In some sense, today "resolved" the question of whether Kipchoge has limits, and (more reasonable) which side of two hours those limits lie on. He went out just under 2:00 pace and faded significantly over the second half; ergo, the (legitimate) two hour marathon is just a bit beyond him.
I'm not really going to argue with that conclusion, but I'll present a thought experiment suggesting that he is closer than that simple story suggests. All of this is pretty ad hoc, so if you don't like my reasoning, say so! That's the purpose of a thought experiment
POINT 1) Kipchoge went out faster than ideal for a two-hour clocking.
First, note that 59:51 is not 59:59, and so the opening split was not quite idea. But I'll go beyond that: there's a school that the idea marathon is a slight negative split, and the other recent world records at Berlin were mostly negative splits. The past 8 men's world records (since Khannouchi) have all been set at Berlin, and here are their split differentials:
Kipchoge 2018: 61:06, 60:33 (-33)
Kimetto 2014: 61:45, 61:12 (-33)
Kipsang 2013: 61:32, 61:51 (+19)
Makau 2011: 61:44, 61:54 (+10)
Gebrselassie 2008: 62:03, 61:56 (-7)
Gebrselassie 2007: 62:29, 61:57 (-32)
Tergat 2003: 63:04, 61:51 (-73)
Let's remove Tergat to be conservative; his race was so long ago that I don't think that it's an apt comparison (or you could consider him the counterpoint to Kipchoge's positive split today). The other 6 races have an average negative split of just short of 13 seconds. I think that's a fair assumption. Note that Assefa split 68:13, 67:24 (-49), so there's no evidence that today wouldn't have a good day for a negative split.
POINT 2) Running so fast early takes a toll.
The question is how much time should the adjustment be. I'll use the 2-to-1 adjustment that I've seen some places: if you go out 1 second too fast, you'll give that second back along with another one. Obviously it's nonlinear (see e.g. Belihu), but I think this is fair (although unproven) estimate for Kipchoge's race.
From the previous point, the idea first half for a two-hour clocking would have been 60:06; Kipchoge ran 59:51. That's 15 seconds, so knock his time down to 2:00:54. But that's still not quite right since he would still positive split in this scenario. The correct solution to this "linear programming" problem is that 60:24, 60:12 would have been about the correct splits. He starts 33 seconds slower, and gains twice that back in the latter half. His overall race time is 33 seconds faster, for a final time of 2:00:36.
POINT 3) Going out slower could have helped the pacemaking situation.
Today, all three pacemakers were gone at 25km, and two of the three before that. Even before that, Kipchoge wasn't always quite as shielded as he was during his unofficial sub-2 attempts. I've seen estimates between 1 s/km and 6 s/km for the benefit of drafting. Let's go in between, but mostly with the former, and give Kipchoge a 10 second benefit from better drafting, 1 second for each kilometer between 20km and 30km. This should take into account both the imperfect drafting through 25km and the total lack of draft after 25km.
Since most of this extra drafting comes from the second half, he doesn't quite get 10 seconds from this so as not to "double count" the previous adjustment. Instead, I'll do this adjustment first, and give him 59:50, 61:09, and then apply Point 2 to those numbers. This puts Kipchoge at 60:20, 60:09, for a final time of 2:00:29.
POINT 4) Weather could potentially have been a bit better
Now, we're getting pretty speculative. I can't find the precise weather conditions during the race, but if I remember correctly, it was in the low 50's Fahrenheit. Ideal is thought to be more like 45 degrees, maybe even a touch lower. So 5-10 degrees warmer than ideal, although you might dispute that.
I'll only give 5 seconds here because I'm uncomfortable giving too much when I don't have data to back it up. Maybe you're feeling even more stingy than this; in that case, just remember that the weather can always be better. In fact, with atypical conditions (e.g. right after a storm, strong winds shifting at precisely the right time) I'm sure we could gain enough right here to get under two hours. But I don't want to push it that far, so we'll stick with 5 seconds. Final time: 2:00:24. (Coincidentally, one second under his first Breaking-2 attempt).
POINT 5) Intangibles
This is the section for all those little things that could be just a bit better, and probably would be if Kipchoge really were on the knife's edge. For instance, suppose he had been explicitly aiming for two hours, and suppose he had already run a time close enough to back it up (like the 2:00:24 above). They're going to set it up as close as possible to INEOS without breaking the rules. You can bet that the organizers are going to have more than 3 pacemakers. and perhaps they'll hire a couple super-pacers. Pay Kiplimo or another super-stud to dedicate his season to being able to run 35km at two hour pace. With so many runners under 59 (even 58) nowadays in the half marathon, I believe with a bigger commitment we could have pacing that is far more reliable and effective, particularly in the vital 20-35km section. Or what about using Wavelight? Could that perhaps make an even bigger difference in the marathon than on the track?
Also, who is to say that Kipchoge was completely 100% maxed out today. What happens if he hits 40km on pace for say 2:00:08? How strongly would you bet against him finding a few more seconds somewhere?
Combining all possible factors in this category could give you many seconds, but I'm wary of assuming too much. I'll give him 10 seconds: 5 for race changes that weren't considered in the previous points and 5 for the adrenaline rush of being so close.
FINAL TIME: 2:00:14
Summary: this sounds about right to me. I'm trying to give Kipchoge every "reasonable" advantage, but no "unreasonable" ones. Obviously, we've seen that having the latter does put him under the two-hour mark; I've estimated that the former barely does not. But I also want to mention that while you might quibble with my estimates, they're pretty minimal, and none of them is assuming an absolutely perfect confluence of events.
I'll say this another way that makes it a bit more concrete: I think if Kipchoge, and the Berlin marathon organizers, and all of us go back in time six months, with full knowledge of what happened this time, and if the weather is just the tiniest bit more perfect, and if Kipchoge pushes through the latter stages with every ounce of his prodigious willpower, then everything above happens and 2:00:14 is the time I think he *really would* run.
I truly believe it's possible. And two hours is not far off.