A power meter can be a good tool for a cyclist to use, but I think some people rely on them too much.
I started thinking about this the other day when a beginner/intermediate rider who I just met, told me that his new coach said "if you aren't training with a power meter, then there's absolutely no way that I can help you get faster"
Power meters can be great. When doing longer tempo intervals, they're great for pacing and for getting yourself to push harder at the end, but training with a power meter usually doesn't get you to ride the way you would in any race situation.
Let me use a few examples: 1) Newton Heights (a hilly crit in Victoria), 2) any road race with a single hill in it, and 3) the first 10k of the Test of Metal.
1) At Newton Heights, racers complete multiple laps of a 60-90s climb, a 60-90s descent, and a 60-90s section of flat road. During the opening laps, riders hold 450-550w up the climb (that's for me, being 150lbs), but towards the end of the race, the pace drops and the wattages aren't nearly as high. When you train with a power meter, you try to keep the same wattages throughout the workout, so your first few efforts will seem quite comfortable. This does not simulate racing.
I remember Roddi Lega being dedicated to training with a power meter when he was in Victoria, and over the winter he did lots of steady intervals and hill climbs around 350-400 watts. For most people, those intervals would be hard, but I think Roddi realized that when he jumped into Newton Heights, most people go out way too hard on the opening laps. Roddi might have been saying 'these guys can't sustain this... why are they going so hard?', but that's the way racing is. People are always overly ambitious, and the power most people can sustain at the end of a race might only be 75% of what they could produce at the start. Roddi wasn't used to producing 500-600 watts at the start, so he didn't race well.
2) Any road race with a single hill in it is only going to be hard going up the hill. People in the race realize that this is the only place where riders can be dropped and selections can be made, so in order to drop people, riders will produce a crazy effort up the climb. Obviously riders can go hard on other sections, but for the purpose of my argument and based on past racing experience, many people would agree that pace changes and accelerations on hard sections cause the splits in racing. Anyways, my point is this: you need to practice being right on the rivet and breathing through your eyeballs, just like you would be over the crest of a climb. Training with a power meter usually doesn't encourage this.
3) The Test of Metal pretty much starts with a VO2 max test. Riders coast off the start and are riding easy, until a few kilometers in and the grade of the road starts to slowly rise. It gets a little steeper and a little steeper, until the group thins out and you can hear the few riders around you gasping for air. At the end of the road section, riders sprint to enter the singletrack first and to win a prime at the top. After cresting the hill, heart rates are at max, and there's still 65km to race.
I think power can be used as a good comparison from workout to workout, but holding back at the start of a workout just doesn't seem to make sense. Riders don't usually ease into an attack. Or ease into a start loop at the start of a MTB world cup.
One way that I don't like to see people training with power is when they set power ceilings for rides they'll go on. They'll say "I don't want to go over 300 or 350 watts today". Structure can be good, but too much structure can take the fun out of riding, and it can deter people from making hard efforts when they're tired, which isn't necessarily a bad thing.
Anyways, just some thoughts. I welcome you to question them.