Which was more surprising?

[Miles]

Tough, but I'd say Arkansas.

I wasn't expecting it to be particuarly close (7-8% win for Cotton), but when it was called right after polls closed, I knew the result would be pretty damning.

I was also expecting Orman to be close, but like third parties or Libertarians in general, I had a hunch he was overpolling.

[ShamDam]

I'd have to say Roberts winning by 11. That race was supposed to be a dead heat. Put it this way: shift both races 11 points toward the loser and I would have been much more surprised by Pryor overperforming (losing by 6) than by Kansas being a toss-up.

[Attorney General, LGC Speaker, and Former PPT Dwarven Dragon]

Roberts. I didn't expect Cotton to win by the margin he did, but you could tell he clearly had the momentum in the final days. Roberts, on the other hand, was trailing in polling averages right up to the election, and in the end he won comfortably.

[TDAS04]

Both were very surprising, but I voted option one.  I didn't think that Roberts would win by more than 5, while a double-digit margin for Cotton seemed very plausible shortly before the election.  Pryor still didn't lose by as much as Blanche Lincoln.

[Miles]

Pryor still didn't lose by as much as Blanche Lincoln.

But if you told me before the election that he'd win fewer counties than her, I probably wouldn't have believed you.

[IceSpear]

I saw it as a realistic possibility Pryor would lose by double digits, if not particularly likely. I would've thought Roberts winning by double digits was near impossible.

[ElectionsGuy]

Roberts. I thought that it would be the closest race of the cycle, and the winner would be below 50 no matter what.

I predicted an Orman win, one of my most embarrassing predictions.

[Mr. Smith]

Not really, Orman had a decent campaign and Roberts was quite lousy, he was saved by lots of financial backing and low turnout.

That said, I had expected Roberts to win at the most by 5 points.

And yeah Orman was my prediction as well.

[Xing]

I had a feeling Pryor was toast, and that it wouldn't even be that close. Kansas, on the other hand, really did surprise me. Even though polls can overestimate third party/independent candidates, I didn't think the polls would be THAT far off.

[Nichlemn]

I'd say Arkansas, because incumbents losing by a lot is usually more shocking than an incumbent winning by a comfortable margin. Just a few months before the election, if you'd said that Roberts would win by 11, that would be unsurprising, or even a fair bit under expectations. It was only at the last minute that a bunch of polls suddenly suggested a hitherto unprecedented outcome of a Republican Senator losing in Kansas.

And going by how things seemed to stand at the last minute, 538's forecast supports the idea of Arkansas being more shocking. Both results fell outside the 90% forecasted range, but in the case of Kansas, only just, as he gave a pretty wide confidence interval, stretching from Orman +10 to Roberts +9ish. That makes sense: a race with an Independent and an unpopular incumbent in a red state is difficult to predict. In Arkansas, 538 gave a much smaller confidence interval, about Cotton +0 to Cotton +10, and yet Cotton still obliterated it.

[Miles]

And going by how things seemed to stand at the last minute, 538's forecast supports the idea of Arkansas being more shocking. Both results fell outside the 90% forecasted range, but in the case of Kansas, only just, as he gave a pretty wide confidence interval, stretching from Orman +10 to Roberts +9ish. That makes sense: a race with an Independent and an unpopular incumbent in a red state is difficult to predict. In Arkansas, 538 gave a much smaller confidence interval, about Cotton +0 to Cotton +10, and yet Cotton still obliterated it.

I was also going to add Perdue winning outright as an option. I was expecting a runoff, but 538 had him at 49.7% on election day, so winning outright wouldn't be that surprising.

[© tweed]

Roberts' performance shows how much the two parties have GOTV institutional advantages.  there's a real art/skill to it that $$ can't automatically match.

[Mister Mets]

Cotton's margin. Pryor had seemed relatively competitive, and he was an incumbent.

Roberts was a Republican running in Kansas, and they seemed to paint Orman as a vote for Harry Reid pretty effectively.