It would be better to ignore exit polls as their sampling only uses 1,500-1,800 polling stations out of ~800.000 polling stations. And for parliamentary elections, some polling stations just about to start counting the ballot and it's a long process because there are 4 different ballots and election officers have to validate and allocate various ways of punching the ballots (for political parties only/for candidates only/for both/etc)
A random sample of representative polling stations to form the quick count can be very predictive. They did it in Singapore in 2015 and for each constituency the quick count was within 1% of the final result. Of course the key words here are "random" and "representative"