Thread by @GabrielEpsteinX

The +972/Guardian articles alleging that the IDF’s internal data indicates it has killed fewer than 9,000 militants in Gaza misconstrues the available data, improperly combines IDF and Hamas-run Gaza Health Ministry figures, and fails to ask key questions. Very long thread 🧵

BLUF: Is it possible the IDF is overstating militants killed in Gaza? Yes. But these articles do not present a convincing case about the existence or extent of an undercount, and instead ignore data limitations and common sense to make a flashy claim.

First, militaries do often exaggerate enemy losses and conceal their own casualties, as both sides in the Russia-Ukraine war have. Hamas has also exaggerated operations (see their narrative of the major attack Wednesday) and does not reveal losses.

In Israel, the combination of low casualties and active media (plus social media channels that evade the military censor) make concealing losses near-impossible. The same conditions do not apply to militant losses in Gaza, about which the IDF has only ever given broad figures.

However, the +972/Guardian articles majorly overstate the AMAN database they accessed and ignore the major caveats provided to them by intelligence sources, leading to a faulty analysis. These caveats (to which more can be added) include:

The combatant figure cited lists only named terrorists who match someone in the AMAN database, meaning a sort of double-confirmation: the IDF both knows the name of the person they have killed, and they are confirmed via internal Hamas documentation.

In many cases, though, the IDF may not know the names of militants killed. Many do not carry ID cards with them, are killed in inaccessible areas like tunnels or from the air in places the IDF is not operating, and may not be listed in the Health Ministry’s death toll.

From this, the figures +972/Guardian cite are not a ceiling, but a floor of named Hamas/PIJ fighters killed alone for whom there was specific intelligence and double-confirmation via internal files. This high bar is not going to catch all Hamas/PIJ deaths.

The database also lists only Hamas/PIJ fighters. But many armed groups have lost fighters, including: PFLP, DFLP, PFLP-GC, Mojahedin Brigades, Nasser Salah al-Din Brigades, Fatah al-Intifada, Jaysh al-Umma, plus gangs, affiliated clans, etc. These deaths are not included.

Moreover, the list is based on Hamas/PIJ documents seized from Gaza which are not comprehensive. Israeli intelligence services are adept, but a list of members of organizations dedicated to secrecy will always be a moving target that misses people.

See this document containing many excerpts from different documents here to understand how the AMAN list may have been assembled. It is a patchwork relying on fragmented pre-war documentation, not an up-to-date comprehensive directory. govextra.gov.il/media/d21mw2f3…

One major group this list will miss are new recruits. +972/Guardian cite U.S. intelligence estimates that some 15,000 fighters were recruited to Hamas/PIJ since the start of the war — nearly 1/3 of AMAN’s entire list. If these combatants were killed, they would not be counted.

Moreover, the U.S. intelligence +972/Guardian cite for this itself estimated 10-15,000 Hamas fighters killed as of January 2025, with a roughly equivalent number recruited. It is cherry-picking not to include that detail. reuters.com/world/middle-e…

Non-Qassam and non-Quds Brigades members of Hamas/PIJ are also excluded from the list. +972/Guardian seems to argue that they should not be counted as combatants. But the distinction between “military” and “political” wings of terrorist groups is hardly a real one.

It would be strange to argue that Islamic State or Al-Qaeda members were not terrorists simply because of their role in the organization. Qassam or not, Hamas members are part of a U.S., EU, and Israel-designated terrorist group (all of whom designated both “wings”).

One must also consider combatants without a clear affiliation, many of whom participated in the October 7 attack. This group may not be large, but it is significant and would be missed by such a list.

There is a large gray area in which people may be assessed as combatants in the moment based on rules of engagement or circumstantial evidence, but are not in the end. This reality can be exploited by commanders, but to what extent this has happened is less clear.

A further question the authors should have asked, but did not: Is the AMAN database fully up to date, or is there a backlog in assessing militant deaths and searching for matches? Further, are gaps between fully verified and estimated militant deaths common in other wars?

The articles mention most of these caveats, but then fail to reckon with them in their later analysis. Moreover, the articles make the mistake of trying to combine IDF and Gaza Health Ministry figures. I have previously written about why this is unwise:

https://x.com/GabrielEpsteinX/status/1925619258198720898

The articles argue that the civilian-combatant ratio may be even worse than it appears due to a major undercount of the overall death toll. Yet they assume that all 8,900 confirmed combatant deaths they saw from the AMAN database are in the MOH list. This is highly unlikely.

At a basic level, combining IDF and MOH data is a mistake. The IDF will catch many militant deaths the MOH does not, but does not track civilian deaths comprehensively. Since the MOH doesn't differentiate combatants and civilians, there is no way to know how much overlap exists.

Moreover, both the scope and scale of an undercount in the death toll are unclear. As I’ve written before, there are practical and ideological reasons to believe that many militants have been missed in the count, and that the remaining uncounted deaths skew towards militants.

The overall size of the death toll is also very much in question. Just 1,309 bodies have been recovered from rubble or remote areas in the last 7 months, the only missing-persons estimate sits at 9,500, and the Health Ministry has multiple collection systems to record deaths.

While some studies have projected undercounts of up to 40%, and there are wild claims of a much higher death toll, these don’t fit the observed ground conditions. The more likely scenario is a moderate undercount which skews more towards militants than the counted death toll.

So, it is clear that the AMAN database/list cited by +972/Guardian is incomplete and uses a very high bar, meaning it does not represent the full assessment of militant deaths. There are multiple other problems with the articles.

First, the IDF has made it very clear publicly that it does not regard MOH data as reliable — whether they pay attention to what the MOH says doesn’t determine whether they think it is accurate or not. But even if they did, the assertion that they trust the daily tolls is wrong.

As I’ve written many times before, the contemporaneous daily MOH tolls do not match the later revised figures. In the early months, the death toll was much higher than the daily reports indicated, and in later months it was often lower. Safe to assume, daily variance was high.

This also reflects the fact that as new family reports were processed from Apr-Dec 2024, they were cancelled both from “media reports” and hospital reports figures. After that, they were added directly to the toll. See my May report for more: washingtoninstitute.org/policy-analysi…

Even if one fully trusts the MOH and their methodologies, the contemporaneous daily tolls would still not be accurate. Beyond that, the lists have undergone substantial revision and have to be understood as an incomplete, evolving product. public.flourish.studio/visualisation/…

Finally, relying on secondhand information about Hamas and PIJ’s losses, which they have never publicly disclosed beyond a handful of leadership figures, is poor practice. These are not trustworthy groups, particularly on this subject, and this is not corroborating evidence.

The articles are correct that Israel has often provided varying estimates of combatants killed, and that assertions made by politicians have regularly been unrealistic. I have previously called for greater IDF transparency on this point: aoav.org.uk/2024/flawed-cr…

This list indicates a serious effort to track combatant deaths, and shows an incomplete list with a high evidentiary bar that excludes many groups, misses new recruits, and cannot be compared cleanly to MOH lists. +972/Guardian ignore this and therefore produce a faulty analysis.

See the +972 article here: 972mag.com/israeli-intell…
Guardian here: theguardian.com/world/ng-inter…
Hamas-run GMO missing persons estimate (July 18): t.me/mediagovps/3748

To read more about the knowns and unknowns (and knowable and unknowables) of the Gaza death toll, see my May report: washingtoninstitute.org/policy-analysi…
As well as these threads: x.com/GabrielEpstein…
x.com/GabrielEpstein…

All data and additional sourcing is available on request.