I've never heard of George Parrott other than through the Excel-based running log I've been using for the last several years. And that log was put together by some other guy I've never heard of, named David Hayes. And that log isn't easily found to download, as far as I can see, so this probably makes no sense to you--unless you use the same running log I do.
But anyway, my running log has a page called "Parrott Predictor," which is based on this George Parrott's idea and is supposed to be able to predict your marathon time based on your actual training paces.
According to Hayes' spreadsheet, "George Parrott, known affectionately as 'Coach George' on the Dead Runners Society listserve, presented this idea a couple of years ago. He suggests that a runner can run no faster in the Marathon than he runs his fastest 26.2 miles in a week during training. Calculating what miles are the fastest, adding them up until you get to 26.2, then adding all the times associated with those runs can be tedious. Thanks to ideas contributed by Dale Shoup of DRS, this worksheet does it for you."
No, I don't know anything about the Dead Runners Society, but the only live link I've found to Hayes' spreadsheet is at the DRS website. Running Times magazine had a review of the free running log back in 2005, and there's a Runner's World discussion thread about the aforementioned Parrott Predictor and how bogus it is.
However, that won't stop me from testing its accuracy. Last week, I logged just enough miles to get my first prediction: 4:41:48. I've been saying that 4:30 is my "A goal" for the Akron Marathon, so it seems I have a little catching up to do to quicken my pace from 10:50 to 10:18 per mile.
Last year, my best prediction was 4:01:35, and I ran a 3:58:46 PR. But that was chip time; my gun time was 4:02:10--just 35 seconds slower than the Parrott Prediction. Going back to 2009, my best prediction was 3:51:18, and I was way over that with 4:04:47.
But remember, the description says I can run "no faster" than my fastest 26.2 miles during a training week. If I stick to gun time, which is the true clock time after all, the Parrott Predictor works. For the next, gulp, six weeks (really?!?) I'll be reporting my Parrot Prediction score. Hopefully, we'll see trend toward faster times. At the very least, it will give you a gauge to predict my failure on race day.