In the September 26 Formula 1 roundup, there was a short quote, easy to scroll past between the crash description and the starting grid. Jacques Villeneuve, the 1997 champion, was analyzing Andrea Kimi Antonelli's crash in the first corner at Baku and said: "Walls aren't round, they're sharp — it shouldn't be like that in Formula 1. It would be easy to build them round, like guardrails, but nobody bothers."
The phrase sounds like a retiree grumbling about a stain on a freshly painted wall. What caught my attention was different: the championship leader wrecked his car not in the narrow "castle" section, where the walls stand seven and a half meters apart, but in the first corner — where, according to Villeneuve himself, it's "quite spacious." He wasn't punished by the tightness. He was punished by the geometry of the tire and the block the wall is assembled from.
Then I dug into where they design such walls. And there, a fact emerged that flips the usual picture: a track barrier has two tasks, and they contradict each other by definition.
Intuition says a good barrier is strong. Complete nonsense, and you can see it right in the regulatory document. FIA Standard 3501-2017 "Safety Barriers" describes exactly how a barrier is tested, and this description doesn't say a word about how strong it is.
The test looks like this: a trolley weighing 780 ± 8 kg is equipped with a deformable conical "nose" according to the 2014 and later Formula 1 technical regulations, accelerated, and crashed into the barrier at a set speed. There are six speeds in the matrix: 85, 95, 115, 135, 150, and 165 km/h. The barrier is evaluated not by whether it held up, but by what reached the "driver" — or rather, the volume the document calls the driver's free volume zone.
The acceptance criterion is acceleration. The peak acceleration of the trolley in the direction of impact must not exceed 50 G; deceleration must not stay above 48 G for longer than 3 milliseconds; lateral g-forces on the Y and Z axes are limited to ±30 G. That is, the wall's softness is what gets certified. Softer means it holds worse? Nothing of the sort, because the second task is formulated separately and is not solved by material.
This is where old road engineering steps in. In the European standard EN 1317, introduced in 1998, there are three groups of characteristics, and the first of them is not strength. It's the containment level (who are we stopping at all), the safety level (by biomechanical indices — what's the probability of injury), and the working width: how much space the system will take up at the moment of impact. The practical rule for designers is simple: where space is tight, they take a system with a smaller working width, but with the same safety. Strength turns out to be a consequence of the space budget.
If stiffness isn't the main parameter, then what? The answer is best seen in two crashes, half a year apart and in different hemispheres.
Sochi, October 2015. Carlos Sainz lost his car in the fast twelfth corner and went into the barrier of the thirteenth head-on. The barriers there were the "correct" ones — deformable TecPro blocks, a French development from the early 2000s, homologated by the FIA in 2006. The result was not what the advertising promised: the car slid under the barriers, and the blocks rose around the driver's head. It only stopped after piercing right through the rigid metal Armco guardrail standing behind.
Sebastian Vettel then said a phrase that describes the problem more accurately than any report: "It was shocking to see him so deep in the barrier and that the barrier covered him." And he added the main thing — he doesn't know what the matter is: "I don't know if the problem was how the TecPro was put up or how it broke." Max Verstappen was blunter: the barrier shouldn't rise, because then it could hit the head.
Daytona, February 2015. Kyle Busch crashed on the 112th of 120 laps, hitting the inner wall nose-first, which had no barrier on it at all — just concrete. An open fracture of the right lower leg, a fracture of the left foot, six nights in hospitals, missing the Daytona 500 and another six races.
Track president Joie Chitwood came out to the press that same evening and said what officials rarely say: "Today, we did not live up to our responsibility. We should have had a SAFER barrier here; we didn't. We will fix it. Right now." Overnight, tire packs were laid on an 850-foot section of the wall so that the next day's race could happen at all.
The most unpleasant thing in this story is the premonition, not the crash. Busch said that a few minutes before the start, pulling out of the pit lane, he looked at the concrete and thought: if someone crashes into it, it's going to hurt. And then he found that wall personally. Kevin Harvick had complained about the exact same wall a year earlier — it took a second impact for something to change.
The two cases look opposite: there, a barrier that was too soft; here, none at all. But the failure is the same. What's dangerous is neither hardness nor softness, but an unaccounted point of contact: an end, a joint, a seam, a place where the car catches on something instead of sliding along.
Understanding of this didn't come from motorsport, and this story has an author with a name.
In 1955, at the 24 Hours of Le Mans, Pierre Levegh's car took off after contact with Lance Macklin's Austin-Healey and flew into the crowd. Levegh himself and 84 spectators died — the largest disaster in motorsport history. American John Fitch was racing on the same team, and his memory retained not so much the day itself as the thought that came later. The idea was prompted by the war: during his service, he protected a tent from machine-gun fire with sandbags and noticed how quickly sand stops a bullet. Thus the Fitch Highway Barrier appeared — barrels filled with sand.
It works differently than one might think. The barrels don't withstand the impact, they smear it out: there is less sand in the front row, more in each subsequent one; upon impact, the barrels collapse, the sand scatters, and the car loses speed smoothly instead of stopping instantly. If the car had stopped immediately, the driver would have continued moving — into his own cockpit and chest cavity.
The difference between "withstand" and "crush" is the entire engineering essence of the barrier, and it has survived to our days in three designs. TecPro — modular blocks of high-density polyethylene with a foam core and adjustable ballast, each one meter twenty in height, one meter in width, connected by locking couplers and anchors so that the joints allow limited mutual displacement. Not a single slab, but a flexible chain where elements work through each other.
SAFER (Steel And Foam Energy Reduction, "soft wall") was born in 1998 at the Midwest Roadside Safety Facility of the University of Nebraska and was first installed at the Indianapolis Motor Speedway in May 2002. By 2005, almost all IndyCar and NASCAR ovals were covered with it. Strictly speaking, this is paying back a debt: motorsport took the idea of deformation from road guardrails and returned it to the highway.
Now we can return to Villeneuve and check his diagnosis against the regulatory documents.
First: he is right in describing the problem. Barriers on street circuits, including Baku, are indeed assembled from interlocking segments. And segments have exactly the drawback he talks about: you can't build a perfect arc out of straight blocks, somewhere an edge sticks out. He even names the right model — "like guardrails" on ordinary roads.
Second: he is only right about the design, not the practice. A fully FIA-certified barrier must pass tests on a straight section — standard 3501-2017 explicitly states that the test section is installed straight, while curves and transitions are excluded from the test matrix. However, the geometry that is considered an acceptable assembly tolerance in the lab becomes that very sharp edge on the track.
Furthermore, the FIA mandates embedding markings directly into the barrier during manufacturing — not gluing or engraving them later — as well as a hologram sold only by the federation itself, which only it has the right to remove. The regulations fairly closely ensure that the barrier is exactly the barrier that was tested. Monitors — every single block.
The most beautiful part of the whole story lies exactly where the regulations end — at the end of the barrier.
In road engineering, the joint between two types of guardrails and the end of a guardrail are considered such dangerous places that separate parts are dedicated to them in EN 1317: terminals and transitions, and the requirements for a transition specifically state that its containment level cannot be lower than the worse of the two being connected. The logic is rigid: the system works as long as it is predictable, and falls apart exactly where it ends.
The illustration is a story that reached an appeals court. American company Trinity Industries made the ET-Plus end terminal and in 2005 reduced the guide channels from 5 inches to 4 (from roughly 13 centimeters to 10), without notifying the Federal Highway Administration. The claim was not that the barrier held worse, but that when hit from the end, the system could pierce the car.
The jury agreed in 2014: 175 million dollars, tripled under the False Claims Act to 525 million, plus 138.4 million in civil penalties — 8,250 dollars for each of the 16,771 certifications. Total: 663.4 million. The whistleblower, by the way, was a competitor of Trinity's in the market and claimed 199 million — thirty percent of the amount.
And in September 2017, the Fifth Circuit Court of Appeals overturned everything unanimously. The argument was about money, not engineering: the highway administration continued to reimburse the installation of ET-Plus, fully knowing about the claims. Meaning, the company's statements could not have influenced its payment decisions. Judge Higginbotham formulated it as an aphorism: "When the government at the relevant levels time and again concludes that it was not defrauded, it is not forgiving an established fraud — it is concluding that there was no fraud at all." Trinity's stock rose nine percent that evening.
What holds me in this story isn't the legal reversal. The moral is different: the disaster lived in a four-inch channel, that is, in the edge at the end of the guardrail — in a detail that doesn't make it into flashy reports or press presentations. Exactly where the "sharp wall" of the first corner in Baku is.
The difference between a car sliding along a barrier and a car catching on it is the difference between two very different days. Sliding means energy goes into friction over hundreds of meters. Catching means it goes all at once, and then no softness of the block saves it: those very modules that were supposed to absorb the impact rise up and turn into an additional risk.
At the same time, the problem has a property that hinders engineers the most. A barrier can be designed to be as good as you like, but a track is not a lab, and it ceases to be predictable in two places: in curves, which are officially excluded from the test matrix, and at the ends, where the system terminates. Both of these zones live as background noise: they are remembered as long as nothing has happened.
So Villeneuve is about half right, and the second half is the interesting part. Round walls are indeed not hard to build — the question is that a seam remains between the barrier segments, and a seam is always a place where predictability breaks.
In Baku on Saturday, George Russell won, beating Max Verstappen by 0.196 seconds — the smallest margin in over twenty years, his first win since the Austrian round. Antonelli, who wrecked his car in qualifying, was fighting back from sixteenth place, and on lap 31 Alex Albon crashed his Williams into the TecPro barriers in turn five — again without injury, precisely because the barrier was there and worked as intended.
And now the detail I like more than all the others. The technical specification for these barriers is formulated like this: "let it through slower," not "don't let it through." A wall built with the expectation that it will break is the only kind of wall you can trust. This isn't a metaphor about flexibility of thinking, but a literal norm from a document about a safety system, and I still haven't decided what's more charming about this fact: the cynicism of the engineers or their honesty.