I’m a little surprised that we haven’t covered this in detail before, because it’s so crammed full of Jalop-bait: jumping and flipping cars, AMCs, James Bond, frustrated backwoods law enforcement, and lots of secret math and early computers. The jump is the famous corkscrew from The Man with the Golden Gun.

This famous 1974 jump of an AMC Hornet X over a twisted, broken bridge in Thailand is notable for a number of reasons: first, this was the only time a Bond car (even a temporary one) has been an AMC, which is a glorious event to celebrate regardless.


Then, there’s the wild complexity of the jump itself, which involved the car making a 270 degree twist like a football in flight and landing back on the wheels, and was all accomplished in one take. And, perhaps most importantly, this was the first automotive movie (or any) stunt that was extensively computer-modeled.

Now, the basic stunt — a car making a jump off an angled ramp, spinning on its long axis, and then landing on another ramp — was actually performed before the Bond version:

The US racing driver Jay Milligan conceived the stunt and even performed it in 1972 at the Houston Astrodome in an AMC Javelin, christening the stunt ‘The Astro Spiral Jump”. Milligan contacted the Bond producers with the stunt who promptly protected it preventing it appearing in any other preceding films.


I’m not exactly sure why AMCs have been so associated with this particular stunt — it was an ad placement deal for the movie, and it’s not clear if Milligan just preferred them and encouraged the movie studio to seek out a deal with AMC. Maybe AMCs just have a knack for twisting through the air. Anyway, the basic premise was set, and the studio liked it so much, they even patented the jump so no one would steal their thunder.

Now, setting up a complex jump like this in the controlled environment of a stunt show and doing it on a movie shooting location in Thailand are not exactly the same thing. To duplicate the stunt in the new environment, it would need to be totally re-engineered, and the traditional way of figuring these kinds of things out — making some educated guesses and then trying and crashing a whole bunch — wasn’t the answer here. Or, as an academic paper on the subject bloodlessly states:

“... the destructive aspects of unsuccessful tests would preclude the application of a ‘trial and error’ experimental development.”


Yes, yes, they didn’t want to murder all their stunt drivers. That’s prudent.

So, with wanton automotive and human murder out of the question, what other options did they have? They had one very important one, courtesy of a man and his pioneering work in computer simulations and visualizations, work that would one day lead directly to the physics models in the amazing driving simulators we use today just for fun — racing games like Forza.


The man was Raymond R. McHenry, who was working with Calspan Corp. Calspan started as the Cornell Aeronautical Laborotory, which was instrumenting aircraft to learn about the underlying physics of flight. They company soon began researching much-lower-altitude craft like cars, and started compiling definitive physics models about how cars worked, moved, crashed, and all that exciting stuff.

To calculate all this complex data and reliably test the data without constantly smashing up cars, a computer/mathematical model was developed. Called the HVOSM, for Highway Vehicle Object Simulation Model (‘Highway Vehicle Object’ may be the geekiest way to refer to a car that I’ve ever heard), the HVOSM

... consists of up to 15 degrees of freedom; 6 for the sprung-mass, and up to 9 for the unsprung-masses. The mathematical model is based on fundamental laws of physics (i.e., Newtonian dynamics of rigid bodies) combined with empirical relationships derived from experimental test data (i.e., tire and suspension characteristics, load deflection properties of the vehicle structure). The balance of forces occurring within and applied to components of the system are defined in the form of a set of differential equations which constitute the mathematical model of the system. The HVOSM includes the general three-dimensional motions resulting from vehicle control inputs, traversals of terrain irregularities and collisions with certain types of roadside obstacles.


That’s a pretty comprehensive model, especially when you consider the state of computing in the early 1970s. You may also note that this also sounds very serious and academic, and you’re probably wondering why, exactly, such a serious-seeming organization would have anything to do with a crazy movie stunt. Well, happily, we know why:

Subsequent to completion of development and validation of the HVOSM simulation in 1970, the thrill show ideas were given somewhat more serious consideration. Such an application would constitute both a challenging dynamics problem, similar in nature to a particularly violent single-vehicle accident, and an attention-getting demonstration of capabilities. It also had the appeal of a “fun” project to relieve a steady diet of crash protection studies.

See? Man cannot live by crash protection studies alone. “Fun” is fun, after all.


So, using his remarkable HVOSM software and a healthy desire for “fun” in his diet, McHenry set off to plan and design the movie version of the Astro Spiral Jump. The HVOSM program was used to get remarkably accurate predictions of how the stunt would go, and gave such important information as the angle of the launch and return ramps, the ideal speed of the car (right around 40 MPH — partially due to the restriction of the usable space of the jump), the roll velocity (about 230 degrees/second), and more data. This was the first time computers had been employed to compute any sort of movie stunt like this.

Perhaps even more revolutionary considering the future directions this technology would take, McHenry was one of the first to employ computer-generated visualizations to help evaluate the test runs. Sure, they could have figured it all out via numbers, but this was for a movie stunt, and the visual look of the stunt was absolutely key.

The jump was rendered in very crude but surprisingly effective wireframe vector 3D graphics, either outputted to a plotter or drawn on a CRT. These visualizations were critical in proving the success of the jump, because it looked pretty badass, even in simple wireframes:


Just remember that this is from 1974. Most people then only interacted with a computer to move a little white bar of light to play Pong, and even then it wasn’t until the next year they were likely to be able to do that at home. This was wildly advanced stuff.

Eventually, the time came to turn the models into reality. A suitable river of the right width was located in rural Thailand, and a set of ramps were built to the pre-determined specifications, and then disguised to look like a collapsed bridge with scrap wood. The Hornet was modified with a roll cage, given a central steering position, and an “auxilliary contact point” was added to the rear axle. Into this modified Hornet X they crammed stunt driver Loren “Bumps” Willard, who has an amazing nickname.

Eventually, it was time to shoot the scene. Here’s how it went, from the handheld camera of a crewperson:

That’s right — they nailed it on the first try. Eight cameras caught the action (and then later slowed it down a bit, because it all happens pretty fast, and you really want to see how that car twists, after all) and the Hornet landed exactly as the computer models predicted. It was an incredible, unqualified success.


Well, maybe not unqualified. Here’s how it showed up in the actual movie:

Wait— was that a fucking slide whistle sound effect in there? Holy crap, it was. What kind of sick Foley joke is that? A slide whistle? Is this a Scooby-Doo cartoon? Jeeezis.


Well, if you can ignore the slide whistle — and maybe Roger Moore’s painful Southern accent impression, and maybe that inane stereotyped bumbling fat redneck cop character — then I’m sure you can appreciate what is likely the most amazing movie car-jumping stunt ever.

Also, with games like Forza still relying on data and models from Calspan to this day, and with McHenry’s use of computer-generated visuals, I think you could make a pretty reasonable argument that this stunt was one of the ancestors of modern physics-based driving simulations and games.

A slide whistle?

Contact the author at jason@jalopnik.com.