Clear all

Data analysis

5 Posts
2 Users
Posts: 59
Topic starter

Data analysis of Russian losses with the goal to find a rule of thumb to estimate real Russian losses with daily MoD Ukraine updates:


Chapter I: Personnel

Personnel for 4.8.2023:


MoD Ukraine:

"personnel ‒ about 248490 (+640) persons were liquidated”


Method: Aggregation of daily reports by Ukrainian units.

Probably highly overestimating real kills.


BBC News Russia and Mediazona:

29,217 Russian soldiers who have been killed in Ukraine”

Method: Open source confirmation in Russia.

Highly underestimating real losses.


US Leak:

Western officials and leaks regularly publicly estimate Russian casualties, that is MIA (missing in action), KIA (killed in action) and WIA (wounded in action), somewhere above the MoD Ukraine “liquidated” numbers.

The leak states with “low confidence”:

for 28.2.: KIAus 35500-43000 and WIAus 154000-180000 (MoD U. 28.2..: 149240 KIAukr)

Using 39.250 KIAus and 167.000 WIAus we get a ratio 1:4,25 KIA:WIA

Using 39.250 KIAus and 149240 KIAukr we get a ratio 1:3,80 US:Ukr KIA

Therefore an estimated KIAus for 4.8.2023 at Ratios of 28.2.23 would be: 65392 (=149240/3,80)

and WIAus for 4.8.23: 277916 (= 65392*4,25)

and Sum: 343308

US intelligence is probably the best in the world and their estimates are probably the closest to the truth we can get. But they state themselves that these are stated with “low confidence”.


Rule of thumb:

I personally think that the pentagon leaks were real and so the estimates are what the US believes to be probable.

So they deliver us a got rule of thumb to interpret the KIA numbers by the MoD of Ukraine. To make this estimate more conservative and easy to use:

Estimated real Russian KIAs at a given date: Divide daily MoD Ukraine numbers by 4.

Estimated real Russian WIA at a given date: Equals daily MoD Ukraine numbers.


Posted : August 6, 2023 21:12
Posts: 59
Topic starter

Posted by: @max-beckhaus

Therefore an estimated KIAus for 4.8.2023 at Ratios of 28.2.23 would be: 65392 (=149240/3,80)


Therefore an estimated KIAus for 4.8.2023 at Ratios of 28.2.23 would be: 65392 (=248490/3,80)


I added the brackets later for clarity, well, that didn´t work.

This post was modified 2 months ago by Max Beckhaus
Posted : August 7, 2023 13:04
Posts: 59
Topic starter


for 27.5.23

KIAmed (Killed in action Russians, that is excluding DNR, LNR fighters): “between 40,000 and 55,000 Russian men under the age of 50 died fighting in Ukraine by May 27, 2023”

SWIAmed (seriously wounded Russians, which could not return to service): “we are left with a wounded-to-killed ratio among Russian soldiers that fluctuates between 1.57:1 and 2.84:1“

KIAukr: 206200

Method: Analysing Russian inheritance data for deaths and Russian payments to seriously wounded with the later: „though paperwork complexities and delays with receiving these payments make this calculation thoroughly fuzzy “. They also shows that there may be a bigger MIA problem (using leaked 1st Guard documents), but there is no good data on it.


Using 47500 KIAmed and 206200 KIAukr we get a ratio of 1:4,34 Med:Ukr KIA

Median of KIA:SWIA ratios: 1:2,21 KIA:SWIA med

→ estimates with the caveats of excluding DNR, LNR and an open MIA question and the highlight of giving us data on seriously wounded.


Meduza and US leak:

for 4.8.23:

KiAus: 65392

KIAmed: 57256 (248490/4,34)

KIAmed/KIAus: 88%

The missing 12% could be explained with DNR/LNR losses and/or MIA dropped by the US intelligence as irrelevant but being relevant as the 1st Guard documents and reports of high MIA numbers indicate, of which a lot may actually be dead. E.g. Russian KIAs left on the battlefield and being in Ukrainian hands.

Considering the difficulty of the matter I would argue that the two methods actually support each other. In absence of any good data on MIAs and those being actually KIAs, aswell as on DNR/LNR losses a difference of 12% doesn´t look that bad. More over the rule of thumb would actually be in the middle of both estimates. I personnaly would tend to stay with the US estimates for KIA, since we know that the Meduza KIAs are underestimations and I tend to believe the US intelligence is better than Meduza.


Using KIAus and KIA:SWIA med we get for 4.8.23 with still (very) low confidence:

KIA: 65392

SWIA: 144189 (65392*2,21)

LWIA: 133.727 (277916- 144189)

SWIA: seriously wounded, cannot return to service

LWIA: lightly wounded, can return to service

So this would estimate that Russia lost 209581 soldiers for good.


Updated conservative rule of thumb:

Estimated Russian KIAs at a given date: Divide daily MoD Ukraine numbers by 4.

Estimated Russian WIA at a given date: Equals daily MoD Ukraine numbers.

Estimated seriously wounded at a given date: Divide daily MoD Ukraine numbers by 2.

Estimated disabled to fight at a given date: ¾ of daily MoD Ukraine numbers.


Posted : August 7, 2023 13:09
Posts: 59
Topic starter

Ukrainian personnel losses

US Gen Miley:

Wednesday, 9.11.2023:

He said as many as 40,000 Ukrainian civilians and “well over” 100,000 Russian soldiers have been killed or wounded in the war, now in its ninth month. “Same thing probably on the Ukrainian side,” Milley added.”

US Leak:


15500 – 17500 KIA and 109000 – 113500 WIA

Using 105000nov and 127750feb we get a difference of 22750 Ukrainian casualties in 111 days (from 9.11.22 to 28.2.23) giving around 205/day in that period and 346/day for the 369 days of war till 28.2.23.

It seems plausible, that Ukrainian losses where higher before November 22 with the Kherson offensive and the up to 50000 Russian shells per day.

It is obviously anyone's guess how Ukrainian daily losses may have looked like since 28.2.

Extrapolating the US leak numbers from 28.2.23 to the 7.8.23 we get:

22263 – 25135 KIA and 156558 – 163022 WIA

I would guess that this represents an upper ceiling, since it doesn´t seem plausible to me, that Ukrainian losses were higher on average before 28.2.23 than after it.

The US intelligence numbers for 28.2.23 are all we really have. That is not enough for even the roughest guestimate on real losses today.

Posted : August 7, 2023 19:18
Posts: 5

(this is a translation of my article )


Estimates of Russian and Ukrainian losses, end of August 2023

This war has now been going on for 18 months, and it's time to update the previous estimates I made 6 months ago , at the end of February. I will follow exactly the same methodology, first talking about material losses then human losses.


Material losses

As before, I rely on l">all losses confirmed by the now famous Oryx site . They continued their titanic work, and there are currently more than 16,000 pieces of equipment considered lost (destroyed, damaged or captured) by the belligerents: 11,862 for the Russians, 4,247 for the Ukrainians as of 08/27 /2023.

We are at a ratio of 2.8:1 in favor of the Ukrainians with regard to proven material losses, slightly lower than what it was 6 months ago (3.1:1), but which remains very favorable to Ukraine. But to estimate the real ratio, it is of course necessary to estimate the real losses (of which the documented losses are only a subtotal) of the two sides. As before, I rely on the opinion of military analysts (for example l">Colonel Michel Goya ) who think that it is necessary to multiply the figures of documented losses by a factor of between 1.3 and 2 to obtain the real losses. This gives a range of 15,400 to 23,700 pieces of heavy equipment lost by the Russians, and a range of 5,500 to 8,500 for the Ukrainians.

So there is a ratio of 1.8:1 to 4.3:1 in favor of the Ukrainians with respect to actual material losses. Colonel Goya estimated (in July 2022) that Russian losses were more documented than Ukrainian losses (and therefore that we would be at the lower end of the range), but I find it hard to believe that this is still the case currently, given the eagerness of the Russians to publish the slightest Lancet strike and to photograph the same Bradleys from 10 different angles to give the impression that the Ukrainian losses are immense. If there were so many undocumented Ukrainian losses, the Russians would remedy them as quickly as possible to fuel their cyber war against Western countries. On the contrary, the Ukrainians have been relatively discreet since the start of the counter-offensive to try to preserve (as best they can) C" target="_blank" rel="noopener">OPSEC.

Finally, I remind you that this ratio is also a first indication regarding human losses; in fact, for two armies equipped at approximately the same level, the human and material losses are more or less proportional, with some possible deviations.


Human losses

I will mainly focus on the estimates of the number of deaths, using the same methods as before (where these methods were explained).

First method: estimation based on the number of equipment lost

Following this method, and using our estimate of material losses, we obtain:
For the Russians: from 123,200 to 189,600 deaths
For the Ukrainians: from 44,000 to 68,000 deaths

NB: 4">Ragnar Gudmundsson arrives at an estimate of 136,400 using a similar method.

Second method: by correcting the figures given by the Ukrainians

The Ukrainian Ministry of Defense publishes every day a table of Russian losses (material and human): 260,820 Russians eliminated (as of 08/27/2023). As we can compare what is declared to what is confirmed by oryx (for material losses), this gives an estimate of the reliability of these figures. 4" target="_blank" rel="noopener">Oryx currently confirms around 35% of Russian losses declared by the Ukrainian Defense Ministry. 6 months ago, this percentage was 45%.

In my opinion, this drop is largely explained by the Ukrainian army specifically targeting Russian artillery (which has a lower confirmation rate than armor) rather than a drop in the reliability of Ukrainian figures. But as I cannot formally prove this, I will do the calculations using the 35% rate, even if this leads to a small underestimate of Russian losses.

The inverse of 35% is 2.9; in other words, the losses announced by the Ukrainians are, on average, 2.9 times higher than the Oryx figures. So, as the real material losses (equipment) are 1.3 to 2 times those observed, it is enough to multiply the figures given by the Ukrainian ministry by 1.3/2.9 and 2/2.9 to have the limits lower and higher than actual losses. Assuming that the same corrective factor can be applied to human losses, we obtain

For the Russians: from 118,600 to 182,500 deaths
For the Ukrainians: this method cannot be used.

Method 2 bis: we can also choose a larger interval by estimating the real losses between the percentage of confirmed losses (35%) and 100% of the total announced by the Ukrainians. Which gives Russian losses of between 91,300 and 260,800 deaths.

Third method: evaluation based on DNR losses

The so-called “Donetsk People's Republic” (DNR) published detailed tables of its losses until the beginning of December 2022, and gave the figure of almost 4,000 deaths. We can deduce an estimate of total Russian losses on that date (around 80,000 deaths). 

However, as the count stopped almost 9 months ago (halfway through the conflict), there is no new information since my previous assessment. We can at best make a linear extrapolation and then estimate the current Russian losses at around 160,000 deaths.

Fourth method: assessment based on the number of Russian officers killed

Another method used by Xavier Tytelman. Use the number of e">Russian officers confirmed dead in Ukraine (2700). Double this number (to consider all those whose deaths have not been confirmed). Then multiply by 30 (according to X. Tytelman even if he confuses "losses" and "deaths" in his analysis); we then obtain 162,000 Russian deaths .


Fifth method: rely on estimates made by intelligence services/media

Over the past 6 months, there have been fewer published estimates; however, some are very interesting and we will detail them here. 

First of all, honor to our host: on 08/16/2023, Q" target="_blank" rel="noopener">Colonel Goya estimated that the Ukrainians had “more than 60,000 deaths”. He did not give a precise figure for the Russians, only saying that they had "a little more" deaths than the Ukrainians. l" target="_blank" rel="noopener">The New York Times reported on 08/18/2023 that American officers give the following estimate:
For the Russians: 120,000 deaths
For Ukrainians: 70,000 deaths 0" target="_blank" rel="noopener">A study published in PNAS on 08/14/2023 estimated that, after one year of war (February 23, 2023), the Russians had 76,687 dead (95% confidence interval: 38,670 – 139,772) and Ukrainians 17,223 (6,219 – 39,105). These are significantly lower numbers than my own estimate as of this date, especially for Ukrainians. The study gives a loss ratio estimated between 2.9:1 and 4.9:1 in favor of the Ukrainians, which is approximately the same interval as that which I had established, for material losses (of 2:1 at 5:1). Regarding the number of deaths, if we make a linear extrapolation of the median values ​​calculated in this study, we obtain:
For the Russians: 115,000 deaths
For Ukrainians: 25,800 deaths
Finally, s" target="_blank" rel="noopener">the study published by Mediazona/BBC on 07/10/2023 attracted a lot of attention. It estimates that by the end of May 2023, there were approximately 47,000 Russian deaths (95% confidence interval: 40,000 – 55,000). The methodology seems very serious: they are based on a database listing inheritances in Russia, and look at the differential between men and women (in other words: they assess the excess mortality of men of fighting age). But the result is so far below all other estimates that it is doubtful. 
So certainly, we have to take the date into account (a linear extrapolation updates their estimate to around 56,400 Russian deaths at the end of August) and they don't count the DNR/LNR deaths. Does that suffice to explain the difference? As we saw above, the DNR had almost 4,000 deaths at the beginning of December 2022. Even if we multiply this figure by 2 (to take into account the LNR) and then again by 2 (to take into account the 9 months that have passed since the beginning of December 2022 ), we obtain approximately 16,000 DNR/LNR deaths at the current date, which would give an estimate of 72,400 Russian deaths at the end of August, a figure still much lower than other estimates. 
There is also the question of "missing". The authors of the Mediazona/BBC study tracked some of them (those of the 1st Guards Armored Army, in March 2022) and observed that, out of this sample, only half of the "missing" are at ultimately declared dead, sometimes after a very long delay. The very large number of Russian "missing" (the Ukrainians claim that they have tens of thousands of Russian bodies) could therefore also explain why this study (which is based on the "officially dead" Russians) underestimates the actual number of deaths.
Finally, there is possibly a strong social/geographic bias: in fact, we know that deceased Russian soldiers come proportionately more from the poorest regions of Russia (often in Asia). However, as the authors base themselves on the inheritance register (well, only a part but which they think is representative), it is possible that these poor regions are also under-represented in this base (the authors do not indicate that they sought to correct geographical biases), especially if the deceased soldiers left no inheritance due to their poverty. This bias could therefore also explain why the study in question underestimates the number of Russian deaths.




Although they only coincide imperfectly, these different estimates give an order of magnitude of the material and human losses: around 140,000 dead on the Russian side, a little less than half on the Ukrainian side.

In terms of likelihood, I would say that at the moment (end of February 2023):

Probable intervals (confidence index > 60%)
- the Russians probably have 120,000 to 160,000 deaths
- the Ukrainians probably have 55,000 to 75,000 deaths

Very likely intervals (confidence index > 95%)
- Russians most likely have 100,000 to 180,000 deaths
- Ukrainians most likely have 40,000 to 90,000 deaths

NB: we are only talking about military losses here. On the Ukrainian side, we must also add the very significant civilian losses.

Posted : August 30, 2023 22:47
Max Beckhaus reacted