There isn't much polling available at the moment on the new Canadian political scene (post Martin and united right), just a Reid and a Leger Poll taken around the time of the creation of the Conservative Party of Canada. Both polls had the Liberals around 60% in Ontario. At that level of support, using my forecast model, we find that the Liberals would sweep most, although not all constituencies - the NDP still win one or two (those being Windsor-St. Clair which they won in 2000 and Trinity-Spadina).
However, the Liberals have registered 60% in polls between elections before and not done as well on election day. If instead we assume that they will win the same level of support as in 2000 when they received 51.5% of the vote in Ontario, and that the NDP receives say 16%, the support actually registered in Ontario in an Environics poll released in October, then we find that the NDP would win 6 seats including both Ottawa Centre, where Ed Broadbent would defeat the Liberal (likely to be Martin insider Richard Mahoney) by 5 percentage points, and Toronto Danforth, where Jack Layton would defeat Liberal Dennis Mills by a narrow margin. This calculus, however, takes no account whatsoever of Broadbent or Layton's drawing power as individuals. It is based on overall shifts of support since the last election.
Although the NDP won neither seat in 2000, their fourth best performance in Ontario that year was in Toronto Danforth, and their sixth best in Ottawa Centre. While the NDP has not won either constituency since the 1980's, both ridings have been relative sources of strength for the NDP over the years. It should not come as any surprise then, that if the NDP picks up 15 or 16% in this year's election (a level close to the NDP's long-term average in Ontario) one should expect to see victories for the party in these two ridings.
I suggest if someone wants to bet some money on the two New Democrats' Liberal opponents, accept the wager - quickly.