Mon 2025-Sep-22

Ukraine War: 1,100,000 Russian Dead

Tagged: MathInTheNews / Politics / Sadness / Statistics

Russian dead in Ukraine: now more than 1,100,000.

Huītzilōpōchtli & Moloch Continue Their Reign

Here on this Crummy Little Blog That Nobody Reads (CLBTNR), we’ve for some time been having… opinions about Ukraine. ^[1] ^[2] ^[3] ^[4] ^[5] ^[6] ^[7] ^[8] ^[9] ^[10] ^[11] ^[12] ^[13] ^[14]

It is, very regrettably, time for an update.

As we’ve been saying for a couple years now, we’re tracking the Russian casualty figures provided by the Ukrainian Ministry of Defence:

We’ve been using the Ukrainian Ministry of Defence’s figures for Russian casualties. Of course, one could object that they are a biased source. But when we looked at other sources, they were kind of in the middle of the pack. Some sources were wildly inflated, while others had extremely hard to believe small numbers. (The latter were institutions that insisted on a very high standard of evidence, like geo-location, identities, photographs, and so on. This leads to a very severe under-count, so people usually take their estimates and apply a multiplier derived… somehow.) The Ukraine MoD says they count, every day. That might even be true. But they seem reasonably credible compared to others, and they are the ones on the spot. So we’ll take their figures.

Sadly, it appears the Russians have now wasted the lives of another 100,000 of their young men for a total of 1.1 million so far in their invasion of Ukraine:

As we’ve previously remarked, this is continuing to blow a hole in Russian demographics of 18-44 year old males, i.e., those more or less qualified for military service. By our reckoning:

\[144\ \mbox{million Russians} \times 0.5\ \mbox{males} \times 1/3\ \mbox{military age} = 23.7\ \mbox{million}\]

So 1.1 million dead is about $100\% \times 1.1 \mbox{million} / 23.7 \mbox{million} = 4.6\%$ of the military-age male population.

Add to that reports that approximately 900,000 Russian men of military age have left Russia to avoid the draft.
Then add that Russian casualties from COVID-19 seem to be around 400,000 ^[15]. Applying that to 1/2 the population to get males and then 1/3 to get those of military age means a loss of about an additional 67k men of military age unavailable due to being dead from COVID-19.

Together the Russian emigration and COVID-19 deaths give us about an extra million, for a total loss of about 2.1 million, or about 8.8% of the Russian men of military service age.

That’s… staggering. As we’ve quoted Bertrand Russell on this subject before:

“The mark of a civilized man is the capacity to read a column of numbers and weep.” – often attributed to Bertrand Russell, who said things like this if not exactly this, but would in any case almost certainly agree with the sentiment

When we look at the deaths in Russia’s invasion of Ukraine, the deaths in Israel’s invasion of Gaza, the wholesale dismantling of public health in the US in favor of superstition, summary military executions of people in boats in the Caribbean, and masked men kidnapping people from the streets of the US with no due process at all… it is arguable that we can no longer claim to be civilized. We are, perhaps, just clever barbarians?

We have let right-wing populism turn us into vicious, superstitious, and venomously racist savages. And somehow we have done this on a planet-wide basis.

Comparison Against Our Casualty Rate Model

Back in 2023, I started building a regression model tracking number of Russian casualties versus the number of days from the start of the invasion. By 2023-May-17, I’d updated the model a couple of times, and had 114 data points on the first 116 days of the war. At that point, I stopped updating in detail, because it was a lot of work to extract the data from the images posted on X/Twitter by the Ukrainian Ministry of Defence.

The model as of that date is shown here (click to embiggen). As you can see visually, the linear fit is pretty good. The casualty figures snake around a bit, sometimes above and sometimes below the line. But the fit is always pretty good, with nice tight confidence limits.

The statistics of the fit bear this out, showing an excellent $F$-statistic $p$-value for the overall fit and good $t$-statistic $p$-values for each coefficient. The adjusted $R^2 \sim 99.43\%$ is surprisingly excellent, i.e., the model was highly predictive.

Residuals:
    Min      1Q  Median      3Q     Max 
-3077.9 -1720.8    72.2  1169.4  3299.7 

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept) 1.225e+05  3.318e+02   369.3   <2e-16 ***
x           6.996e+02  4.984e+00   140.4   <2e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 1767 on 112 degrees of freedom
Multiple R-squared:  0.9943,	Adjusted R-squared:  0.9943 
F-statistic: 1.97e+04 on 1 and 112 DF,  p-value: < 2.2e-16

The model back then showed a surprisingly steady casualty rate of about 700 ± 10 Russians/day.

That was at what turns out to have been just the beginning of this madness. One thing leads to another, and here we are having moved on from day 116 to approximately day 1000.

The plot here (click to embiggen) shows:

the original data (blue points, lower left),
the line fitted to the original data (black line, gray prediction confidence interval), and
the new points, approximately every 100,000 deaths, plotted in red points.

The striking feature here is that when I started plotting the new (but sparse) points at about day 450, the death rate increased relative to the original model. If you squint diagonally back along the red points, there’s no argument any more for a nonlinear increase, but there is a very good argument for a linear increase, i.e., a new rate of deaths/day. (Yes, there are statistical tests of linearity/nonlinearity; no, I’m not going to bother.) We’ve gone from about 700 Russian deaths per day to a little over 1000 Russian deaths per day. That’s not a little thing: it’s a 43% increase in the Russian death rate!

If we were to re-fit a model, clearly it should have a breakpoint at 450 days, with separate linear fits before & after. (A scrupulously principled and honest approach would also fit where the breakpoint is, not impose it at 450 days.)

We don’t know if the Ukrainians have just gotten so much better at war (arguably true, especially with drones) or if the Russians have resorted to more and more human wave tactics (also arguably true). Either way, it’s more human sacrifice.

The Weekend Conclusion

Not all of this is the fault of the US or Europe, but we have certainly prolonged the conflict by being just… flighty about Ukrainian support.

On this matter as with so regrettably many others, all of us have much of which to be ashamed.

The treatment for shame is tikkun olam: repentance and repair of the damage we’ve all done to the world. But we certainly don’t seem interested in walking down that path, do we?

(Ceterum censeo, Trump incarcerandam esse.)