I've been meaning to write a bit about the figures I use on this site for ages. Firstly, a note about the cycle computer and GPS that I use.
The cycle computer is a relatively cheap bike computer that was (I think) about ten pounds a few years ago. It's a wired computer that counts the number of wheel revolutions using a magnet on the spoke and a sensor on the fork, and tells me things like distance (per trip and cumulative), average (mean) speed, max speed, time spent moving, cumulative time spent moving and the time.
The GPS I use is a Garmin etrex Legend that I bought a couple of years ago - more info here on the Garmin website.
As a physics graduate, one of the first things that was drummed into me at university was that precision and accuracy are very different things. I might say that my front door is 8.38320333725164 metres high; that is a very precise figure, but almost entirely inaccurate. On the other hand, I might say that my door is about 2m high; that's more accurate, but not terribly precise. It's worth saying though that the figure of 2m may well be a good enough number to use, depending on what I'm doing; obviously if I'm buying a new door for my doorframe then 2m isn't precise enough, but if I'm estimating how tall a building is and using a door as a unit of measurement then approximating it to about 2m is probably good enough.
So how does all of this relate to my cycle rides? Well, firstly, neither my GPS nor cycle computer quote errors on the numbers they give out - which isn't a problem, as long as one is aware of the maths behind it and what they're actually doing.
The computer is basically a glorified counter; it just counts the number of revolutions of my wheel and multiplies it by a number (the circumference of the wheel) to calculate the distance (and speed etc) that I've travelled. Its precision is normally shown to two decimal places - perhaps it'll say I've been 8.53 miles or 9.32 miles. To cut a long story short, in practice this means that it's probably reasonably correct to the first decimal place - it's not going to be massively wrong, but particularly over longer distances any small error in its calculation will be multiplied up once per wheel revolution. As it happens, when I've been comparing cycle computers with friends who've ridden the same ride as I on occasion together (and bearing in mind their computers have similar issues as mine) we've discovered that they normally agree to within about 1 or 2%. This isn't actually too bad so gives a reason to think that perhaps they're not too far off.
You'd think that the GPS gives a more accurate reading of distance travelled, wouldn't you? After all, it's based on spangly technological satellite technology, so it must be better! Well, in a sense it is... and in a sense it isn't. The first thing to note is that in the grand scheme of things my GPS is a cheap unit; unlike the bike computer, it checks to see where I am every few seconds and works everything out from that. That's all well and good until it loses reception for a moment - perhaps I've gone in a tunnel or through trees - and it has to guess what I'm doing. The thing is, I might be cycling round in circles (I don't tend to do this, but bear with me) or back and forth out of sight, but when it next sees me it assumes I've gone in a straight line; this can lead to fairly major discrepancies as it thinks I've not travelled as far as I actually have, or says I've got a slower average speed than I have. There are other issues, too - I've reduced the sampling rate a bit so that it saves the batteries, for example. The consequence of this is that the GPS is an excellent tool for finding out where I am at any given point in time (either on its internal map or by taking the OS grid reference and looking at a real map) and tracking one's route for analysis later, but not so great for detailed speed and distance information for any given route - although it's good for order of magnitude distances and speeds.
The practical upshot of all of this is that I normally quote both sets of info that I have (if I've got them both!) on this site so I, and you if you like, can analyse the raw data yourself. Anyway, since things like my personal best are all measured using the same computer (I use the cycle computer on virtually all rides), it doesn't matter if they're slightly out from the true value of the distance; if I know my previous best was 32m and I've now done 32.5m, I know I've cycled further on this occasion. After all, if I wanted to know more precisely how far I've travelled I'd be using a more expensive cycle computer or other timing device, but that's verging on the obsessive... (and yes, I know I might be bordering it already, there's no need to tell me!)