Your tracklog seems to have it as 370 meters, though I can't tell for sure what lane you were in, likely an outer one. Google Earth gives 322 meters for the inner lane, 364 meters for the outer. It's the inner lane that counts, right?
My Garmin always registers long on a track, who knows why. So that makes 333 a more credible number.
I think a track is measured in lane one, a certain distance out from the curb (a foot?).
Reminds me of a certain friend of ours who convinced himself that he could run 400M in 60 seconds.
My Google Earth measurement was not precise enough for it to matter what part of the lane I was in, of course.
As an experiment, it might be interesting to try a track session wearing your watch on the other wrist, and one with it on your torso somehow (in a pocket or something, and not on a swinging arm), and see if that affects the overreading.
I think in this case there is more than a simple case of an extra 5m per lap being added or subtracted because of the watch being on an outside or inside wrist instead of on top of the head. Looking at the traces running to and from the track on Enfield Road and Exposition Blvd. it is clear that the watch is doing not too bad a job of at least being on the correct side of the street on Enfield Road, and right on the sidewalk and turning in at the correct place on the driveway into the track while approaching the workout, but then getting displaced terribly when turning around and approaching the track. Probably Peter's body getting in the way of receiving a direct signal from one of the satellites, and the watch either getting a delayed reflected signal, or simply trying to do a best guess while down one satellite from what it really needed. Clearly the travel on both ends of the track are grossly displaced from where they should appear on the orthophotos in the Google Earth view, and the 1,2 side of the track displaced considerably further away from the 3,4 side than it should be. I'm also wondering whether Peter is using the Garmin in a "smart" mode where it is deciding when it needs to take another position point, or whether it is set in a mode of taking a position every x seconds (probably preferable). The corners of the Garmin trace look very squared off relative to the track (which already has rather small radius turns), and it may be throwing in some predictive guesses based on the straightaways into estimating where it needs the next points, making the measurement too long because of approximating the oval with something more rectangular around the outside. In any case, the Garmin trace seems a lot happier on the part of the run not on the track when Peter is running to the southwest, rather than in other directions! I've generally found I get a lot nicer agreement of my runs as recorded by gps datalogger compared with where I thought I went if I wear it in a hat on top of my head, rather than lower down where the view of some satellites may be obstructed by my body.
PG is correct that a 400m track distance should be measured at .3 meters from the curb. My Garmin rarely gives me the same distance for the numerous 400m intervals I have been running this year. Can be up to 3 meters off either way. ( I wear it on the left wrist)
3 meters over a 400 meter loop honestly doesn't sound disappointing to me when you think about what's going on here.
Wouldn't it be much easier to measure the curb and then do some arithmetic?
It may well be that that's how they do it.