FR 945 accuracy of race time predictions

It's called a prediction, not a crystal ball.

"But what about a hilly course on a hot day where you only ran for training purposes?"

This kind of run might be a good candidate for using the Trail Run app to keep the activity out of VO2max estimation and hence race prediction. I think that you could then edit the activity in Garmin Connect to the appropriate type without affecting the data on the watch.

I just started using Runalyze a week ago based on a comment in these forums, and, wow, I'm impressed. But there does seem to be a major limitation in that you can't discern a trail run from a road run. Sure, you can probably define a custom activity type or a tag (I'm still learning) and manually edit each run, but it's a pain for someone who does both regularly. Of course, while Garmin at least has separate activity types, they're lumped together in reports anyway.

In Runalyze you have a checkbox in the edit screen of an activity where you can exclude the trail run from VO2max calculations.

About the type: just ask on Facebook or Twitter or their forum. The developers are very responsive.

yes in a word

I have seen a few anomalies like doing an officially measured race faster than the prediction after it but generally it is pretty much in the ballpark. My best 5K this year is 17:57 and 10K 37:37 and watch currently predicting 18:04 and 38:08 which is probably about right as I got ill after doing those in quick succession and feel am starting to get back to that fitness. VO2 Max too - was around 59/60 when did those races dropped to about 56 when was ill and now back up to 59. With the 935 it used to predict times I could only do 20 years ago when was 30 something not 52!

It's actually getting worse.  The algorithm doesn't understand running easy, and coupled with terrible data from the wrist sensor (aka unreasonable high HR data - ie 40bpm higher than finger measurement ) I'm getting only useless data.   I'm apparently 'overreaching and my VO2 max has come down despite running 21k with 11k @ 3:38 avg pace yesterday pretty comfortably and then running 42k easy today at the ned of a week of 182km starting with the 34min 10k last Sun.   

The watch thinks I'm a "deadman walking" and cannot cope with high level training I guess, despite the fact that I can clearly hit my paces in every workout.   I've now turned off the performance widget and training status because the inaccuracy is annoying.  

Considering my resting pulse is consistent at 37 each day I feel fine running and hitting my desired workout paces I highly doubt I'm on the edge of life if you get my drift.

Garmin for RHR but my manual check laying down isn't so far off.  My manual check is really during a workout when garmin says I'm dying at 185bmp on an easy run as I'm talking quickly (aka hard to do at 185!) and my finger says 140 or so.

Further "detraining" metrics: since running 42k on Saturday I've run probably over 60kms.  Today I ran 5 x 800 @ 2:38 (like 3:17/k pace), which is at my higher end effort.

Second run today easy recovery around 5:00/km with HR avg 136.   And yet Garmin still says I'm detraining/unproductive.

So 

1) Increasing my distance of my LR by like 6k

2) increasing weekly mileage from last week 

3) increasing speed

4) fully recovered 

5) detraining/vo2 coming down/performance coming down further than it was before.  

I tried to turn off the widgets but when you save an activity it shows it to your face anyways, and again in Garmin connect mobile.  

"terrible data from the wrist sensor (aka unreasonable high HR data - ie 40bpm higher than finger measurement)"

This is why you see what you see. It's nothing new that WHR works for a few, but not for most. The best way to ensure you can get close to reliable data is to use something that tracks your heart rate better- a chest strap. If you don't do that, you're destined to be continually disappointed.

Feeding the algorithm garbage data will result in garbage results.