PMOP.

The argument that the Kyiv Main Directorate of the MVS could not give orders to Crimean units is not supported by the documents. In November–December 2013, regional Berkut units were attached to the Kyiv Main Directorate by telegrams from the deputy minister, and each time the initiative came from the Main Directorate itself 2023 verdict, para. 5.6; according to a witness, the attached units’ actions on the ground were directed by a colonel on behalf of the head of the Main Directorate 2023 verdict, para. 5.5.1.

A separate in absentia verdict of September 4, 2026 (Shevchenkivskyi court; officers of the Sevastopol Berkut company, which the prosecution identifies with the group with green markings, were convicted) established that on the evening of February 20, from about 19:00 to 05:00 on February 21, the company was issued 40 AKS-74U rifles and 12,000 rounds, which were then taken to Crimea in absentia verdict of the Shevchenkivskyi court (2026); the company commander gives the date as February 22 in a video statement examined by the court in absentia verdict of the Shevchenkivskyi court (2026). These rifles have no bearing on the morning episodes, either way.

The only armed man in blue camouflage among the special company—“єдиний озброєний у такому одязі” (the only armed person in such clothing) 2023 verdict, para. 5.5.1. According to the court’s finding, he received his AKMS from special company officers—“від бійців якого відразу отримала АКМС” (from whose fighters it immediately received AKMS rifles) 2023 verdict, para. 5.3.2. At 09:05 he ordered the special company officers to fire; the court called the order “явно злочинним через вимогу її невибіркового застосування” (manifestly criminal because it required its indiscriminate use), but aimed at evacuating the palace rather than at a terrorist act; that he himself received such an order from above is “недоведений” (unproven) 2023 verdict, para. 7.

The court tracks him on video from 09:03 to 09:27:40 near the palace and the barricades, then behind the concrete barricade 2023 verdict, para. 5.3.2. In Vysota at 09:17:05 a man in blue camouflage can be seen firing an AKMS next to a group of special company officers (description and identification by Claude); this is consistent with the court’s description but establishes nothing new.

The verdict sentenced the defendant—a lieutenant colonel, the regiment’s deputy commander for personnel (the same anonymized label in the heading of the verdict and in its operative part)—to life imprisonment 2023 verdict, introductory part, para. 1: he was convicted under Part 2 of Art. 365 and subparas. 1, 5, 12 of Part 2 of Art. 115 of the Criminal Code, with no reference to acting under orders (Part 4 of Art. 41), and stripped of the rank of “підполковник міліції” (lieutenant colonel of the militia) 2023 verdict, operative part. In justifying the sentence, the court describes his role “як начальника, який очолив протиправні” (as the commander who led the unlawful) actions of subordinates and notes that he “ухилився від суду” (evaded trial) 2023 verdict, para. 15. Names in the verdict are anonymized.

The Omega commander assigned a sniper pair, a sapper and a machine gunner to Zakhidnyi because the detachment’s snipers “були без снайперської зброї, мали лише АК 5.45мм” (had no sniper weapons, only 5.45 mm AKs) 2023 verdict, para. 5.5.1.

Based on video, the court describes how at 09:16–09:20 “бійців в чорній формі без жовтих пов`язок” (fighters in black uniforms without yellow armbands), armed with sniper rifles, join the special company 2023 verdict, para. 5.3.2; at the snow barricade, “із застосуванням окремими з них вогнепальної зброї в бік активістів” (with some of them using firearms in the direction of the activists) 2023 verdict, para. 5.3.2; from 09:30 at the concrete barricade, men without armbands with AK-74s, sniper rifles and a machine gun 2023 verdict, para. 5.3.2. Zakhidnyi had no sniper rifles, so in Claude’s estimate the men without armbands carrying sniper rifles at 09:16–09:30 were more likely Omega-4.

The Omega-4 sniper says that behind the snow barricade he “позицій не займав” (did not take up positions) 2023 verdict, para. 5.5.1: by his account, he himself watched the hotel from the concrete barricade, while his partner periodically moved around to observe. This does not conflict with the video: if the man in the frames is from the pair, he is more likely the partner. Another member of the same four-man group, the commander of Omega’s special-equipment team, testified that at the barricades the Omega snipers took up various positions and reported the situation by radio; he did not say which positions 2023 verdict, para. 5.5.1. Shortly before 09:50, among the special company officers, “тривалий час” (for a long time) one can see an “особа в чорній формі без жовтих пов`язок на руках, яка була озброєна СГД калібру 7,62х54 мм” (a person in a black uniform without yellow armbands, armed with an SGD of 7.62x54 mm calibre) 2023 verdict, para. 5.4.91. According to the SBI, the Omega-4 machine gunner fired around 09:25 and around 09:43; this is a charge (the case is in the Podilskyi court), not a court finding.

The Internal Troops and Omega also had markings: yellow or gray squares on the arm or shoulder (testimony: “із жовтими квадратами на формі” (with yellow squares on the uniform) 2023 verdict, para. 5.5.1, “із жовтим квадратом на плечі” (with a yellow square on the shoulder) 2023 verdict, para. 5.5.1; defendant: “жовті квадрати для ВВ” (yellow squares for the Internal Troops) 2023 verdict, para. 5.1). In a poor-quality frame such a square could be mistaken for the special company’s yellow armband.

Reference frames (frame descriptions by Claude): the special company’s wide yellow armband on the left sleeve—09:08:16 and 09:20:11; the green marking of the group with green markings—08:51:30 and 08:59:54; a yellow square on the shoulder of an officer with the word “МІЛІЦІЯ” (MILITSIYA) by the concrete blocks—10:10:07; a gray square on the shoulder of a security officer by the concrete blocks at 09:59:43 (frame G-33.3). Red items on uniforms, according to the Omega sapper’s testimony, are first-aid kit handles, not identifying marks (2023 verdict, para. 5.5.1). Frames from Vysota (frame descriptions by Claude) show snipers in black with no visible armband: 09:19:18 (lying in the alley, scoped rifle on a bipod), 09:20:06 (kneeling by the parapet; left sleeve not visible), around 09:26 (behind a barricade, probably the snow one).

At 09:55:48 a frame (description by Claude) shows a group in black by a brick wall on Bankova Street, no armbands visible, the first man holding a pistol; for these minutes the court describes the arrival of a unit “без жовтих пов`язок” (without yellow armbands) with pistols, AK-74s and sniper rifles 2023 verdict, para. 5.3.2. At 10:11:14 a frame (description by Claude) shows APCs and a KamAZ truck cab at the concrete barricade; the Omega commander describes 2–3 Berkut officers on the blocks behind the KamAZ truck bed 2023 verdict, para. 5.5.1, but he himself arrived around 11:00. The affiliation of the men without armbands cannot be determined from these frames.

According to its commander, the main Omega unit arrived on Bankova Street around 11:00 2023 verdict, para. 5.5.1; a journalist saw “кілька десятків” (several dozen) people in black at the corner of Bankova Street as early as 10:10—whose they were has not been established 2023 verdict, para. 5.5.1. The unit took up positions in the club only after the lull began at 10:31:50.

The commander says the snipers on the 3rd floor of the club came under fire “з 5-7 поверхів” (from the 5th–7th floors) of the hotel and from the palace attic; the bullets were not recovered 2023 verdict, para. 5.5.1. This unit is linked to the only case against a security-force sniper: the death of Khrapachenko (see the table). Alpha and UDO were on the roofs of the Presidential Administration and the Cabinet of Ministers; almost nothing in the episode estimates rests on their presence, except the 10:21–10:31 window (see below).

Table 3.5. Armed groups on the Maidan side: the same table.

The court’s general finding on leadership: from the Maidan stage, the shots at the security forces “взагалі ніяк не коментувалися” (were not commented on at all), and there was no control over the armed forward groups 2023 verdict, para. 5.3.2. The court did not establish who commanded the armed groups.

Until 09:16 the security forces by the palace stood in the open, 5–50 m from the shooters with guns; that is how a special company officer was killed. From 09:20:40 they were behind the snow barricade—a heap of debris and sacks opposite the metro—and from 09:30 behind the concrete one: it had been built back on February 19, from 14:12 to 19:37 according to the NBU camera 2023 verdict, para. 5.3.1. These are blocks up to 3 m high 2023 verdict, para. 5.4.70 behind three trucks, with an APC in the gap and two more APCs in the rear 2023 verdict, paras. 1, 5.3.1, 5.3.2.

The court describes the fire from behind it as follows: “висунення голови стрілка лише в необхідних межах” (the shooter exposing his head only to the extent necessary) 2023 verdict, para. 5.4.72. By Claude’s calculation, the protesters were 60–165 m away from there, and the barricade stood 5–7 m higher than Arkada.

Therefore the absence of wounded security officers after 09:20 is a weak argument against fire from the Maidan side: the likelihood ratio, in Claude’s estimate, is about 1.2–2.4, whereas without cover the argument would be nearly decisive. Positive traces of fire at the security forces exist for 09:20–09:30 2023 verdict, paras. 5.3.2, 5.4.84; after 09:31 there are almost none—there is the court’s general finding that the “скритий характер обстрілу” (the covert nature of the fire) persisted, with no source named 2023 verdict, para. 5.3.2, and Alpha’s testimony about coming under fire around 10:00.

According to the court’s finding, the hotel was an “об`єктом масованого обстрілу” (the target of massive fire): the vast majority of 7.62×39 hits came from the top of the street and from the metro, where the security forces stood, but room 825 was hit from the direction of the Conservatory, room 925 from the direction of the Main Post Office, and shots were fired from room 1116 toward the barricades 2023 verdict, para. 5.3.3. On the 2nd-floor windowsills of the Cabinet of Ministers club (No. 7) there are marks left by sniper rifle rests and two 5 mm holes, and on the roof, five metal balls (homemade shot); time not established 2023 verdict, para. 5.3.3. From the direction of building No. 8, “поодинокі” (isolated) traces of fire 2023 verdict, para. 5.3.3. More in Part 4.

In 18 of the 48 episodes, the share of “other security forces” is 25% or more. Below are the “special company / other security forces / Maidan side” shares in percent and who, in Claude’s estimate, is behind the other-security-forces share (based on the episode rationales). Summed over all 48 episodes, the other-security-forces shares come to about 11 deaths; these 18 episodes account for about 8.

Table 3.6

Summed by unit (Claude’s estimate):

• the group with green markings accounts for about 1.9 expected deaths (3 episodes);
• security forces in the area of the alley, the snow barricade and the adjacent buildings (No. 7–8) at 09:17–09:30—primarily Omega-4, Zakhidnyi and the deputy commander of the Berkut regiment (PMOP)—about 2.9 (8 episodes);
• the man with the 7.62×54 rifle at the concrete barricade around 09:54—about 1.0 (2 episodes);
• security forces at the concrete barricade and in the adjacent buildings (club No. 7, Cabinet of Ministers roof) at 10:21–10:28—about 1.4 (4 episodes);
• the main Omega unit—about 0.85 (Khrapachenko).

The remaining ≈2.9 are spread in small shares across 30 episodes in which the special company or the Maidan side is clearly more likely.

Table 3.8.

Table 3.9.

• Were protesters wounded by gunfire before 8:57? There are no such episodes in the verdict. Katchanovski, citing the investigation without naming a document, writes of 10 wounded before 9:00—and that window also includes 8:57–9:00. Three and a half hours of shooting at the Stela make early injuries plausible. Claude’s assumption: 50%.
• If there were, the share of fire from the Maidan side. For the Maidan: accidental hits when firing at the security forces through the crowd, and the claim about “other sectors.” For the security forces: the prosecution’s claim about buckshot in the non-lethal weapons, which the court, however, did not confirm. Assumption: 55%.
• If the first were the 8:57–9:00 episodes, the security forces’ share. For individual episodes Claude gives the security forces 60–90%, but the episodes are linked: the same minutes, the same place, the same group of shooters in the frame. Assumption: 80%.

Security forces = e × (1 − m) + (1 − e) × c = 0.5 × 0.45 + 0.5 × 0.8 ≈ 63%, Maidan side ≈ 37%. The special company is near zero in both branches.

The conclusions of the forensic medical commissions on the circumstances of the shot (sector, relative positions) were made without a ballistics expert, for 121 episodes, and the court ruled them inadmissible in that part 2023 verdict, para. 5.4.129; it did not rule the data on time and place inadmissible, treating them not as expert conclusions but as a restatement of investigation materials 2023 verdict, para. 5.4.129.

The court listed the errors in these examinations 2023 verdict, para. 5.4.129; most of them shift the sector forward, toward the security forces’ positions. The rule was applied in both directions: for example, for Ushnevych the court accepted a comprehensive examination involving a ballistics expert (sector: the hotel) and rejected a later medical one with a different sector 2023 verdict, para. 5.4.74. At the same time, the court draws situational conclusions itself—both toward the hotel and toward the security forces.

The 2014 examinations and the automated TAIS system (January 2015) did not link the bullets to Berkut rifles. Manual comparisons with reference samples in December 2015–2016 did 2023 verdict, para. 5.3.3; the Kharkiv examinations of 2019 and 2021, carried out without the earlier experts, largely confirmed them and corrected them in a few places 2023 verdict, para. 5.3.3. The court found the Kharkiv examinations reliable and the 2014 examinations and TAIS erroneous in this respect because of their equipment: microscopes with magnification up to 60–64× versus an electronic comparison microscope up to 150× 2023 verdict, para. 5.3.3; the experts explained the discrepancies by barrel wear and the limitations of the system 2023 verdict, para. 5.3.3. This is a technical justification, not falsification; but the relevant rifles have been destroyed—cut into parts, so control test-firing is impossible 2023 verdict, paras. 5.2.2, 5.3.3—and the examination can no longer be independently re-checked. The estimates account for this as a common risk of reference-sample error.

The 29 bullets and fragments (plus one more fragment that remained at the laboratory) seized during the inspection of Hotel Ukraina are inadmissible: there is no document showing where and how they were seized 2023 verdict, para. 5.3.3. This is a loss for everyone: they could have shown both fire at the hotel and fire from it. The six-groove bullet in the Ushnevych case was rejected because its connection to the episode was not confirmed 2023 verdict, para. 5.4.74. The timing of the radio communications overlaid in Vysota was not accepted 2023 verdict, paras. 5.3.1, 5.5.1. The court did not examine the video reconstruction of three deaths (by Claude’s cross-check, the one by SITU): the victims’ side did not file a motion for it 2023 verdict, para. 13.10.

Some of the court’s findings are made for a group of episodes at once:

• eight episodes at 09:02–09:09 are combined into one group; the involvement of “the special company and other security forces with AKMS rifles” in the whole group is inferred from the evidence taken together: one identified bullet (in the Saienko case), 7.62×39 caliber in three others, visible fire and the general direction 2023 verdict, para. 5.4.20;
• the Blok episode is included in the “28 episodes” group 2023 verdict, para. 5.4.50, although for this episode itself ballistics ruled out the special company’s involvement 2023 verdict, para. 5.4.50; the group finding about “the special company and other security forces” 2023 verdict, para. 5.4.70 was extended to it as well;
• the paragraph on the situation at the Arkada parapet is repeated almost verbatim in 13 sections (with variations, in roughly two dozen more) 2023 verdict, paras. 5.4.75, 5.4.76, 5.4.92;
• five woundings at Arkada within less than two minutes (≈10:07–10:09) are resolved with a single common formula 2023 verdict, para. 5.4.105.

Therefore, in the estimates, a group finding is not carried over to an individual episode in either direction (section “Rules of analysis”).

Sometimes the court’s summary or finding is firmer than what it rests on. A witness says that Moisei “скоріш за все присів, але дивився… в напрямку алейки” (most likely crouched, but was looking… in the direction of the alley) 2023 verdict, para. 5.4.32, while the court says he was turned “передньою частиною тіла” (with the front of his body) toward the alley 2023 verdict, para. 5.4.32. In the Tsarok case, the victims’ representative admits that the position of the body “залишається не з'ясованим” (remains unclarified) 2023 verdict, para. 5.4.109, while the court writes: “безспірно встановлено” (established beyond dispute) that the shot came from the direction of the concrete barricades, that is, from the security-force area 2023 verdict, para. 5.4.109.

In the Aksenyn and Baidovskyi cases, the court first stipulates that the direction of the wound track has not been clearly determined or that a ricochet is not ruled out, and then, right next to it or a few paragraphs later, draws a categorical conclusion about the direction 2023 verdict, paras. 5.4.28, 5.4.55. The reverse also happens: the Omega commander (according to the video recording of his questioning in court on April 25, 2019), in the transcript of the questioning, speaks of a single shot at dusk, around 16:00–17:00, more categorically (“На моє враження, так, однозначно постріл був проведений з нашого боку…” (My impression is that yes, the shot was unequivocally fired from our side…)) than in the court’s summary 2023 verdict, para. 5.5.1; the transcript is automatic. Therefore, where possible, the estimates rely on the testimony itself rather than on the summary.

The verdict is a text of almost a million words, and it contains mechanical errors:

• expert examination numbers from other episodes in a boilerplate sentence (for example, for Opanasiuk and Kornieiev 2023 verdict, paras. 5.4.63, 5.4.66);
• the time “09:07:59” from the video author’s caption (02:06 into the recording), reproduced without qualification, although 9 seconds of recording earlier (01:57) the same author’s caption gives 08:59:34: 8 min 25 s pass by the captions, 9 s by the recording 2023 verdict, para. 5.4.4;
• “19 лютого 2020 року” (19 February 2020) instead of 2014 2023 verdict, para. 5.4.19;
• “seven” instead of “eight” episodes in the group 2023 verdict, para. 5.4.19;
• the same person under different anonymized numbers within one section;
• here and there, real surnames break through the anonymization.

Anonymized numbers (ОСОБА_N) are not matched to names in the dossier, except for the dead and wounded whose episodes the court examines separately.

The court found justified the caveat of the prosecutor and the victims’ representative: the data on activists’ injuries are incomplete, since they count only official requests for medical help, and many of the wounded did not dare to seek it; serious wounds, however, are harder to conceal 2023 verdict, para. 5.6.

From 2014 to 2021, the same bullets went through several waves of examinations. They diverged from one another, and the court itself records this: “прямі невідповідності щодо результатів ідентифікації зброї” (direct inconsistencies regarding the results of weapon identification) (2023 verdict, para. 5.3.3). A summary by group is in the cards below: each group has two estimates—how likely it is that it is erroneous or unreliable, and how likely that it is falsified. The probabilities of error for the ballistic groups come from the ballistics analysis (Claude’s estimate based on the internal consistency of the examinations; Claude did not compare the marks itself). The probabilities of falsification and the estimates for the medical examinations are new estimates by Claude; the grounds are in the “Why these estimates” row of each card and in the paragraphs after the cards.

For groups B and C (under 2%): their errors did not help the prosecution, and the experts explained the discrepancies in court, including by admitting their own errors (2023 verdict, para. 5.3.3). For group D (≈3%): at the level of the upper bound of the Kharkiv estimate—the same experts, and matches changed within one line. For the Kharkiv examination (1–3%): a different lead institution, the earlier experts excluded, corrections not all in one direction (see below). These arguments do not cover substitution of an object before the examination; that is estimated separately, based on the chain of custody.

Comparing bullets with each other in spring 2014 produced a mass false match: eleven objects were deemed to have been fired from one barrel, while the later examinations distributed them across seven barrels. Of the ten dead whom Katchanovski attributes to a single weapon (Katchanovski, ch. 6, p. 185), nine, according to the final examinations, were killed by six different special company rifles, and the court did not link the jacket from the Shymko case to the body (episode data; 2023 verdict, para. 5.3.3). The negative TAIS result is a failed search, not a proven exclusion.

The court explained the contradictions in the early examinations by imperfect equipment and the large number of objects: microscopes up to 60–64× versus a comparison microscope up to 150× (2023 verdict, para. 5.3.3), and rejected the conclusions themselves as unsupported by the case materials and contradicting the Kharkiv ones (2023 verdict, para. 5.3.3). In court the experts also explained something else: once comparison material appeared—the 2012 test-firings—it became possible to distinguish stable barrel marks from wear marks; this is how an expert explained the “34 випадки зміни груп” (34 instances of group changes) (2023 verdict, para. 5.3.3).

• It was ordered by the jury court itself on November 22, 2016, on a joint motion of the parties, including the defense, and the court excluded by name from the commission the experts who had already taken part in examinations in the case (2023 verdict, para. 5.3.3).
• The lead institution is a different one: the Bokarius Institute in Kharkiv. The equipment is a comparison microscope up to 150× (2023 verdict, para. 5.3.3).
• The work took a long time: from the order to the second report (on cartridge cases, March 1, 2021), more than four years passed (2023 verdict, para. 5.3.3).
• The result does not mechanically repeat the earlier ones. Some corrections favor the defense: the Blok fragments were found not to have been fired from a Kalashnikov-type weapon, and the court excluded the special company in this episode (2023 verdict, paras. 5.3.3, 5.4.50). Four bullets that the examination of December 16, 2015 attributed to one rifle had already been reassigned to another by the examination of June 14, 2016, and Kharkiv confirmed the corrected match, not the original one (2023 verdict, para. 5.3.3).
• The matches add up by place and time. If the matches were random errors, bullets from one barrel would be scattered along the street and across the hours. For most barrels this is not the case: barrel A—the footbridge and the alley by the palace, 09:08–09:25 (Saienko, Ilkiv, Baidovskyi); barrel B—the alley at the end of the green fence, 09:28 (Kemskyi, Dziavulskyi); barrel C—at the Arkada parapet, 09:44–09:46 (the thigh wound of Ushnevych, Varenytsia); barrel D—at the metro entrance, 10:07 (Pankiv and the wounded nearby). This is an independent argument in favor of the matches. The exception within barrel A is Saienko: the other hits from this barrel came from the direction of the palace, from the left, while Saienko was wounded on the right side (2023 verdict, para. 5.4.20), so his match is rated as less reliable (≈18%, episode card). This argument does not protect against one kind of error: if the reference samples themselves had been mislabeled, the clustering would remain, but the person the barrel is assigned to would change. For the question of whether this is a special company rifle, that does not matter: all the candidates are regiment weapons.
• The reference samples were checked selectively: the experts test-fired 381 of the regiment’s rifles anew and found not a single discrepancy with the 2012 reference samples (2023 verdict, para. 5.2.2). By Claude’s calculation, this limits the share of mislabeled reference samples to about 0.8% (at 95% confidence).

• The comparison was made only against the 2012 reference samples. About 20 of the relevant rifles were not submitted for examination: they were destroyed, their parts were found sawn apart, and control test-firing from them is impossible (2023 verdict, paras. 5.2.2, 5.3.3; on the fate of the weapons, see Part 8, “What happened to the weapons, February 21–24”). For these barrels there is nothing to check the reference samples against. This is the main residual risk, common to all episodes with a bullet (≈4%, Claude’s estimate).
• The defense objected: the packages with reference samples were not sealed, and the labels were damaged or altered (2023 verdict, paras. 4, 5.2.2). For the 381 re-checked rifles this was refuted by test-firings; for the missing ones it was not.
• Wear: according to an expert, after about 200 shots the fine marks of the barrel bore change, and after 1,000 identification becomes difficult (2023 verdict, para. 5.3.3). Comparing worn barrels against reference samples two years old amounts to choosing the best match among a few candidates, and the risk of a false match here is higher than in standard checks.
• The earlier conclusions were in the case file. In 2015–2016 the investigator attached them to the orders for new examinations (2023 verdict, para. 5.3.3), and the court, in assessing the Kharkiv examination, itself writes that it confirmed most of the earlier conclusions (2023 verdict, para. 5.3.3). The verdict does not say whether the Kharkiv commission worked blind.
• The court did not inspect the bullets themselves, only photo tables: the prosecution did not submit the physical evidence for examination (2023 verdict, para. 5.3.3).
• The court’s statement that the new equipment “фактично усуває можливість людської помилки” (virtually eliminates the possibility of human error) (2023 verdict, para. 5.3.3) is stronger than any equipment allows. It does not affect the estimates: they build in a risk of error, not zero.

An examination can establish which rifle a bullet was fired from. Who held that rifle on February 20 is a different question, and the answer to it is weaker:

• The defense directly disputed that each officer had his own assigned rifle (2023 verdict, para. 4).
• In several episodes the court itself established that the rifle’s owner could not have fired at the moment of the shot—he was wounded or in hospital—and attributed the shot to an unidentified special company officer (“невстановленого співробітника РСП” (an unidentified officer of the RSP)) who had taken possession of that rifle (2023 verdict, paras. 5.4.38, 5.4.78, 5.4.122).
• According to the prosecutor, the deputy commander of the Berkut regiment unlawfully obtained a rifle assigned to another special company officer (2023 verdict, para. 2).
• The weapons issue log and the substitute cards were destroyed on the order of the regiment commander on February 23 (2023 verdict, para. 11). According to the materials of the 2014 internal investigation cited in the in absentia verdict against the special company commander, 372 Kalashnikov rifles were assigned on paper to officers of the regiment (in absentia verdict against the special company commander (2026)), against 578 rifles on the regiment’s books (2023 verdict, para. 5.2.2), and several officers explained that in the rush the duty officer “міг переплутати та видати незакріплену за ними зброю” (could have mixed up and issued weapons not assigned to them) (in absentia verdict against the special company commander (2026); these are the officers’ own explanations, as given in the in absentia verdict).
• A sniper in the special company’s staff list is a position, not necessarily a rifle. Two of the defendants were, by position, snipers of assault platoons (2023 verdict, paras. 5.1, 5.2.1). One testified that he received the AKMS assigned to him (2023 verdict, para. 5.1); the other, that until February 21 he had an AKMS and a pistol and received his SVD rifle only on February 21 (2023 verdict, para. 5.1). According to the prosecution, the SVD was received together with the AKMS on February 18 (2023 verdict, para. 2); the court did not resolve this dispute.

Therefore attribution from a barrel to a specific officer is unreliable in principle, while the link from barrel to special company is much more reliable: all candidate barrels are assigned to the unit, and most of the missing rifles belonged to special company officers (19 of 24, Part 8). The episode estimates account for this separately: where, according to the court, the rifle’s owner was wounded or in hospital and someone else fired his rifle, ≈5% was added for the possibility that the shooter was not a special company officer (Claude’s estimate).

In 12 of the 48 deaths, a bullet or its jacket from the body (or from clothing on the body) has been linked by expert examinations to a specific special company rifle. According to the ballistics analysis performed by Claude and the episode cards, the probability that such a match is false ranges from ≈4.5% (Dziavulskyi: two bullets from two different rifles, both of which would have to be false) and ≈8% (an intact bullet with a good chain of custody: Pankiv) to ≈18% (a bullet from the folds of clothing and the rifle of a wounded owner: Dmytriv; for Saienko, because of the contradiction in wound direction, see above); this figure includes the common reference-sample error of ≈4%. These errors are not independent: a reference-sample error would overturn several episodes at once. Therefore, even in the episodes with a bullet, the special company is assigned no more than 92% (Appendix A, “Rules of analysis”).

It was not only the early bullet ballistics that was rejected: one early examination that went against the security forces was rejected too. The examination of June 4, 2014 found lead in the barrels of some IZh-81 service shotguns—a sign of firing buckshot, shot or a slug (2023 verdict, para. 5.3.3); the examination of November 25, 2014 concluded that the lead traces in the IZh-81 and Fort-500 shotguns could not have come from standard 12-gauge cartridges (2023 verdict, para. 5.3.3). The court rejected both because their relevance to the case was not proven:

• the IZh-81 shotguns do not figure in the charges, and their connection to special company officers has not been established (2023 verdict, para. 5.3.3);
• some of the Fort-500 shotguns arrived in Kyiv from other regions in January 2014, and the time the traces were formed is unknown—they could have appeared in January as well (2023 verdict, para. 5.3.3);
• one shotgun on the list was registered as missing after February 20, and it is unclear how it ended up being examined;
• the initial report on the Fort-500 was not submitted to the court (2023 verdict, para. 5.3.3).

The court established “безспірно” (beyond dispute) that some special company officers were armed with Fort-500 shotguns, but stipulated that this in itself does not mean the shotguns were loaded with buckshot; the use of Teren rubber bullets from them has been proven (2023 verdict, para. 5.3.3). The court linked the buckshot in the Arutiunian case to another group of security forces (Part 3, “Who was where”).

In Part 9 his claims about the examinations are compiled into a table covering seven groups, and for each it is estimated separately how likely it is that the examination is erroneous or unreliable and that it was falsified (all probabilities are Claude’s estimates). Overall: where he speaks of unreliability, he is most often right, and the court agrees with him on this; where he speaks of falsification, the sources show no direct signs of it. In detail: Part 9, “General arguments”.

• Right, and the court agrees.

  The firing sectors in the 2015 commission forensic medical examinations (FMEs) are unreliable (over 90%): the court ruled them inadmissible (2023 verdict, para. 5.4.129). For the Ushnevych bullet, the chain of custody is broken (over 90%): the court itself established that clearly different bullets were examined (2023 verdict, para. 5.3.3), but this defect does not concern the special company—the court excluded its involvement in Ushnevych’s fatal wounds (2023 verdict, para. 5.4.74). The bullet holes in helmets and shields were not examined ballistically—the gap is confirmed (2023 verdict, para. 5.4.93). The contradictions between the waves of examinations are also real (2023 verdict, para. 5.3.3).

• Falsification not supported.

  He calls the Kharkiv examinations of 2019 and 2021 fabricated (Katchanovski, ch. 8, p. 220): falsification 1–3%, error for an individual bullet 4.5–18%. The revision of about 40 earlier conclusions (Katchanovski, ch. 8, p. 229) was not unexplained: the experts explained the discrepancies in court (2023 verdict, para. 5.3.3), and the Kharkiv examinations, in the court’s assessment, largely confirmed the conclusions of the earlier ones (2023 verdict, para. 5.3.3). The SITU model: falsification under 3%; the wounds on the models are placed as in the FME texts. The medical part of the FMEs: falsification under 2%, error 5–10%, while his retelling of the wounds diverges from it in places (for Parashchuk: Katchanovski, ch. 3, p. 67; 2023 verdict, para. 5.4.87). For the sectors in the FMEs (5–15%) and the Ushnevych bullet (5–20%), intent cannot be ruled out, but the court discarded these materials and did not rely on them. The early ballistics of 2014–early 2015, by contrast, he trusts (Katchanovski, ch. 6, p. 185), although it is erroneous with a probability of 80–90% and the court rejected it (2023 verdict, para. 5.3.3).

• What this means for trust in the ballistics.

  His critique knocks out the sectors in the forensic medical examinations and individual items with a broken chain of custody—the court had already excluded those—but not the Kharkiv matches of bullets to special company rifles: their risk remains a risk of error (4.5–18% per episode), not of fabrication. And a mistaken bullet match does not automatically shift the shot to the Maidan side.

The court accepted as reliable the following order, from the bottom up, starting at the Maidan: on the right, Hotel Ukraina; on the left, the October Palace and Kinopalats behind it; on the left, the alley with the green fence leading from the palace to the roadway toward the NBU; across the street, the snow barricade and the concrete barricade, “перед і за перехрестям з вул. Ольгинська” (before and beyond the intersection with Olhynska St.); on the right, the building with a dome, the entrance to Khreshchatyk metro station; across Olhynska Street on the right, building No. 8; on the left, building No. 7, the Cabinet of Ministers club, with the NBU behind it 2023 verdict, para. 5.3.2. Two defendants confirmed the same order 2023 verdict, para. 5.1. The hotel is a 15-story building 2023 verdict, para. 5.3.3; building No. 8 has two stories 2023 verdict, para. 5.3.3. Below, “left” and “right” are as seen from the bottom, from the Maidan.

The court took elevations above sea level from the publicly available Google Earth program, which the parties also used 2023 verdict, paras. 5.4.98, 5.4.107, 5.4.125. The court did not factor in the height of the barricade itself (the blocks and trucks) 2023 verdict, paras. 5.4.98, 5.4.107.

Table 4.1

The street rises from the Maidan toward the government quarter: about 18 m from Globus to the concrete barricade, about 5 m from the sidewalk at the Arkada to the barricade. The verdict gives no floor-by-floor heights for the hotel, the Cabinet of Ministers club, building No. 8 or the palace. The court applied the height difference to both sides. For Polianskyi, it found that the elevation data “об`єктивно корелюють” (objectively correlate) with the vertical direction of the wound track 2023 verdict, para. 5.4.107. For Melnychuk, the court accepted the defense’s argument as well-founded: an aimed shot was possible from the hotel but not from the concrete barricade, because of barricade elements, trees and people 2023 verdict, para. 5.4.125. The court then called the same height difference consistent with the wound track 2023 verdict, para. 5.4.125.

Court measurements during investigative experiments:

• the place where Ushnevych was wounded at the Arkada parapet is 87.83 m from the southeast corner of the hotel 2023 verdict, para. 5.4.74;
• the concrete blocks (represented in the experiment by a MAZ truck) are 6.55 m from the fence of the Cabinet of Ministers club and 166.01 m from the nearest corner of the hotel 2023 verdict, paras. 5.4.83, 5.4.93.

Calculation from the court’s coordinates (Claude’s estimate; coordinate precision ±1″, i.e., about ±30 m in latitude and ±20 m in longitude):

• from the sidewalk at the Arkada to the left part of the concrete barricade: about 100 m up the street, to the southeast;
• from the Arkada to the palace terrace: about 170 m down and across the street, with the terrace 5 m lower;
• from the barricade to the palace terrace: about 270 m, a drop of 10 m; to Globus: about 285 m, a drop of 18 m.

The downward angle from the barricade to a person sitting at the Arkada is about 3°, not counting the height of the blocks and trucks. From the upper floors of the hotel it is much steeper, but neither the floor nor the firing point was established in any episode.

• In video from February 21 taken from the concrete barricade, the area at the Arkada parapet is in unobstructed view, “навіть, стоячи на дорозі” (even standing on the road) 2023 verdict, para. 5.3.3.
• In the other direction, from the flower beds at the Arkada, the upper part of the concrete barricade (the truck bodies) is visible above the snow barricade 2023 verdict, para. 5.3.3.
• The Omega-4 machine gunner, questioned as a witness, acknowledged that from his position at the blocks, “з землі” (from the ground), the area at the parapet was in unobstructed view 2023 verdict, para. 5.5.1; asked follow-up questions, that he saw mostly the top of the parapet and does not remember its base (hearing of May 21, 2019, ≈05:43–05:44). The NBU camera filmed him at the passage between the blocks at 09:33:55 2023 verdict, para. 5.5.1.
• The Arkada parapet is clearly higher than 110 cm. For this reason the court accepted the defense’s doubts as well-founded and found that a wound to Ushnevych’s abdomen above the parapet was objectively impossible 2023 verdict, para. 5.4.74. A person sitting right behind such a parapet is protected from fire coming down the street up to a height of roughly one meter. This only works if the posture is established, and for Ushnevych it was established from a single eyewitness (see card 5.4.74).

“Could” means that three conditions are met at once: presence at that minute, direct line of sight, and a caliber that does not contradict known data. ✔: presence and line of sight confirmed by video, an examination or several testimonies; ◐: line of sight exists, but presence at that minute is disputed or only claimed; ✖: no line of sight or no presence. The positions themselves and their sources are given in Part 3. The table is Claude’s summary based on the analysis of positions, not a finding of the court.

Table 4.2

The time (≈09:10) and place of Oleksii Bratushka’s death were not established by the court, and he is not included in the table. The table shows who had the opportunity to hit, not who hit. The weight of each possibility in the estimates depends on the wound, posture and ballistics of the particular episode (Part 5).

Most of the examinations were carried out on February 24–26, 2014, some in March, June, August and October 2014–2015. The direction in the reports was established by “візуванням” (by sighting) with a laser pointer and recorded as “міг бути утворений” (could have been formed), that is, as possible rather than the only one. None of the traces below has a time, except the city-light billboard and the shield in the photo.

Table 4.3

The court formulated its summary of the traces as follows: the street from the hotel to Olhynska was “об`єктом масованого обстрілу з різних напрямків” (the target of massive fire from various directions). Toward the hotel and the Maidan, fire came “найбільш масово” (most massively) from the top of the street, from the direction of the concrete barricade, “що умовно охоплюється напрямком з боку будівлі НБУ та клубу КМУ” (which is broadly covered by the direction from the NBU building and the Cabinet of Ministers club); “часто” from the alley; “поодинокі” (isolated) from the roof of the NBU, from the direction of building No. 8, and from the roofs at an anonymized address 2023 verdict, para. 5.3.3. In the opposite direction, toward the government quarter, shots were fired from the direction of the palace and from the Maidan side 2023 verdict, para. 5.3.3. This is a summary for the whole day. For any single episode it does not say who was firing at that minute.

On the synchronized video (screen 3/2, 48:38) the court records “візуалізацію на дереві слідів пострілу з боку готелю «Україна» в напрямку їх групи зі спини” (the visualisation on the tree of traces of a shot from the direction of the Hotel Ukraina towards their group from behind). After this, the group turns toward the hotel windows; the footage runs until 09:41:44 2023 verdict, para. 5.4.75. On this, together with the general reaction of the people, the court bases its finding about the wound in the back of a wounded man from the same group 2023 verdict, para. 5.4.75.

The court does not specify which tree this is. On the first tree below the flower bed, the examination of February 26, 2014 found no traces from the direction of the hotel 2023 verdict, para. 5.4.75. There is no contradiction if the trace was on a different tree, but it is not confirmed by an examination either. Therefore, in Claude’s estimate, this is not a separate physical trace but part of the observation of the group’s reaction.

Table 4.4. Cartridge cases. Cartridge cases show where shots were fired from, but not at what.

Table 4.5. Buildings: what was examined and what was found, on both sides.

• The top of the street. A great deal of fire from the top of the street at the place where people died near the Arkada is physically proven: the tree, the flower bed, shields, posts, a city-light billboard, cartridge cases at the barricades. The line of sight from the barricade to the Arkada is proven by video. What is not proven is the time of most of the traces and the link between specific traces and specific deaths.
• The Cabinet of Ministers club. Seven traces on the tree near the Arkada coming from the direction of its 2nd–3rd floors, and rests on the windowsills, are physical evidence of fire from the club. Who was in the club before the main Omega unit arrived has not been established. Therefore, for the 09:43–10:12 window, this is a possibility, not a proven position.
• The hotel. Fire from the hotel is physically confirmed by two traces: room 1116 and a car in the parking lot. The RT and BBC videos add to them. But no trace and no recording has a time and target that would link it to a specific death. Fire at the hotel from the top of the street was far heavier.
• Other buildings on the Maidan side. Horodetskoho Street, Muzeinyi Lane and the Internatsional were examined late and found empty. The Arkada roof, the Conservatory and the palace attic were not examined at all. A late examination that found nothing and no examination at all are different things, but neither confirms nor refutes the claims about shooters.
• Building No. 8. The only examination inside a building on the security-force side in which anything was seized, the examination of the attic, was excluded by the court on procedural grounds, and what was found there is unknown. The time when the special company officers entered the building has not been established. This is a gap of the same kind as the late or never-conducted examinations of buildings on the Maidan side: it proves nothing either way.

• Which tree on screen 3/2 of Vysota (48:38) bears the trace “з боку готелю” (from the direction of the hotel) at about 09:41, and the examination reports for the other trees on the right-hand sidewalk. That would make clear whether there is a physical trace of fire from the direction of the hotel at the group at the Arkada, or only the people’s reaction.
• Dating of the chip on the hotel side of the Arkada parapet.
• Comparison of the Fort-14–17 cartridge case by the club, the 9×19 case and the 7.62×54 case from the concrete barricade with the weapons of Omega, Alpha, UDO and the Internal Troops special detachments.
• What was found in the attic of building No. 8 and what the swabs would have shown.
• Who occupied the 2nd–3rd floors of the Cabinet of Ministers club before ≈11:00.
• The time of the shot from room 1116 and the targets of the shots in the RT video.
• Heights of the firing points (floor of the hotel, floor of the club, the Arkada roof); the downward angles could then be compared with the wound tracks.

One can come across the claim that the early expert examinations of 2014–2015 pointed against the special company and that the court rejected them. For the bullets, this is not so. The 2014 examinations attributed the bullets from the bodies to a single weapon, and in January 2015 the TAIS system concluded that the bullets did not come from the regiment’s 578 rifles. Neither finding linked the bullets to specific special company rifles: the TAIS finding directly favored the defense, and the single-weapon finding did not help the prosecution, whose case was that many officers had fired.

The court found them erroneous: it attributed the contradictions in the early examinations to imperfect equipment and the large number of objects, and rejected the findings themselves as unconfirmed and contradicting the Kharkiv ones (2023 verdict, para. 5.3.3), while the links to special company rifles appeared only in December 2015 and were confirmed by the Kharkiv examinations (in detail: the section “Ballistics and the reliability of expert findings”).

Among the early examinations, it was others that pointed against the security forces: the 2014 examinations of lead traces in service shotguns. The court rejected those too, but for a different reason: it was not proven that these shotguns and these traces related to February 20 (2023 verdict, para. 5.3.3). Bottom line: the court rejected early examinations in both directions—the bullet examinations, which did not help the prosecution, as refuted by examinations on better equipment, and the lead-trace examinations, which favored the prosecution, on grounds of procedure and relevance.

Below are the court’s decisions that can be challenged and decisions the court made against a side’s expectations, measured by one yardstick. Points already covered in Section 4.1 and Appendix C are not repeated here: the rule on sectoral expert findings, the exclusion of about 30 objects from the hotel, the acquittal of all defendants under Arts. 340 and 258, the recognition of the morning fire at the security forces, the word “безспірно” (beyond dispute) for Chmilenko, the wording about the Kharkiv ballistics, and the right to rebellion.

Table 4.11