Questions about ICM

I recently started playing around with ICM and have a couple questions. This will be a bit detailed.

Why are we adding together the (win * new prize pool equity), (loss * new prize pool equity), and (tied * new prize pool equity) when calculating its effect on our prize pool equity? My question is more about adding the loss column. More specifically, I'm thinking that if a hand wins more often, then it also loses less often, so the difference between hands is minimal.

Say your hand would win 62% and lose 36%. And winning gives you PE (prize pool equity) of 40% and losing gives you PE of 20% (everyone's pretty much moderate stacks, give or take). That gives you (.62 * .40) + (.36 * .20) -> .248 + .072 -> .32

Now say your hand is slightly better, winning 70% and losing 28%. Same winning and losing PE. Now you have .336, only a difference of .016.

32% with the worse hand and 33.6% for the better hand.

And then let's say your fold PE was 31.5%. Calling would only be .5% better in the first case, and only 2.1% in the second case. If I recall correctly, we clearly should be calling if we can gain 5% PE, but not if we only gain 2%.

I've read the post by dethgrind a few times, but he doesn't explain this part. Any assistance would be great.

[wildzer0]

I'm kind of confused by your examples here. Maybe if you could give an example that included stack sizes, blind sizes, etc. it would make more sense.

In general though, if I'm understanding your question correctly, I push with a .5% equity increase or better. Often the decision is close but it's always best to go with the decision that has the higher EV. Again, I'm not quite understanding your example, are you calling a push or pushing yourself? If you're pushing yourself, one extremely important number you're missing is how often your opponent will fold to your push.