Using betting market data from Betfair, I have the NO vote in the Scottish independence election at 84% to succeed. And, I can break that down:
6% for 0% to 40% of the vote for YES
26% for 40% to 45% of the vote for YES
51% for 45% to 50% of the vote for YES
12% for 50% to 55% of the vote for YES
3% for 55% to 60% of the vote for YES
2% for 60% to 100% of the vote for YES
There are two key points. First, this is a really wide range. The polls are all clustered at NO being a few points ahead, but the markets are giving 32% that NO wins by 10 percentage points or more! Second, if the polls are way off, the swing is going to favor the NO vote. Intelligent review of the polls are both surmising and determining this possibility.
To provide some perspective on this crazy error possibility, let us consider what an 84% election the in US looks like right now. Gary Peters is 83% to defeat Terri Lynn Land for the open Michigan senate election. My current fundamental and poll-based forecast has Peters getting about 54% of the vote. I would put the possibility at him getting 60% approaching 0%!