Talking about predicting marathon finish times, here’s a good piece by McMillan, covering fast finish long runs, sub distance races like half marathons, and Yasso 800’s. My personal take on this is, fast finish long runs are a great workout for your overall aerobic endurance and particularly for your mental strength towards the finish, but not necessarily a marathon predictor. Yasso 800’s are a great workout to train your 5k and 10k race speed as well as maximum oxygen uptake, but also not a particularly good marathon predictor. Sub distance races work much better, for me at least (read: anecdotal evidence), and they are used by coaches like Herbert Steffny and Jack Daniels with success. Mind you, these predictors alone don’t guarantee that you can run a marathon in the predicted finish time, if you don’t do your long steady runs and everything else.
Yasso 800’s are often critizised for not being related to marathon running at all. But if you consider how Jack Daniels uses Vdot to normalize and compare finish times for different distances, you can see that these values are indeed related – but Daniels uses them in the reverse sense. If you run a race at a given time, you can determine your future training paces from there. So for example if I finish a 10k in 42:00 minutes, my Vdot is approximately 49, and I will use the training paces in the Vdot=49 row from then on, until I finish a race in a time associated with a higher Vdot, after which I will switch training paces again to the corresponding Vdot row. So – Daniels will derive training paces from finish times, but not vice versa.
So – being able to run 800 meter intervals or repetitions at a certain pace covers only a very small aspect of marathon training, namely subdistance training. And 10x 800 meter intervals are equal to 8x 1000 meter intervals, which are pretty tough. Steffny doesn’t even suggest them.