Our earlier predictions against the makers’ datasheets

We took property predictions our methods had already made for commercial coolants and graded them against the values the makers publish. Most missed, by a lot. We publish the misses because they show how far our methods sit from real data.

Graded after the fact against the makers’ datasheets, most of our earlier property predictions missed their bars; one held.

Each row is one earlier prediction. The shaded band is its pass bar around the maker’s value (the centre line), on a scale that treats too high and too low alike; the dot is our estimate. One dot lands inside its band.

A dotted underline marks a number read straight from a published file when this page was built.

In this post
  1. What it shows
  2. Why it matters
  3. Who should care
  4. The limits, in the lab’s own words
  5. Prior art
  6. Evidence
  7. Related results

What it shows

We graded 9 of our earlier predictions for three commercial fluids against the makers’ published datasheets. A surface tension passed if it fell within plus or minus 15% of the published value, and a dielectric constant within plus or minus 25%.

Only 1 passed. It was a dielectric constant: we estimated 2 for the perfluorinated liquid, and its maker publishes 1.75.

The misses are the larger story. Every surface-tension prediction came out high. Our estimate of the hydrofluoro-olefin fluid’s dielectric constant was 2, and its maker publishes 32.

Why it matters

A computed coolant is only worth testing if the computer gets real fluids roughly right. This grading tells us, and a reader, where ours does not: surface tension from our molecular simulations runs high on fluorinated fluids, and our quick permittivity estimator can be wrong by an order of magnitude.

Who should care

  • Anyone qualifying a computed coolant: these are the kinds of error a computed property can carry before it is measured.
  • A partner lab planning bench work: the misses say which properties to measure first.
  • An investor: the grades show where the lab’s methods stand against outside data, misses included.

The post on how we check a simulation sets these grades beside the solver check that held.

The limits, in the lab’s own words

Quoted word for word from the published file named under each quote. Read the result with them.

Post-hoc grades: every prediction predates the bar, and the bars were chosen after the published values were read. They measure how far this pipeline sits from manufacturer data; they are not pre-registered tests.

From Our earlier predictions graded against manufacturers’ datasheets.

Prior art

The published work this result uses or is checked against. Each link was opened and checked when this page was written.

Evidence

Each number with a dotted underline was read from one of these published files, field by field, when this page was built.

All results

Ask about a result, or check one yourself

Every result on this site links the published file it was read from. Licensing, acquisition and partnership enquiries go to one address, and a person reads it.

Write to us Read the results

How the numbers on this page are checked
Each underlined number is read from a published file when the site is built, and the build stops if the page and the file disagree. A number would be marked if
  • its file were still loading in your browser,
  • the question behind it had not been looked into,
  • it had been looked for and not found,
  • its file held no value for it,
  • its file were missing or had changed,
  • two files disagreed about what it describes,
  • its sample were too small for the claim,
  • two files gave it different values,
  • its file could not be published,
  • the file it is read from had last changed more than thirty days before this build,
  • the question did not apply to this page or
  • the program that makes it had stopped with an error; no underlined number on this page is marked.
A number in plain text, without the underline, is not checked this way and carries no mark; the date each file last changed is in the list of published files.