I didn't really know how it all works and found the following duped across a number of websites:
"The penetrometer reading is obtained using the following method: A 1 kilogram weight is released by a trigger action and falls 1 metre down a shaft, which in-turn hits a 1 centimetre square rod into the soil profile. The 1 centimetre square rod has 1 centimetre increments as the measurement for the depth the rod that has entered into the soil profile. This action is undertaken 3 times in the one position, therefore giving 3 readings eg: 2.5, 4.5, 6.5.
A formula is used to calculate the reading and is based on the average figure for the entire course. To obtain this figure there are 6 readings taken every 200m around course. Three of the 6 readings are taken 2m from the running rail and then the remaining 3 readings are taken 4m from the running rail. Each reading is recorded and then all figures are entered into a formula that will provide an average. This average then becomes your race day penetrometer reading and will correspond to your set ratings – Fast / Good / Dead / Slow / Heavy. Eg: Track – 'Good', Penetrometer – 4.25"
Some of the websites with that lifted text then go on to say that "readings will differ from track to track due to varying soil types and rail positions. Therefore you cannot compare readings against different tracks ie Randwick compared to Caulfield"
While the averaging of penetrometer readings gets you a single figure, it is difficult to equate it with a broad statement like "Track with good grass coverage and cushion" (Numerical Rating #3). The penetrometer readings don't have anything to say about grass coverage, for one!
Check out the map below for another reason the whole thing's a fudge (I tried to find current ones on TasRacing's website, but no luck; this was from a post in another forum):
If you used the same average method to gain a single figure but only over the relevant race distances, you'd get different figures if you were running a 900m vs a 1400m (what relevance does the firmness of the rest of the track have if the horse isn't even going to run on it?!).
In short, penetrometer readings add a veneer of scientific approach, but seem to be of little account in the awarding of final Track Rating. It gets you in the ballpark at best, but then local/expected/weather conditions are spun into the final opinion - which are even then only indicative as they aren't necessarily specific to each race distance.