Of the 3 5K races I ran in July, my best time was 24:00. According to the McMillan Running calculator, my equivalent times for other distances (based on a 24:00 5K) are: 6:55 mile, 49:51 10K, 1:50:56 Half Marathon and 3:53:58 Marathon. These are encouraging numbers, assuming you can actually rely on this data.
However, I find the accuracy of the race prediction calculators varies, sometimes widely. Personally, I've attributed the variation to the lack of specific training for a particular race. For instance, if your 10K race prediction is 49:51, but you don’t train specifically and adequately to race the 10K distance, the race prediction time may vary somewhat. I think this is especially true when you use short race distances to predict your marathon time.
It follows that I can only hope to think my (current) marathon prediction time of 3:53 is accurate if I am adequately and specifically trained to run the marathon. So, what does it mean to be inadequately trained for the marathon? What training factors are more likely to cause the deviation?
Most people agree that training mileage in a marathon program is the most significant factor in the accuracy of the predictions. I came across one study that looked at the marathon to 10K ratio versus mileage. It proposes using the following conversion factors when using your 10K time to predict your marathon finishing time:
55 mpw: 4.9
60 mpw: 4.75-4.85
70 mpw: 4.70-4.80
80-100 mpw: 4.55-4.65
The higher mileage the runners in the survey ran, the lower the marathon-to-10k correlation ratio, i.e., the faster their marathons were relative to a 10k time. This supports the premise that perhaps total training mileage is the most important factor for marathon success. It's also interesting to note that race equivalency calculators are based on the high mileage run by very advanced, highly competitive runners.
On a personal level, this just confirms my choice to increase my training volume to (peak at) 80-100 mpw is a good one:-).