Casualty exchange ratios: distinguishing fact from fiction

While most estimates are useless and propagandistic, there is some strong evidence and solid analysis worth considering.

As I stated in an earlier piece, “There is zero evidence to support any of the casualty estimates published regarding the Russo-Ukrainian War.”¹ I stand by this claim. Reading the Wikipedia page on the subject reveals a slew of activists who fail to distinguish what they want to be true from what is true.² If someone wants to argue that Russia has suffered greater casualties in the war, they can find dozens of estimate supporting that claim. Similarly, if someone wants to argue Ukraine has suffered greater casualties, they have a whole menu of estimates to choose from.

Image from The Star.³

This agenda-driven information environment is nothing new in war. Since the beginning of mass media and propaganda, almost all armies exaggerated enemy losses while strictly censoring friendly losses. Their fans and haters parrot the number that suits their agenda, portraying their preferred winner as invincible and their preferred loser as tactically inept and disposable. During WWII, Nazi Germany exaggerated Soviet losses to a cartoonish extent. Similarly, Israel today is exaggerating Hamas losses. Reporting on both conflicts have genuinely permanently altered the conventional public understanding of the truth.⁴

Since many people have strongly held beliefs about loss ratios, it is hard to cut through the nonsense of both sides and actually come to a realistic picture of the truth. I will do my best to clear up the confusion with respect to the Russo-Ukrainian War. A few recent pieces of media by Oleksii Arestovych, UALosses, and Defense Politics Asia (DPA) provided interesting insights into how casualty exchange ratios can be realistically estimated using a combination of verified casualty figures and some extrapolation. UALosses makes a compelling case, but Arestovych and DPA’s claims deserve some scrutiny.

In a somewhat recent interview with Patrick Bet-David, exiled Ukrainian politician Oleskeii Arestovych claimed Russia is certain to have suffered greater casualties because “The attacker always loses more than the defender.”⁵ This claim has zero basis in military history. It is a myth that originated from the doctrinal necessity of a 3:1 manpower ratio needed to guarantee an army could capture and hold territory.⁶

Indeed, the opposite of Arestovych’s claim is often true. For example, two of the largest land operations ever conducted, Barbarossa (1941) and Bagration (1944), each had 2:1 casualty exchange ratios in favor of the attacker. In Barbarossa, the Germans managed to kill 566,000 Soviets while only sustaining 182,000 KIA (killed in action).⁷ Similarly, during Bagration, the Soviets managed to kill 381,000 Germans while only sustaining 180,000 KIA.⁸ While modern weapon systems are more precise and lethal, and surveillance certainly makes maneuver less stealthy, there is no reasonable basis to expect the casualty exchange ratios to be the exact opposite of the historical precedent merely because of a misconception. Therefore, we must examine the verified figures and extrapolate ratios from them.

I assessed in early January 2025 that the verified casualty exchange ratio was ~1:1. Given the available verified evidence at the time, this was a reasonable thing to believe since the verified Russian fatalities from Mediazona⁹ and the verified Ukrainian fatalities from UALosses¹⁰ each stood at approximately 70,000. Mediazona and UALosses only count dead fighters that they have independently verified by name, so these are infinitely more useful numbers than estimates put out by governments or newspapers, which are worth as much as the paper they are written on.

In my estimate, I included the caveat that it is unlikely Russian losses exceeded Ukrainian losses since Russia has had artillery superiority over Ukraine since the start of the war and drone superiority for the second half of the war. Since these two weapon systems are understood to be responsible for the majority of casualties in this war, it is reasonable to predict greater casualties on the disadvantaged side. I did not have clear evidence of greater Ukrainian losses, but I argued it could be assumed given the context.

UALosses announced that they have changed their verified count to include MIA (missing in action), in addition to KIA. This was justified on the belief that verified KIA figures alone are “no longer representative of the real level of losses.” While the new figure has a “lower degree of confidence,” it is more likely to be accurate since “bodies are retrieved only a quarter of third of the time, which precludes official status as KIA.”¹¹ Given this change, it is no longer possible to directly compare UALosses to the Mediazona count.

Previously, since UALosses was a count of dead regular army troops and since Mediazona had the total dead regular army troops (once you subtract PMC and inmate PMC volunteers, which are not counted in UALosses), a casualty exchange ratio of ~1:1 was observable throughout the entirety of the war. There is no verified count of PMC or foreign volunteers KIA on the Ukrainian side, so the default Mediazona figure cannot be compare to UALosses in good faith.

Since the Battle of Avdiivka, Russia has been capturing Ukrainian territory at an accelerated rate and the Ukrainian Ministry of Defense has been sending huge numbers of troops to plug gaps, increasing the number of targets for the Russians. These two facts have two expected results. (1) The locations in which people die tend to end up under Russian control. Therefore, a large portion of Ukrainian MIA soldiers are likely to have been KIA. (2) Ukrainian casualties are likely to increase in proportion to Russian casualties since they are filling gaps and increasing the volume of targets for Russian forces. Because of these two observations, it can be deduced that Ukrainian losses have likely increased in proportion relative to Russian losses since early 2024.

For that reason, I do not fault UALosses for changing their methodology. Yet, two facts muddy the use of its database as a “verified” casualty count. Some unknown percentage of those MIA are definitely still alive, either having fled or been captured. Additionally, it is no longer comparable to the Mediazona count since it is not simply the known number of cadavers. Therefore, it is now impossible to know the likely casualty exchange ratio purely from the verified numbers. It incorporates statistical assumptions, rendering it unverified and no longer useful as a tool for deducing a realistic ratio. Despite this, it is the only decent count we have.

Based on the current figures from Mediazona and UALosses, we have a casualty exchange ratio of 1:1.7 in Russia’s favor given that have definitely been 75,631 Russian soldiers killed in action and there have probably been 127,290 Ukrainian soldiers killed in action. For all we know, the Russian MIA could be equally substantial, reducing this figure back to 1:1, but this is unlikely for the reasoned stated above. My conclusion is the real casualty exchange ratio is somewhere between 1:1 and 1:1.5 in Russia’s favor.

Image from UALosses.

In a recent vlog, Wyatt Mingji Lim of Defense Politics Asia, who I consider to be the single best analyst on the Russo-Ukrainian War, extrapolated significantly higher KIA figures on both sides, and used artificial intelligence to corroborate his figures. While AI language models are fallible, as anyone who has used them for math or coding is aware, Wyatt was nonetheless able to extract some interesting information using Grok 3.

Wyatt personally estimated that Ukraine has suffered 1.2 million KIA and that Russia has suffered 500,000. These figures are so much higher than the verified figures that I am reflexively skeptical of their veracity. Grok was similarly skeptical, but eventually came to a similar conclusion as Wyatt once he explained his logic. They both conclude Ukraine can hold on for, at most, another year before their manpower shortage will cripple their capabilities.¹²

Wyatt showed his process for training Grok to be critical of mainstream sources and deduce what a realistic figure is using population size and a detailed history of the war. He had the program calculate how much available manpower Ukraine likely had based on its population size and other factors, narrowing it down to roughly 2,000,000 men. Then he asked the program to calculate how much manpower would likely be lost through each offensive push and from holding the frontline. Accounting for a 30% attrition rate in offensives and cyclical daily frontline attrition, Grok came to a final tally of 900,000 Ukrainian KIA. It similarly calculated the Russian figures based on population, duration of offensives, and estimated daily attrition, coming to a total of 250,000 KIA.

My problem with this approach for casualty estimation is that, based on my experience using them, AI language models are polite and will concede to your argument if you point out their mistakes. They typically present the most widely read mainstream sources, which are quite often propagandistic, but then quickly abandon them as soon as you point out their flaws. This was certainly the case with Wyatt’s session on Grok 3. He pointed out that the mainstream sources were wrong, and the model quickly conceded to Wyatt’s numbers based purely on their logic, rather than finding supporting evidence. It is an interesting approach by Wyatt, but it doesn’t prove anything because we do not have verified evidence of anywhere near that many casualties.

Image from Kalibrated.¹³

Wyatt and Grok’s estimates are that Ukraine has suffered two to four times more casualties than Russia. While I also assess that Ukraine has suffered significantly more casualties than Russia (1.5 times), two to four times is beyond what I believe can be reasonably deduced given the available evidence.

I do not accept the premise that it is two to four higher because both army’s have used roughly equivalent style of warfare for the duration of the conflict with few exceptions. Both have been armed with huge volumes of precision explosive weaponry. Both have done occasional costly offensives that resulted in temporarily higher casualties like Russia’s initial invasion or Ukraine’s Kursk salient,¹⁴ but they both generally stayed safe. Therefore, I do not believe it can be argued, given the available evidence, that either side has suffered more than twice the casualties of the other.

If you, dear reader, think my conclusion is wrong, I would be happy to entertain arguments and accept additional sources. I do not claim to know the real answer here, but I encourage everyone to have a high degree of skepticism when they find a source claiming one side has suffered significantly greater casualties than the other.

1. https://substack.com/home/post/p-156049510?source=queue

2. https://en.wikipedia.org/wiki/Casualties_of_the_Russo-Ukrainian_War

3. https://www.thestar.com.my/lifestyle/living/2023/04/11/of-love-pain-and-loss-at-historic-ukraine-cemetery

4. During the war, German propagandists characterized SS officers as “invincible” and Soviet soldiers as “hordes” that did “human wave” assaults. Soviet human wave assaults were a myth. Indeed, German infantry more often did large infantry charges. They would crowd with such density that they often literally tripped over each other. Nonetheless, German propaganda has colored the American and European conception of WWII history, making it the version repeated by American media, notably Enemy at the Gates.

https://mitpressbookstore.mit.edu/book/9780700621217

IDF press claims to have killed thousands of Hamas militants while sustaining fewer than one thousands KIA themselves. However, the IDF frequently count civilians as “terrorists” to exaggerate enemy losses by a multiple of ~20. Therefore, they likely have a casualty exchange ratio of 1:1 or worse.

https://www.haaretz.com/israel-news/2024-12-18/ty-article-magazine/.premium/idf-soldiers-expose-arbitrary-killings-and-rampant-lawlessness-in-gazas-netzarim-corridor/00000193-da7f-de86-a9f3-fefff2e50000

5

6. https://www.jstor.org/stable/2538780

7. https://en.wikipedia.org/wiki/Operation_Barbarossa

8. https://en.wikipedia.org/wiki/Operation_Bagration

Before anyone says, “Why would you believe the Soviet count? That’s clearly propaganda to hide the heavy losses from meat wave assaults.” First, the Soviets never used meat wave assaults. Only the Germans did. Second, the Soviets were the ones capturing the territory and counting the bodies, so they would know how many died. If anything, the Germans would be the ones likely to underrepresent their losses. This was often the case according the historical record, and muddied the American history of the war. Captured Nazi officers were often the source of American historical records about the Eastern Front. All these points are detailed in this lecture below.

9. https://en.zona.media/article/2022/05/20/casualties_eng-trl

10. https://ualosses.org/en/soldiers/

11. https://x.com/LossesUA/status/1892902871609340069

12.

13. https://web.telegram.org/k/#@kalibrated

14. The Telegram footage from the beginning of this battle until now show a huge number of Ukrainian casualties and a moderate number of Russian casualties. Armored columns of Ukrainian BMP, tanks, and infantry have been routinely attrited using drones including fiber-optic drones which were introduced in this battle. Despite claims to the contrary, there is no evidence Russian forces have suffered a level of attrition as high or higher than Ukrainian forces in this battle.