7 Comments

Good job but what I find difficult to estimate given the present situation is the tempo of casualties and equipment losses since the summer. It surely is horrific for both sides and what it is objectively is difficult to assess. I assume the units fighting around Bakhmut are certainly at the very least weakened. How far it is impossible to assess. Thanks for your work Henry. always appreciated.

Expand full comment

Yeah it's quite obvious that Russian losses are in the 12-15k range, DPR/LPR add an additional 6-8k, then volunteer/Wagner/etc add an additional 5-10k at most.

Ukraine's losses on the other hand are likely minimum 60-80k for KIA and another 60-100k for irrecoverably lost (limbs gone etc).

Expand full comment

Nice work. Would an assumption of what % of actual losses are captured by Oryx have any effect? For example, assume he captures 50%, then assume he captures 75%. I suspect your relationships are not sensitive to this assumption. BTW, if the Oryx % captured was 55%, the Ukraine estimates of Russian tank losses would be right on the money.

Expand full comment

It depends on the timeframe. When others and myself looked at the data specific for 1st Guards Tank Army around Sumy in the initial phase of the war, about 70% of the losses mentioned in the captured reported could be located in the Oryx data. Ukrainian personnel are far more likely to take the time to take pictures of destroyed Russian equipment if they have secured the area. So you see more in the Oryx data during the Sumy, Kyiv, and Kharkiv portions of the war. Then it decreases, both due to Ukraine being on the defensive and the Oryx team not being caught up with all the data. Also, only a percentage of what is in the overall Oryx dataset can be attributed to specific unit. With 1st GTA we were fortunate in that almost all the vehicles in the area could be matched to units. Ultimately, I haven't finalized exactly how I'm going to account for it, but I would say 70% at the maximum for February through the early spring perhaps, then decreasing as we get closer to the present.

Expand full comment

Interesting. It seems to me that lends a bit more credibility to AFU loss estimates. Oryx's cumulative Rus tank losses today are about 52.3% of AFU cumulative estimate. As of late June 2022 that same comparison percentage was 52.5%. So since late June it's been remarkably consistent at about 52%. As you said, it was quite a bit higher during certain earlier periods. Wish I could share a graphic. So maybe the Oryx capture rate is lower than 70% now, but it surely can't be lower than 55%. And if it is lower than 70%, then the AFU estimates seem pretty darn credible. I know that rolling 5 day averages of the AFU loss data suggests very interesting and credible stories. Such as large UAV loss spikes exactly coinciding with Gerasimov's Ilium field visit and Surovikin taking command, possibly reflecting UAV battlefield assessments.

Expand full comment

Great piece that I will follow with interest.

Russia are using the mobile crematoria heavily and the order of another 21 was confirmed 6 days ago. These are used to distort the death rate of the military losses as well as the civilians massacred. There are civilian reports of these crematoriums used 24/7 and the ashes buried in the new building going up in the occupied territories.

I am sure the families will be told that the soldiers were captured or that they deserted, as we know Russia is creative in lying and obvuscating.

Sadly the real picture will take decades to come out, if it ever does.

Expand full comment

Great stuff

"1st GTA lost many of its vehicles through breakdown rather than fighting, resulting in more missing and captured than might otherwise be the case"

I was thinking that revuewing Oryx's dataset it should be easy (if time consuming) to categorise vehicles as intact or destroyed. That would then allow for a bit more granularity in losses. There are of course still problems - destroyed in action or abandoned and then destroyed - and for the IFV the question of whether infantry were loaded or not.

A quick first iteration would be to ctaegorise just on the Oryx terminology of destroyed vs captured. Possibly looking up a table of crew size for the equipment.

A proxy for deaths could be russian mobile phone numbers that connected to the network and then did not reconnect. The hypothesis being non-reconnection = death, capture, severe injury. Again a margin of error and phone records would need to be available

Expand full comment