But did punters misinterpret the ICM voting data?
Three weeks ago during the first Salmond-Darling TV encounter the first indications that NO was having the best of it came on the Betfair betting exchange where full data on trading is made available instantly and where you are able to track it.
In the two hours of the STV hosted confrontation a lot of money was traded and YES moved sharply backwards from the 22% chance position it had reached in the aftermath of the Glasgow games. The ICM poll that came afterwards merely confirmed what punters had seen for themselves – Salmond was losing.
So last night I kept a close at the markets which barely moved throughout the 90 minutes of the debate. YES started at a Betfair price of 7.4 (a 13.5% chance) and finished at 7.4.
The movement came after the the debate was concluded and the Guardian published details of its ICM poll of 505 people who had watched it.
As well as the “who did best” the 51-49% leads for NO on referendum voting intention before and after the debate in the poll were widely reported. This led to an assumption on PB and elsewhere that the referendum voting intention findings now had YES and NO very close. People were mistakenly comparing the debate figures released with other ICM referendum polling.
With its debate results the firm issued the following guidance:-
“It should be stated this this sample was pre-recruited on the basis of watching the debate and being willing to answer questions on it immediately after the debate ended. While we have â€˜forcedâ€™ it via weighting to be representative of all Scots, it SHOULD NOT be seen as a normal vote intention poll as it is premised on a different population type i.e the profile and nature of Scots who watched the debate is different to a fully nationally representative sample of Scots.
If punters moved to YES because the ICM debate sample appeared to be split down the middle they were drawing the wrong conclusion.
UPDATE: YES price back to almost where it was before debate started
The chart above has been revised.