The Merseyside projectTurning the Rainhill rules into tests
Eight conditions from 1829, four engines, and a test suite that finds problems in the rules as well as the machines.
The conditions, as Smiles prints them, include: the engine "must effectually consume its own smoke"; if it weighs six tons it must draw twenty tons "(including the tender and water-tank) at ten miles an hour"; it must not weigh more than six tons, but a lighter engine "would be preferred on its drawing a proportionate load"; four wheels are allowed only at four and a half tons; and "a mercurial gauge must be affixed". The trial itself was twenty trips on a level piece of line "about two miles in length", averaging at least ten miles an hour. The learner writes one small Python function for each rule that can be checked from numbers, and records what Smiles says each engine did.
Our tests run on the figures in Smiles's account, 26 September 2026| Engine | All 20 trips | Weight rule | Speed of 10 mph | Load, read literally |
|---|
| Rocket | Pass | Pass (4.5 tons) | Pass | Fail: 13 tons against 15 |
| Sanspareil | Fail: stopped at trip 8 | Fail: over the four-wheel limit | Pass (about 14 mph) | Cannot decide |
| Novelty | Fail: pipe burst | Pass (about 3 tons) | Cannot decide | Fail: about 7 tons against 10.2 |
| Perseverance | Fail: withdrawn | Cannot decide | Fail: 5 or 6 mph | Cannot decide |
Three things go wrong on the way, and each one is the real lesson. First, the very first run crashed: Smiles gives no times for Novelty's trial, and a test that expected a number fell over. A well-built test has three answers, not two: pass, fail, or cannot decide from the evidence. Second, the arithmetic of the course does not add up. Twenty trips on two miles of line is forty miles, not the seventy Smiles states. The account of Rocket's run settles it: each trip went "backward and forward", so a trip was a round journey of three and a half miles.
Third, and most interesting, our test fails Rocket, the engine that won. Reading "a proportionate load" as straight proportion, a 4.5-ton engine should draw 15 tons, and Rocket drew "about thirteen tons" of wagons. But the judges said Rocket met every condition. When a test fails a result that everyone accepted, the first suspect is the test. Condition 2 counts the tender and water tank inside the twenty tons, and our test left them out; the book does not say how the judges set the load. Speed has the same trap. Rocket's seventy miles took 3 hours 51 minutes of running, 18.18 miles an hour; add the 57 minutes spent raising steam and it becomes 14.58, close to the fifteen Smiles reports. Both pass, but a specification has to say which clock counts.
Ages 8 to 11
Write three rules for a toy car race, test each car against them on paper, and argue about the rule that could be read two ways.
Ages 11 to 15
Store the four engines in Python dictionaries, write a function per rule, and print a pass or fail table like the one above.
Ages 15 and up
Add a third answer for missing data, write the tests with a proper testing framework, and document every assumption the rules forced you to make.
Whose figures these are
The conditions and every figure for the engines come from Smiles's book, which was written long after the event and is one account among several. The tests, their readings of the rules and the pass or fail results are ours, made for teaching. They are not a new verdict on the trials.
