Some websites answer a question with a paragraph. Stone Age Skills answered one by rubbing dozens of species of wood together hundreds of times.
Read the hand-drill friction-fire study
The page documents an experiment with the hand-drill method of friction fire. The author gathered spindle material from 52 native and 22 non-native species of trees, shrubs, forbs and lianas, plus one grass, and tested them against four kinds of hearthboard material.
That produced 370 possible wood combinations.
Each attempt was rated by how difficult it was to produce an ember. The author also measured relative wood density, compared green, dead and nascent growth, tested different hearthboard densities, and tried producing embers without cutting the usual notch into the socket.
No metal tools were used in the trials.
A personal webpage behaving like a field study
The useful part is not simply the conclusion that some woods work better than others.
The page preserves the messy process behind the conclusion.
Fifteen spindle woods emerged as especially effective, including yarrow, horseweed, box elder, California buckeye, mule fat, redwood, cattail, mugwort and Douglas fir. The author then compared those results against measured density and found that lighter spindle woods generally performed better in the tests.
The page is careful about the limits of those observations. It explicitly calls some conclusions anecdotal and notes that density can vary even between pieces from the same tree.
That caution makes the material more useful, not less.
This is the sort of technical hobby documentation that old personal sites did unusually well: one person gets obsessed with a practical question, runs far more tests than a casual article would justify, publishes the failures along with the successes, and leaves the data online for whoever arrives next.
CacheRat’s 1,967 Ancient Web Domains research list contains many pages built from that same mixture of curiosity and stubbornness.
Stone Age Skills turns “what wood should I use?” into a small experimental dataset.
That is a much better answer than a listicle.