Rocket testing, the Washington Post, and the funniest fact check of all time
Please fix this, Mr. Bezos
Last October, Amazon and Blue Origin founder and Washington Post owner Jeff Bezos published this essay explaining his decision to not have the Post endorse a Presidential candidate. He noted that the Post and the wider media is widely perceived as politically biased, but seemed to think this was merely a perception problem and that the Post is fundamentally credible. He contrasted them with “unresearched podcasts” and “unverified news sources”, as though the Post is a reliable (and “verified”?) news source that powers its reporting with research that podcasters and new media don’t or cannot do.
I’m confident that the Post is extremely unreliable — much less reliable than a conservative outlet like the Daily Wire or the indy Free Press — and that its stark leftist bias yields many straightforwardly false claims, along with severe biases in selecting, framing, and writing news articles. (I’d bet thousands of dollars on it being less reliable than those other outlets, but it would take funding to do the rigorous research.)
There’s no better example than the Post’s “fact checker” Glenn Kessler. Since fact checking should be the cleanest and most disciplined of its journalistic operations, the fact that the Post pays for and publishes artifacts like the example below is an indictment of its overall commitment to and ability to produce journalism. It’s confusing that Mr. Bezos hasn’t noticed the frequent false claims in his paper, or the frequent false fact checks. Had he read this one, I’m sure he would’ve been stunned and horrified. Here we go…
In 2017, Kessler decided to fact check Trump’s claim about America’s missile defense system, the Ground-Based Interceptor (GBI): “We have missiles that can knock out a missile in the air 97 percent of the time, and if you send two of them, it’s going to get knocked down.” (Link)
The context was heightened tensions with North Korea.
Kessler also found Congressional testimony where a general gave a “high-90s” estimate for the system’s intercept probability.
Background: The GBI is designed to intercept incoming nuclear warheads in space, before they re-enter the atmosphere. The US has deployed at least 44 of them in Alaska and California, and they’re meant to defend against smaller nuclear powers like North Korea and accidental/rogue launches. (Russia and China have too many missiles and warheads to intercept.)
Kessler said: “Several experts said the high-90s claim appears to be based on faulty math.”
There is no math – not anything public he would know about. His math:
He looked at the program’s 18-year test launch history and… divided the number of successful interceptor launch tests by the total number of tests.
He thought this gave him the intercept probability of the currently deployed system.
He also thought this is what the Air Force did – their “faulty math”.
The Air Force didn’t divide successes by total tests to get its estimates. Those tests weren’t commensurable – they weren’t the same missile from test to test, and the tests themselves were different in design and goals. Missile and rocket testing is iterative – they learn from each test, especially failures, and make changes to the missile and software as appropriate for the next test. This testing spanned 18 years – the vehicle they first launched in 1999 had little in common with the 2017 version, and used a surrogate booster. A 2002 test failed because the kill vehicle didn’t separate, and in 2005 the arms holding the missile in the silo didn’t retract, so the missile didn’t launch — those resolved failures have nothing to do with the intercept probability of the system deployed a decade later. The USAF developed at least three different publicly known versions of the GBI. You can’t divide anything by anything here to get the intercept probability of deployed system. You might as well divide the missile’s height by its mass times π. There’s nothing you can do with a calculator here.
The intercept probability of those missiles is determined by a classified methodology that draws from technical test data and various engineering domains, as well as intelligence assessments of adversary ICBMs, including their expected trajectories, maneuverability, and decoy characteristics. All of that is classified. The interceptors are one piece of a multifaceted system including large ground-based radars, launch detection and tracking satellites, guidance uplinks, etc. Moreover, all those factors are dynamic, so intercept probability can change given a change in any of the components or adversary systems. It’s not possible to fact check a President’s claims here, and stunningly reckless to do so in the middle of a showdown with a volatile nuclear power.
Incredibly, it gets worse. Kessler’s mindless division gave him a 56% estimate (10 successes out of 18 tests).
56% is not 97%, but he thought he could get to 97% with extra steps, so he stipulated four interceptors launched against one incoming nuke. (Remember, 56% is a completely meaningless number here.)
The equation given the independent probabilities is 1 − (1 − 0.56)⁴. That doesn’t work for Kessler because it yields 96.3%, not 97%. So then he concluded:
“MDA appears to be rounding up to 60 percent.”
He means that he thinks that the US Air Force Missile Defense Agency took the 56% successful test rate from his division and rounded up. He genuinely believes they did his division thing, and that they would round up to the nearest multiple of 10% in computing an intercept probability.
With that assumption of insane, certainly criminal rounding, he lands on 97%. 1 − (1 − 0.6)⁴ = 97.4%
This is amazing. He thinks this is where Trump’s number came from, and he proceeds to pick it apart with his leftist NGO “experts”, who, incredibly, also thought this arithmetic was what the USAF actually did.
This artifact belongs in a museum. Just the odds that Kessler could land on 97% with his abuse of a calculator are very low – it’s an incredible coincidence and a candidate for the funniest fact check of all time.
Kessler gave Trump “Four Pinocchios” – the max – and this fact check is in the Post’s repository of thousands of purported false or misleading claims by Trump (and probably in Wikipedia’s wildly false article on false claims by Trump). That repository is full of false fact checks. (Mr. Bezos, when I say false, I mean straighforwardly false. Many others are misleading, and some are invalid, but many are simply false.)
This fact check was invalid — an arbitrary, painfully stupid artifact — and also false in saying that Trump made a false claim (and that he lied, using the unprofessional Pinocchios rating system that calls people liars without evidence). Kessler had no basis for it — he didn’t falsify the claim, and it’s probably not possible to falsify it short of a whistleblower investigation uncovering fraud in the tests or estimates.
Root causes
I see four root causes here:
Kessler doesn’t know anything about rocket/missile testing, which led him to make an embarrassing mistake. (It’s common for leftist “fact checkers” and “misinformation researchers” to adjudicate on topics they know nothing about, even deigning to fact check peer-reviewed scientific journal articles.)
Kessler is a partisan leftist and is extremely motivated to claim that Trump lied about something. (We can show how he lets leftist politicians get away with stunning lies — actual lies in that we can be confident they know what they’re saying is false.)
Kessler often only cites other leftists as “experts”, many of whom are not experts. This is common with leftist “fact checkers”. I call the method LQL: Leftists Quoting Leftists. His experts here were stunningly incompetent. (There’s a lot to say about the authoritarian, anti-reason epistemology of the left, and their new dogma around “experts” who seem to mostly make false claims, but that’s for another day.)
No one at the Post noticed, or noticed his many other false fact checks. There’s no active intelligence at the Post of the sort that can see basic, severe problems like this. His fact check of the Biden-Trump debate last year is another example where it’s obvious that no one is attentively reading his work and checking his sources — if they are, they’re not able to reason well or grasp basic descriptive statistics.
Mr. Bezos, you’re obviously familiar with rocket testing. You know how insane this fact check was. The Post isn’t operating at a standard of research and fact checking that would compare favorably to Joe Rogan, much less a more focused, expert podcast or writer. Moreover, Rogan tends to correct his false claims when he learns they were false. The Post does not. I emailed Kessler in July, 2022 about this one — no response, no correction. I’ve tagged him at least four times on X on it over the last couple of years — no response, no correction. This is incredible. And again, there are many, many other false fact checks, false claims in news articles1, and lots of interesting forms of invalidity and deception2.
Joe Duarte grew up in small copper mining towns in Southern Arizona, earned his PhD in social psychology, and focuses on political bias in media and academic research. You can contact him at gravity at protonmail.com.
Afterword
With a team and an LLM, we could cleanly and unassailably assess the Post’s repository, and similar ones at NYT and elsewhere, using a rigorous epistemic framework and solving the domain knowledge problems. I’m confident that most of the fact checks unfavorable to Trump are themselves false or invalid, and that Trump’s actual false claim count is much lower than widely assumed. More importantly, there’s a large corpus of false fact checks and claims in news articles that go well beyond Trump — this is systemic problem with America’s information environment, it was created by the left, and it needs fixing.
I’m writing a project proposal to systematically assess the reliability of media outlets, including self-styled “fact checking” groups, as well as various forms of bias. No one has done this — all current efforts that might sound like this are scams like NewsGuard. No one has actually rolled up their sleeves and systematically, validly assessed news headlines, articles, and shows for false claims, and any leftist group or academics that attempted to do so would end up ignoring most false claims by leftists, while mirroring Kessler’s record in false assessments of falsity by conservatives, Republicans, etc. For example, see the fraudulent and false (separate issues) “misinformation” research by Rand, et al. here.
So, Mr. Bezos if you want to have a huge impact on media reliability and credibility in the United States, supporting my project or something like it would be the best way I can think of. It’s long overdue — a country of America’s size, wealth, and role should have several redundant, rigorous, thorough assessments of media reliability, and we have nothing.
Yesterday’s headline “House votes to ban transgender students from girls’ sports” is false (link). The only people banned from girls sports (really a loss of federal funding) are boys. Trans-identified girls, e.g. who “identify as” boys, are free to play in girls’ sports, since they’re girls. The bill has no awareness of “trans”, a proprietary leftist construct. So the headline is plainly false, as is the article. It’s still up.
Read the Post’s pre-election article reporting that most voters were closer to Harris than Trump on “the issues” — I’m sure you understand why it’s false and invalid.