The graphs I produced, such as this one
seem to show clustering of results around what the cassette offers (its a Shimano 105 11-28 so has 11-12-13-14-15-17-19-21-24-28).
The next question is: if these are my results from actual ride data, what would a theoretical approach look like?
So I took the data from a longer ride (roughly 65km over 2.25 hours giving about 10,000 data points), and again plotted speed vs cadence: NB speed in kph, cadence in Hz.
Again, I used Google Spreadsheets to calculate theoretical values. What I did was to create a matrix of speed vs teeth, and each data point is the cadence in Hz, populated using the formula discussed previously:
which I've plotted. Each coloured line represents a particular sprocket
So is the analytical model valid? Here's the two charts overlaid and scaled
Next: is the pedal cadence I've arbitrarily set OK? I test this in a bit...