Strava is
The Strava power estimates work very similar to the PowerCal from Cycleops. It makes an estimate based on instantaneous changes in heart rate. The better your HR strap the better the estimate. The longer you time sample the better the estimate.
So it depend on how you want to use it. If you want to compare ride efforts longer than 1 hour; at moderate and above efforts; then it's not too bad and often it's very good. If you can get some rides with a power meter then you can judge how much it's off from real which allows you to use the numbers between 30 minutes and 1 hour.
A PowerCal is actually more useful than strava's estimates (even thought they use the same theory). That's because Strava will think it's power meter and do things like factor it into your training logs and rolling averages, and turn of features that are only allowed for real meter; like fatigue versus fitness. The Strava estimates are excluded from such analysis so as not to polute the data of people using real meters on some bikes but not others.
When it comes to real-time training data. The estimates aren't as useful. This is becuae from 30 second average power and below the readings from Strava and PowerCal are on the high side of garbage, and you get very high power readings than never happened. This is why you'll probably never see the Strava App Estimate power on the fly. It certainly could but it wouldn't be very useful.
The estimates are best rides ranging from 20-100 miles with good consistent effort. For those the average power estimates for me tend to be 25-30 watts high. If you take shorter ride and mash around in a big gear you can get some spectacular garbage data.
Here's the tour da tonka; 50 miles moderate effort in a crowd with too many rest stops and plenty of DF's to draft from all day.
Here's a 50 miler solo in the wind out of shape at the start of the season with a real meter.
Here's a night time; putter ride at 14mph in a gigantic gear estimate
Here's an almost double metric of moderate effort.
If you look at the short 9 mile ride; the numbers are just dumb. I was in street clothes; at night, in the dark with 45 degree weather; the numbers are high because the night time is more dangerous and external stimulus makes the heart rate jump. So the graph looks like I went for 9 mile sprint when in fact I was just riding through the dark neighborhoods for the sake of riding.
On the 3 long rides you can see that after 1 hour the numbers align to be in the 150 watts range which is what I typically hold on for long distance cruising where I don't try to go for broke. Below 1 hour the estimates tend to run high; and between 5 minute and 1 minute it's high; and then below 1 minute it's garbage.
So Estimated power is also really only good for rides that are over 1 hour long and from middle effort on up. It will tell you pretty accurately how your 1-5 hour fitness is; and you can actually compare effort on a longer rides. Useful if you rode like a machine last week with no wind and averaged 20 mph; and this week you pulled a similar effort in a head wind; felt awful and averaged 16 mph. Pull up the power graphs and find out that you did .4 watts/kg better on the slow ride and you better understand how you are doing.
In short it's a good intro to learning how to use the data. Once you decide you want to Train with power you will need the real thing, you'll need to read a book or too; and you'll need about 3 months of base data (collect and don't analysis); then you can start training with it; to go faster.
You can't "train" with a PowerCal; but you can gather more usable baseline data than you can with the Strava Estimates; so if you are thinking about getting into Power training; get a powercal; start gathering data; and read the books on the topic. If you understand all that and start to analyze your data you'll know if you want a meter and whether you'll actually use it.
