I'm pretty sure ICMizer looks up how close your specific opponents play to nash, it probably figured out that they are too tight in this spot so it recommends taking a huge(!) -edge. It also has statistics about how often your stack size is going to cash (and stacks when you double), and recommends that you should gamble. But there are also a bunch of other factors why I think this is a great spot to push wide.
1) We are the shortest stack
2) We have to pay the blinds soon
3) We can't push that much next turn
4) We surprisingly don't actually gain that much when we fold, and another player is knocked out, same for next turn (checked this with an ICM calc).
5) The value of doubling up is huge because then we are 2nd in chips (1st if BB calls), and our play in the blinds and future game is much more profitable.
6) Unlike the other players, if we double up we around double our equity in the tournament.
7) We actually benefit more than any other player if we steal the blinds. It also makes our BB/SB play more profitable, because people are going to be pushing less into us.
Our Equity: 11.846%
Win blinds: 13.480%
DBL via BU: 23.894%
DBL via SB: 23.606%
DBL via BB: 22.518% (but we are chip leader)
All other players except the BB have a reasonable BF against us, so they wont be calling too often. The BB has a low BF, so he is calling quite often, but if he does call and we win, we are the chip leader. If you look in HR calc the BB nash calling range can be (9%-23%), it's really flat for those ranges. Also our best response here is really flat too, any strategy from pushing 9% to 50% is pretty much the same (-0.03 to push 50%).
Part of the reason I think this is a great spot to push really wide is that a lot of people seem to think it's not, so in general opponents wont put you on a wide range. It really doesn't matter to much since the BB can call with 9%-23% and it doesn't really matter to us.
So I'm not pushing ATC here, it's possibly a good strategy though. Personally, I think it's optimal to be pushing 50% here though.