Torsional stiffness matters more than weight
Why twist governs how a car behaves, and how it is measured.
Torsional stiffness, a frame's resistance to being twisted along its length, matters more than weight because a chassis that twists under load lets the frame itself absorb the movement a suspension was designed to control, which makes every carefully chosen spring and damper setting behave unpredictably regardless of how little the whole car weighs.
Carroll Smith's Engineer to Win devotes most of its length to how things fail, which is an unusual emphasis for a book about going quickly. The chapter on chassis stiffness in particular reads less like a design guide and more like a list of ways a light, clever frame quietly stops doing its job, and it was the first place I saw torsional stiffness described as something a suspension depends on rather than something a chassis merely happens to have.
The frame as an uncalibrated extra spring
A car's suspension is calibrated to control how each wheel moves in response to the road, using springs and dampers whose rates were chosen for a reason. That calibration only means anything if the frame connecting the four suspension mounting points holds its shape while the car is cornering, braking and running over uneven ground, all of which load the chassis unevenly and try to twist it along its length. If the frame is stiff enough, essentially all of that twisting load gets carried by the suspension itself, exactly as intended, and the wheel movements the driver feels match the movements the suspension was designed to produce. If the frame is not stiff enough, some of that load instead twists the chassis itself, and the frame quietly starts behaving like an extra, uncalibrated spring wired in alongside the real suspension, one with a rate nobody chose and a behaviour that changes with every new load case. Weight reduction has an obvious, easily measured payoff. Torsional stiffness has a payoff that is invisible until it is missing, which is exactly why it is the one a design can get wrong without any single number on a drawing looking incorrect.
A kitchen table on an uneven floor
A four-legged kitchen table on an uneven floor makes the same problem visible without any suspension at all. If the tabletop is genuinely rigid, one leg simply lifts a small, definite gap off the floor, and that gap can be closed cleanly with a shim under the short leg, exactly the kind of local, controlled adjustment a suspension is built to make. If the tabletop itself is a little flexible, though, the table does not wobble in that clean, fixable way. Instead the whole top subtly twists to let all four legs touch the floor, and the table feels mushy and unpredictable under an elbow leaning on one corner, because now two different things are absorbing the unevenness, the shim that was never fitted and the tabletop's own unwanted give, and nobody can tell from feel alone how much of the wobble either one is responsible for. A rigid tabletop turns floor unevenness into a problem the legs can solve. A flexible one turns it into a problem that the whole structure is quietly, and unhelpfully, solving on its own terms.
Twist measured against suspension movement
Torsional stiffness is usually described as the angle a chassis twists through for a given twisting load applied across its two ends, and the practical target for a competition chassis is to make that angle small enough that the frame's own twist stays a small fraction of the movement the softest suspension setting produces at the same corner. Once a frame reaches that point, further stiffening buys only a diminishing improvement in how predictably the suspension behaves, while every extra kilogram spent getting there is a kilogram the car now has to accelerate, brake and turn everywhere else on the circuit. The trade being made is therefore a specific, checkable ratio between how much the frame twists and how much the suspension is allowed to move, which is far easier to reason about than stiffness against weight in the abstract.
Weighing frames by how they behave
Because the payoff from torsional stiffness is invisible until it is missing, it tends to be judged wrongly by comparing raw chassis weight between designs rather than by comparing how each design actually behaves under load. A frame that is a little heavier but clearly stiffer will usually out-handle a lighter frame that quietly gives away some of its suspension's authority to uncontrolled twist, because the driver feels consistency and predictability rather than pure lightness. Torsional stiffness is therefore measured directly, with a chassis mounted at three corners and a known twisting load applied at the fourth, rather than inferred from wall thickness or tube diameter alone, since two frames built from identical-looking tube can end up with very different stiffness depending entirely on how those tubes are triangulated together.
A driver adjusting to a new chassis notices this before any engineer does, because a car whose frame is soaking up part of every corner feels different from lap to lap even when nothing on the setup sheet has changed. Chasing that inconsistency with spring rates or damper clicks usually fails, because no combination of suspension settings can fully compensate for a frame that is itself behaving like an unmeasured, unpredictable spring. The only real fix is to design the stiffness in from the start, which is why torsional stiffness is treated as a target to be hit early in a chassis design rather than a property to be checked and tolerated once the frame is already welded together.