It has been more than a year since I have done a race, but my magic mile was 7:06 at the end of July (which was an improvement on my 7:55 magic mile at the beginning of May when we started training).
Congrats on the improvement! That's fantastic. If you can continue to drop that MM time, then it's definitely going to increase your chances.
So on Galloway's calculator a 7:06 MM is about a 4:02 marathon when fully trained. A mile is really really hard to project to the marathon, but my initial reaction is that it isn't impossible. I used the 7:06 MM to figure out what an equivalent 5k would be (around a 23:50). If you could run a 23:50 5k, then my calculator (based on Ian Williams and Vickers data sets) suggests the following probabilities based on a normal distribution of runners who can do a 23:50 5k:
10% chance of 3:53 or better
25% chance of 4:08 or better
50% chance of 4:37 or better
The further distance you have to compare (like a 10k or HM), the better it will be able to project. The huge range given is indicative that not everyone who is capable of a 23:50 is capable of sub-4. But someone with an equal 1:49 HM would suddenly have a range of 3:48-4:02 instead of 3:53-4:37. Since your training would probably fall on the lower end of what comparable runners are doing, I would say you're probably more likely to be in the 4:37 area than the 3:53 area. Runners who fall into the top 25% area typically have these characteristics:
Ian Williams: An Updated Race Equivalency Calculator Attempt
Conclusions
The conclusions we can draw from this:
-If HM performance is equal, women are likelier to finish with a faster M time than men.
-Runners of all abilities are capable of a 1.06 or less, and roughly the top 10% of all subgroups from 1:20 HM'ers to 2:00 HM'ers were roughly the same R value (or relative performance).
-Faster runners are better converters with a lower R overall average. Makes sense then why Rigel came up with 1.06 since the elite runners available to him would have been a similar pool to the faster runners in Williams dataset.
-Runners on the slower side of the HM performances tend to have more variability as a group because of the bad converters in their groups, not because of the lack of good converters. So more people on the slower side of HM performance training inappropriately for marathon performance.
-Roughly 5:00 to 5:30 hours per week on average for a marathon training plan is considered "typical" or "sufficient" by Williams.
-Those who run more than 5:00-5:30 hours per week are more successful at being good converters than are runners who run less than 5:00-5:30 hours per week.
-Those who do 5L around 100 barely appear different than those around lesser or higher numbers. The 5L would suggest it is lower on the predictive nature than other variables.
-Those who have 5L be a lower % of total mileage from 16 weeks tend to be the best converters. The faster runners also tend to be the ones with lower %5L values. Relying less on the long runs and more balance yields a better relative performance.
-Those who train at 40-80 seconds slower than race pace more often than not will be a good converter and have a R less than 1.15.
So based on this information, I'd say you're probably going to need more training than the 12 week Galloway 3-day plan to go sub-4 based on your current fitness. By no means a guarantee, but you would be an outlier in the data set.