Backtest · Engines · 1972–2026

Fifty-four years of data. The rotary never had one good year.

From the first measurable point in 1972 to today, the Wankel rotary never once out-improved the piston engine it was meant to replace: 5–6%/yr against a rising 10–13%. The method calls losers as clearly as winners.

Computed from patents public at each date · no hindsight
01 · The consensus vs the rate

Real money filed into the rotary for decades. The rate was flat the entire time.

Combustion (piston)Rotary (Wankel)
The consensus · what the industry filed
New patent families per publication year
The rate · the signal
Estimated improvement rate per year, computed from patents public at each date
GetFocus platform export · 1972–2026 · what the platform would have said at the time, computed from patents public at each date
02 · The timeline · against the rate

The industry's biggest engine bet, next to the number that said no.

The data series opens in 1972, as GM's programme was at full speed. The verdict predates every cheque that followed.

1957

NSU runs the first Wankel.

Smooth, compact, few moving parts, on paper the piston engine's obvious successor. The industry begins to believe.

PRE-SERIES
Data opens 1972
1961–70

The licensing wave.

Dozens of manufacturers buy Wankel licences: Curtiss-Wright, Mercedes-Benz, Toyota and General Motors among them. The strongest consensus any alternative engine has ever drawn.

PRE-SERIES
Data opens 1972
1967

The Ro80 wins Car of the Year, then eats its engines.

NSU's rotary flagship suffers mass early-life rotor-seal failures. Warranty claims bleed the company.

PRE-SERIES
Data opens 1972
1969

NSU disappears.

Financially drained, NSU is folded into Volkswagen's Audi. The rotary claims its first company before the data series even opens.

PRE-SERIES
Data opens 1972
1970

GM pays $50 million anyway.

In November 1970 GM licenses the Wankel, roughly $400M in today's money, and president Ed Cole projects production in three years. The first platform reading lands two years later.

THE RATE READ (1972)
PISTON 6.6% · ROTARY 4.8%
1973–74

The oil crisis calls the bluff.

Rotary fuel consumption becomes untenable; GM shelves the programme as Cole retires; Mercedes cancels the C111 line. The rate had never moved.

THE RATE READ (1974)
PISTON 8.1% · ROTARY 5.4%
1978–2012

Mazda carries it alone.

Engineering heroics, including a Le Mans win in 1991, against a rate that never budges. RX-8 production ends in 2012, closing 45 years of rotary manufacture.

THE RATE READ (2012)
PISTON 13.0% · ROTARY 5.5%
2023

The coda.

The rotary returns in Mazda's MX-30 R-EV, not driving the wheels, but as a small onboard generator. A supporting role was all the rate ever supported.

THE RATE READ (2026)
PISTON 12.8% · ROTARY 5.8%
03 · The reasonable arguments at the time

Smart people chose the loser for defensible reasons.

The rotary attracted more manufacturers than any alternative engine in history. The arguments were real. The rate never was.

“Fewer moving parts must win.”

Mechanical elegance reads like destiny. But elegance is not a learning curve: apex seals, chamber shape, fuel economy and emissions never entered a fast improvement loop, and the flat rate proved it, year after year.

“Everyone has licensed it.”

When Mercedes, Toyota, GM and Curtiss-Wright all pay for the same technology, dissent feels reckless. Consensus was the filing count. The rate disagreed with all of them at once.

“Racing proves it works.”

Rotaries won races, including Le Mans 1991. Race wins prove peak performance on one day, not the compounding of cost and performance that decides markets.

04 · The bill

What a measurable dead end cost anyway.

$50M
GM's Wankel licence, November 1970, roughly $400M today
1
Carmaker destroyed: NSU, absorbed by Volkswagen in 1969
45 yrs
Of Mazda rotary production, 1967–2012, against a flat rate
$50 million in 1970 dollars: GM's licence alone, paid two years before the first data point confirmed the dead end; the programme was cancelled by 1974.
NSU's independence: Ro80 warranty failures drained the Wankel's inventor into a 1969 takeover.
Dozens of licensees' R&D programmes: across two decades, almost none of which produced a production engine.
The one that persisted: Mazda built rotaries until 2012; the engine survives today only as a range-extender generator.
05 · Beneath the rate

A twenty-year cycle time is what a dead end looks like.

We forecast the improvement rate from two citation metrics, cycle time and knowledge flow. Both were visible the whole time.

Cycle time
Years between invention generations, shorter is faster
Knowledge flow
How much later work builds on each invention, higher is bigger leaps

The rotary's cycle time sat near twenty years for half a century, each invention generation took a generation. Its knowledge flow never crossed 0.3: later work simply did not build on it. Compare the piston engine, unglamorous and compounding, its cycle time tightening decade after decade. The verdict was never close.

Provenance. Every patent carries a filing date, so the record can be replayed: every improvement rate on this page uses only the patents that were public at the time. No hindsight enters the series. Chart data: GetFocus platform exports, August 2026. The historical facts below come from the public record.

Sources
  1. Contemporary reporting on GM's Wankel licence, November 1970 ($50M; Ed Cole's three-year production projection).
  2. Automotive histories of the NSU Ro80 (Car of the Year 1968; rotor-seal warranty failures) and NSU's 1969 absorption into Volkswagen/Audi.
  3. Mazda: rotary production history, 1967–2012; Le Mans victory 1991; MX-30 R-EV rotary range extender, 2023.
  4. GetFocus platform exports: piston vs rotary engine datasets, 1972–2026.
Seen early

The method calls losers too. Ask it about the bet you're least sure of.

Bring us one live technology decision. Within 48 hours you get a free preliminary Decision Brief: a first-pass radar, improvement rates and suggested triggers.

Free · 48 hours · preliminary scoping — validated radars come with a subscription