Text taken from the wartears.org website, via Russian military journalist Boris Rozhin:
You can see a graphic on the conclusions of wartears.org in the above like. Alas, it is in Russian language
Attention!
The calculation of the mathematical model is temporarily suspended because we have received a large amount of accurate data on the dead and are working on model changes in accordance with the changed incoming accurate data.
Also in this second version of the model, estimates of the number of deaths are overestimated by approximately 25-30%. We have significantly corrected and improved the model by publishing the third version of . Which, see above, was not the last.
Since May 25 of this year, the WarTears.org project has been helping people find their relatives who served in the VFU and stopped communicating for one reason or another. During this time, the Project has collected a database of more than 159,000 records, received more than 8,500 requests for the search for people and helped to find information on more than 2,000 military personnel (as of April 2023, and these numbers are growing daily).
Note : hereinafter we use the abbreviation VFU - Armed Formations of Ukraine. We do this because it is often not clear which organizational hierarchy a person might belong to. These can be the Armed Forces of Ukraine (Armed Forces of Ukraine), NGU (National Guard of Ukraine), TRO (territorial defense), SVR (Foreign Intelligence Service of Ukraine). In addition, there are nationalist formations, for example, the Azov regiment (which is banned in the Russian Federation and significantly, but not completely overlaps with military unit A3057 ), the Kraken battalion , the Transcarpathian Sich; volunteer battalions (dobrobat "Solnyshko"), organizations of mercenaries and just foreigners, far from always included in official structures.
We are looking for everyone. That is why hereinafter we use the general term VFU.
We decided to try to estimate the losses and the total number of HFU based on the data obtained by statistical methods. What is interesting here is that it is quite difficult to do this directly because of the many reasons that can significantly distort the statistics. For example:
However, some of the records, especially the records of those killed, prisoners and claims from relatives, we manually check for uniqueness, so duplicates may affect the estimate of the total number, but not the estimate of losses.
METHODOLOGY
We want to note right away that we did not estimate the number of wounded evacuated to the territories controlled by Kyiv, as well as the number of deserters due to the lack of sufficient data. Therefore, the mathematical model and its estimates below were made only for the dead Ukrainians and Ukrainians who were taken prisoner. It is these people - dead or captured - that make up the general population of the model in the statistical and probabilistic sense.
We separately estimated the number of people in one or another narrow and unambiguously defined category.
We make the assumption that the distribution of people searched through the WarTears.org project and the distribution of data obtained by the project from open sources are independent. For the most part, we get information about the prisoners and the dead from lists that have surfaced in open sources and from Telegram channels that post any such information that falls into their field of vision. A small part of the information comes from the media, but it mostly concerns either officers or nationalists.
In connection with the latter, we had to count the officers separately - about this and other possible adjustments below, in the "Numerical result" section.
For each state (dead, prisoners), the score was calculated by the formula: T = R / (A / Q)
Where:
T- the total number of VFU soldiers in this state,
R- the number of records about this state in our database,
Q- the total number of search requests from relatives,
A- the number of applications by which the soldier was found and is in this state.
In other words, we a priori assume that the probability that we have an entry in the database about someone is the same as the probability that it was found in our database, provided that we searched.
In addition, we estimated the total number of AFU - active, dead and captured - according to the same method, assuming that everyone who is in our base is or was in one way or another in the AFU. Based on this and the previous two estimates, we estimated the active population of HFU by simple subtraction. Since we know that our database contains duplicate entries about active members of the VFU, the resulting estimate is an upper estimate. As you work, duplicates will be removed from the database, and the score will decrease.
NUMERICAL RESULT
We have found that this model does not apply to certain categories. Chief among these categories are officers. There are very few requests to search for officers in our database, but out of more than 3,500 records of the dead, more than 1,200 are officers. This showed a fundamentally different distribution of probabilities in the category, and therefore we assumed that we know about all or almost all cases of officer deaths. In the model, the number of such records was taken into account directly, without any coefficients.
We also compared our estimates with the performance of Sergei Shoigu. As of September 21, 2022, our model gave an estimate of 6 thousand more (67 thousand versus 61 thousand) than the official data of the Russian Ministry of Defense. We hypothesize that this may be evidence of categories other than officers, for whom the number of search applications may be abnormally low. Such categories may be the core of the Ukrainian army, which has undergone British training, as well as representatives of nationalist groups such as "Azov", "Kraken" or "Transcarpathian Sich" - all of them are given increased attention in the media. We believe that these people in terms of statistics are somewhere between officers and ordinary mobilized. However, in the absence of strict criteria, we did not introduce another correction. If the reader considers it necessary, then we can reduce our estimates of the dead by about 10%.
On November 30, 2022, according to the statement of the head of the European Commission von der Leyen (very quickly edited, but the original version remained on the Web ), the losses of the Ukrainian military amounted to more than 100,000 people. Evaluation of our mat. models for this day : 102,350 dead.
Another coincidence with the statements of officials occurred this year: according to the Turkish press, as of January 14, 2023, MOSSAD estimated the number of dead AFU fighters at 157,000 people. According to our mat. models for the same day the death toll was 161,769 people.
Below we provide daily updated graphs for estimating the number of dead and captured soldiers of the VFU. To see the numerical value of the ratings on a specific day, hover over the reference point for that day.
Attention! The graphs show the change in our assessment, which includes both a change in the real state of things and a change in our understanding of what is happening . Therefore, one should not be surprised at any downward changes in the estimate of the number of dead, or sudden jumps - most likely on this day we found and processed another list of names that appeared on the network.
Further, based on the totality of records, we ventured to make estimates of the total number of HFUs.
We believe that this is an upper estimate that can be adjusted downward for a number of reasons, for example:
there is a noticeable number of duplicate entries in our database. In the current version, however, we did not take into account this factor, which, again, can significantly reduce the estimate of the total number;
from general considerations, search requests will be much more likely for those who serve in the front-line units than for those who, conditionally, protect Lvov.
CONCLUSION
The real losses of the VFU, in addition to these numbers, include the wounded, deserters and actually missing. We assume that the total irretrievable losses of VFU (both dead and wounded, unfit for military service) may be twice as high as our estimate. The numerical composition, perhaps, is exaggerated.
It is also necessary to remember that this is a delayed smoothed estimate . It takes into account the fact that relatives applied to our project. This can happen both in two weeks and in five months, as was the case with the attack on the location of 137th Separate Mechanized Brigade in Nikolaev on March 18, 2022. The jumps in our estimates, for example, on October 5 or 24, are directly related to the fact that we stumbled upon some a fairly large list of the dead and made it at once.
We did not make an assessment of the missing and other categories. In the case of the missing, we know for sure many cases where the Kiev regime informed relatives that people were missing, while this person died during a rocket attack on a location deep in the rear.
For the time being, we consider the assessment of other categories to be premature, because the number of known cases in them is not enough to obtain statistically significant values.
We also understand that the estimate of the total number of HFU is likely to be overestimated. We plan to clean up the database from duplicate entries and, accordingly, adjust the estimate.