This has been researched in the past (second google hit for "Strava power accuracy"):
http://stevebalescyclingblog.wordpress.com/...power-vs-real-power/ Based on some of the replies in this thread, I'm going to go with the individual is regurgitating what they've heard and not what Strava says vs. their own power meter. If you know that a computer algorithm is only able to calculate for one specific condition - specifically, mitigate your expectations - Strava isn't *that* bad. On my last ride in generally rolling terrain, with low wind, a longish steep climb (1mi at 10+%) and loads of flat fast riding (10ish miles of <±1%) on a 40 mile ride along the coast, my Quarq was ~10% lower than I would expect Strava to calculate for 1hr power (Strava, in my experience over estimates power on the short end, say up to maybe 30s power but that's to be expected based on calculations which input noisy data). To be fair, I should strip the power data from the Garmin file and have Strava calculate it, but that's what the link above is so much easier to read. Depending on nutrition, sleep (I have an 8 month old), or pacing, the ride could have brought my 1hr power 10% lower without much difficulty.
Bottom line, use common sense when deciding if Strava power is accurate - did you fight a huge headwind or catch a nice ride on a tailwind, have a very different weight than Strava thinks you do, are you looking at the power over a 0.1 mi segment where you mashed it, draft a car/rider, etc...? Of course Strava won't know those things, the only disappointment is people would expect Strava to know. Most likely, Strava calculates energy required to overcome gravity in hills, drag based solely on speed, and general friction which should work for most people - significantly changing those parameters will throw the calculation off (you can see this looking at Strava power between a road bike and mountain bike which have very different friction values). In my experience, if I'm riding hills or flats with no winds, it's reasonably close.
Alternatively, there are calculators online where you can input your own parameters, CdA, Crr, speed, windspeed, distance, time, etc... and it will calculate power (or speed/time/whatever depending on what you input); those calculators can be powerful as well but also are only as good as the inputs. Remember, computers will output exactly what you tell them - junk in, junk out. For being free, juxtaposed with the cost of a powermeter, non-power users (a group which I used to belong) should be so fortunate.