Quico says: By popular request, here's the Constitutional Reform Referendum poll chart. (It's a bit hard to read like this, but you can click to enlarge.)
As you can see, there are two clearly different sets of answers here. When you ask voters in general if they favor or oppose the constitutional reform, you get a clear, consistent majority against it:
But, at the start of the campaign, most people who opposed the reforms were saying they wouldn't vote. So, until the last few weeks, the "Sí" camp had a majority among likely voters.
The trend, however, is for more and more "No" voters to say they've decided to turn out to vote after all. The rise in people's stated willingness to vote has transformed the race into either a Dead Heat (according to Hinterlaces and Datos) or a into clear lead for the "No" side (according to Datanalisis, Keller and a rather optimistic seeming Mercanalisis poll.)
So what we're facing is a remarkably fluid race: one that will come down to turnout.
Causes for caution: as you can see, many of these polls were carried out in urban areas only, and a couple of them just in the Caracas Metro area. The opposition always does relatively better in urban than in rural settings. Over a fifth of Venezuelans live in towns and villages of less than 20,000 people: they almost never get polled, and they trend heavily in favor of the government. Moreover, identifying "likely voters" is a notoriously haphazard game: more art than science. Venezuelan pollsters usually rely on people's self-reported willingness to vote, which is an iffy proxy at best.
A note of optimism: Though the race appears to be very close, the momentum is clearly with the No side. The government is bleeding votes fast, as you can see in IVAD's dramatic tracking poll results for the capital:
The magnitudes here are not significant, since it's a Caracas-only poll. But the trend is significant, and very striking.
A plea for help: Notice any mistakes in these charts? Have details of a credible poll I'm missing? Please send them in. (Oh, and the first person to leak Consultores 21's super private final poll gets a bottle of something.)
Finally, those of you minded to dismiss "opposition pollsters" due to bias are urged to check out how they did ahead of last year's presidential election.