Opinion polls have a reputation for inaccuracy in India, but how inaccurate are they in reality? Has a meta-study been conducted on them? The reason I ask is that as I see it, there are two main steps where the polls can go wrong. Either they get the vote percentages wrong, or they make an error in translating the votes into seats. My ill-informed guess is that the first step is easier to get right than the second. It is possible that they get the vote percentage right and screw up in their model that will translate it into seats, or a small error in the first leads to a large error in the second. It should be possible to test this hypothesis by looking at the data for the last few elections. I am willing to bet (though not a large amount) that the non-shady organizations that conduct the polls get the vote percentage right. (I am also willing to accept the possibility that there are in fact no non-shady pollsters.)
The reason for this speculation is that it occurs to me that if I am right, this time the polls should be very very inaccurate, because the delimitation must have completely screwed up whatever models they have to translate votes into seats.