A 30-car coal drag climbed the helix fine solo, but under a 2.2% grade on a 22″ curve the lead third of the train stayed on the rail while three or four cars near the middle got dragged inward, straight across the inside rail, and derailed. It only happened pulling, never pushing, and always on the curve — never on the tangent approach.
Pushing didn’t reproduce it, so this was a tension failure, not a buff (compression) one. The derailing cars were the lightest in the consist (empty hoppers at ~2.8 oz) and sat where the drawbar pull was still high but the train was deep into the curve. Swapping two light cars for weighted ones at that position stopped the derailment — a strong hint the problem was weight-versus-pull, not a track defect. Rolling each suspect car solo through the curve showed no binding, ruling out a gauge or coupler-height fault.
On a curve, the coupler between two cars does not pull straight along each car — it pulls along the chord. That puts a lateral (inward) component on every car equal to roughly the drawbar tension times the coupler’s angular offset. The inward pull is resisted only by the car’s own weight pressing the wheel flanges against the railhead. When the lateral force approaches the car’s weight, the inner wheels unload and the car gets pulled off the inside of the curve — classic stringlining.
The light hoppers lost that contest: high line tension (30 cars up a 2.2% grade) × the 22″ curve angle produced a lateral force comparable to their 2.8 oz. Prototype railroads see the identical failure — it is why they cap train length and tension on sharp curves and distribute power.
Three changes, in order of payoff:
- Weight the light cars to NMRA RP-20.1 (1 oz + 0.5 oz per inch of length). More weight directly raises the lateral force the car can resist before unloading.
- Cut the drawbar tension at the problem spot — add a mid-train helper so the forward cars aren’t dragging the whole tail through the curve on one coupler path.
- Ease the curve or the grade where you can; both feed the tension and the angle that produce the lateral force.
Stringlining is a sideways failure driven by coupler geometry, not just “too steep.” If light cars pull off the inside of a curve under load, suspect lateral force before you blame the track — and reach for weight and power distribution before you blame the loco.