Race Predictor - Way off?

Mine also is quite off. Last year personal best: 1h:42min, this year predicted 1h:24min, while realistically looking I am able to run 1h:35min or in best case1:30 (but when I train with such pace (4m:15s/1km) I barerly can make 5km :)). I rather use it as overall progression to compare between different years, but I don' use exact predictions.

The race predictor uses your VO2max. If that is inaccurate, then the race predictor will fail.
A good thing to know and understand is that calculated performance metrics and their relation to real performance is a very individual matter. It might work pretty well for the majority, but be off for some individuals for no particular reason other than "everyone is unique".
It is also possible that you run well below your true capability, for various reasons.

Just to add something more to the discussion, I find the race predictor to be quite accurate in my case. It tends to understimate my capabilties on short distances and overestimate slightly on long distances, but generally it falls into the right ballpark.

I agree with TMK2 - its very dependent on VO2.

My best 5k time is only 30 secs of my predicted max, and my 10K is about 2 mins off. Not too bad.

From what I've read in these forums and as tmk2 said, race predictor is based on your VO2max.
My actual half marathon time and race predictor time is about the same as yours, so I suppose your vo2max is around 46-47.
I think the problem with my half marathon time is that, even though i dont run out of breath, my legs get tired so I cannot keep a good pace after about 12 km.
So I'm guessing that Race predictor only counts your aerobic fitness and can't really tell if your legs get burned out or not.

gchris7 The predictor is based on VO2 max but also presumably in conjunction with age gender weight etc - also as I'm sure you realise they are based upon doing the ideal amount/type of training. At present my Half-Marathon time shows 1:35:40 - that is probably a realistic time that I could reach with 6 months training including long threshold runs and a number of over distance runs as well as shorter intervals. However due to the level of my current training I would struggle to run more than 6 miles and that certainly wouldn't be at 7:20 pace probably more like 9:00

Having said that when my training has been more consistent focussed and higher in volume the predictions for 10K and half Marathon are about right

It's just a straight lookup table -- not even a formula or anything. Doesn't take into account anything other than VO2max and people with the same VO2max will have predictions identical for all race distances:
https://cicerunner.wordpress.com/2014/07/11/garmin-fr620-race-times-from-vo2-max/

Take it as a reflection of your theoretical capabilty. For me, it gives me relatively accurate estimates up until a marathon which I know I am bad at as I rarely long very long distances. For 5k, 10k it's almost spot on my current form - for half marathon I could do the projected time but would need to train differently.

That was my initial reaction but if you think about it it makes sense. If you assume VO2max is the most impactful metric then this sets a certain ballpark of what a person with a given VO2max is "capable of". If your real times are under, you need to work on the specific distance training. If they are above, you have mastered the distance and are beating the average.

I am not sure it's a fair assumption the watch will tell you how YOU will perform. If you want to know that, go outside and run at full pace. But it gives you a good estimation of where you should be, based on a physiological measure that's considered relevant.

Yeah, there are/were online race predictors which either go from VO2Max to race time or vice versa. The watch estimates relative VO2Max in mL/kg/min (which is already normalized for weight), so it's a decent proxy for race time, assuming that the VO2 estimation is legit.

It's not that different in principle from plugging in one race time to the McMillan online calculator and getting your predicted race times at every distance (where age, gender, weight, BMI, etc. are all completely irrelevant).

I have to admit that my equivalent race paces using a calculator like McMillan have always been a lot closer (at least up to the half marathon distance) to reality, compared to my Garmin race predictions from estimated VO2. But then again, I think my strength/durability has always been the limiting factor, not my cardio.

For better results, some people have suggested taking the Garmin predicted VO2, subtracting 3 or 4 units, and checking the same lookup table posted earlier:
https://cicerunner.wordpress.com/2014/07/11/garmin-fr620-race-times-from-vo2-max/