Much Maligned Pollsters on DU were among the most Accurate in Iowa.

[bluewater]

It turned out that Change Research was the most accurate predicting the Iowa caucus. Civiqs also performed well.

Several pollsters that were highly rated by 538, like A+ rated Monmouth University and A- rated Sufolk University, performed significantly worse.

This table shows the root mean square errors (rmse numbers) and the mean absolute errors (mae numbers) for the various pollsters predictions versus the actual results.

LOWER NUMBERS ARE BETTER

Link to tweet

The polls are what the polls are. And some polls were better than others.

[Squinch]

1. How can you tell? We don't have Iowa results yet.

[bluewater]

3. Based on results released so far. Will of course be updated for final results.

From 62% released to 74% released this trend seems to be holding.

[Squinch]

4. 40% of results are still to come. The OP is premature at best.

Irresponsible at worst as it is advocating trusting polls that may be completely wrong.

[bluewater]

7. Your math is off. 62% were released the very first time. ;)

But, time will tell if the 74% results currently released match up well witht the final totals.

[Loki Liesmith]

2. Average the polls

[bluewater]

5. But pay attention to pollsters actual performance too!

Why, even 538 rates pollsters' performances and uses those rating in their own Poll Aggregate Tracker.

[Kentonio]

6. To be fair Iowa polling is always a graveyard for predictions

Selzer apart of course (in normal years!).

[bluewater]

8. Good point. So it's On to New Hampshire!

In about a week we get to see how the pollsters do there.