Data
Not Indiscriminate: What Hamas's Own Numbers Show
The Gaza Health Ministry's named casualty lists carry an age and sex distribution that does not match the population they came from.
"Genocide." "Indiscriminate." "Disproportionate." We've been hearing these words for a long time now. Wikipedia describes the war in Gaza as "a genocide, agreed upon by most scholars." Sounds damning, doesn't it?
To back up these accusations, Amnesty International, South Africa, and other supposedly "impartial" parties say they have rigorous, substantial evidence. What is that evidence? Several unrelated witnesses, all of whom claim there were no military targets in the area that was struck. How would they know? Did all of these very unbiased—and certainly not hostile to Israel—witnesses happen to live in that exact apartment and know for a fact that there were no military targets inside? Very convincing.
So here is my own evidence—drawn from Hamas's own numbers, the very numbers used to accuse us.
I downloaded every casualty list the Gaza Health Ministry has published since January 2024: ten files, which turn out to be nine distinct lists, because two of the files are the same export saved twice. I know that because I hashed them rather than trusting the filenames. The most recent list contains 60,199 named records, each with a national ID number, a date of birth, an age and a sex. Not one record had to be thrown out for a missing or unreadable age or sex.
Every number in this essay was computed from those files by a Jupyter notebook you can read, which also writes the data file the charts below draw from. The published charts and the analysis cannot drift apart, because one generates the other.
If bombs fell on a population without regard for who was beneath them, the dead would be a random sample of the living—and a random sample resembles the population it came from. Gaza's dead do not resemble Gaza.
- Children under 15
- 40% → 25%share of Gaza's population, then share of the named dead
- Males aged 15–59
- 28% → 50%a quarter of the living, half of the dead
- Male deaths per female death
- 1.12 → 3.28under-fives, then the 35–39 band; the living population sits at 1.03
- Lists checked
- 9Jan 2024 to Jul 2025 — every one skewed the same way
The dead are not a cross-section of the living. Each bar pair is one age band. Sand is that band's share of Gaza's population; navy is its share of the named dead. If the killing were a random draw, the navy bars would sit on top of the sand ones.
The two halves of that chart tell different stories, which is the whole point. The female side roughly tracks the population: sand and navy are close in almost every band. The male side does not. Below fifteen the navy bars collapse well under the sand ones; from fifteen through forty-four they overshoot them badly.
Children under fifteen are 40 percent of Gaza's population and 25 percent of the list. Men aged 25 to 54 are 17 percent of the population and 33 percent of the list. Widen that to males aged 15 to 59 and they are 28 percent of the living and half of the dead. Had those 60,199 deaths been drawn at random from Gaza, roughly 24,300 of them would have been children. 14,824 are—six-tenths of what chance alone would have produced.
Expected against actual. How many deaths each group would hold if the dead were drawn at random from the living, against how many it actually holds.
A formal test is almost beside the point at this sample size—with 60,199 records, any real deviation registers as significant—but for the record, a chi-square goodness-of-fit test of the death distribution against the population distribution returns a statistic of 19,308 on 27 degrees of freedom. The interesting part is not that the test rejects. It is which cells do the rejecting: men aged 20 to 44, every one of them recorded at roughly twice their population share, and children under five, recorded at about half theirs.
The sex ratio is the cleanest test
Consider the sex ratio. A boy and a girl of the same age in Gaza live in the same building and shelter in the same room. A collapsing structure does not sort them. So under indiscriminate killing, the ratio of male to female deaths should be flat across every age band, sitting wherever the population's own sex ratio sits—which in Gaza is 1.03 overall and never leaves the narrow range 0.88 to 1.06 in any band. This test barely needs the population denominator at all, which is what makes it the strongest thing here.
In the Ministry's own data the ratio is 1.12 among the under-fives and 1.12 again from five to nine—flat, and effectively at the population ratio, which is exactly what indiscriminate death looks like. Then it breaks upward: 1.44 at ten to fourteen, 2.43 at fifteen to nineteen, 3.00 in the twenties, peaking at 3.28 in the 35–39 band. Then it falls back—2.17 by the early fifties, 1.58 above sixty-five.
The shape is the finding. The line is flat precisely where combatants do not exist, rises across the ages where they do, and comes back down again on the far side. Add to this the fact that Hamas actively recruits children into its ranks—a war crime in its own right, by the way—and the break upward at ten to fourteen stops looking like noise.
The ratio climbs with age, then falls back. Male deaths per female death in each age band. The dashed line is the same ratio in the living population—where a flat, indiscriminate pattern would sit.
One list could be a fluke. Nine cannot.
Everything above comes from a single release. A single release could be an artefact of a single bad export. So I ran the same arithmetic on all nine lists, published nineteen months apart from each other, and the skew is in every one of them—and it gets stronger, without a single reversal, from the first list to the last.
| List | Usable records | % female | % under 15 | % males 15–59 | Male : female |
|---|---|---|---|---|---|
| 5 Jan 2024 | 13,143 | 43.2 | 31.7 | 36.5 | 1.32 |
| 29 Mar 2024 | 20,042 | 39.7 | 26.3 | 42.3 | 1.52 |
| 30 Apr 2024 | 24,240 | 37.4 | 24.9 | 44.7 | 1.68 |
| 30 Jun 2024 | 28,173 | 37.0 | 27.3 | 43.1 | 1.71 |
| 31 Aug 2024 | 34,306 | 36.0 | 26.8 | 44.1 | 1.78 |
| 23 Mar 2025 | 50,020 | 33.0 | 25.3 | 48.5 | 2.03 |
| 15 Jun 2025 | 55,202 | 32.7 | 25.1 | 48.9 | 2.06 |
| 15 Jul 2025 | 58,380 | 32.1 | 24.7 | 49.8 | 2.12 |
| 31 Jul 2025 | 60,199 | 31.9 | 24.6 | 50.0 | 2.13 |
The obvious objection is that these lists are cumulative—each contains the one before it—so a rising cumulative share is a weak claim. Fair. The sharper question is what the newly added names look like in each release, and because every record carries a national ID number, that can be answered exactly. Eight increments, published months apart, each one an independent batch of names:
| Names added between | Records | % under 15 | % males 15–59 | Male : female |
|---|---|---|---|---|
| Jan – Mar 2024 | 7,039 | 18.8 | 52.1 | 2.03 |
| Mar – Apr 2024 | 4,377 | 18.1 | 56.2 | 2.79 |
| Apr – Jun 2024 | 7,523 | 24.9 | 45.4 | 1.97 |
| Jun – Aug 2024 | 6,658 | 24.9 | 48.2 | 2.14 |
| Aug 2024 – Mar 2025 | 18,666 | 22.4 | 54.9 | 2.55 |
| Mar – Jun 2025 | 5,244 | 23.6 | 53.0 | 2.31 |
| Jun – Jul 2025 | 3,195 | 17.3 | 64.5 | 3.76 |
| 15 – 31 Jul 2025 | 1,826 | 21.9 | 56.5 | 2.80 |
Not one of the eight batches resembles the population it was drawn from. The male-to-female ratio never once drops below 1.97, against a living population at 1.03, and in the batch added over June and July 2025 it reaches 3.76. The skew is not an artefact of one release. It is in every release the Ministry has ever published.
While I was matching those IDs I found something else worth saying out loud, because it cuts against the reliability of the lists generally rather than for my argument: names do not only get added between releases. They also get removed. Nearly 4,000 names present on the 30 April 2024 list were gone from the 30 June 2024 list, and about 3,000 more disappeared before March 2025. The Ministry revises. That is a reason to treat any single list as provisional—and it is why nothing here ever merges lists together.
What this proves, and what it doesn't
I'm not claiming every military-aged male is a militant. The excess is a group-level pattern and says nothing about any particular person; men in that age range also die as civilians in ways women do not, because they are likelier to be outdoors, working, moving between shelters or digging in rubble. Nor does any of this measure the total number of dead. It measures the composition of a list.
But the simple fact remains: if the bombing were truly systemic and indiscriminate, the data would show a very specific pattern. Put gender and age bins on the x-axis and death count on the y-axis, and the chart should closely mirror the same chart with population count on the y-axis instead. Yet, by Hamas's own numbers, these two charts are night and day from each other—in every list, in every increment, for nineteen months.
That makes the case, beyond any reasonable doubt, that Israel's actions could not have been indiscriminate. And if they were not indiscriminate, it follows that there could not have been any intent to destroy the people of Gaza, in whole or in part.