Harrington did not by any means invent this play, but may have been the first to write about it in a published work. He also pulls off the most well-televised example of it as well, in a hand at the 2004 WSOP main event final table where Josh Arieh raises thin, Greg Raymer cold-calls thin, and Harrington pushes with 62o and makes them both lay down.
I've waffled back and forth as to whether this is a good play or not. On the surface it's very attractive. Let's say it's the 100-200/25 round. Suppose there is a raise to 700 and a call, and it is folded to you in the BB. You have 2500 in chips (after posting). If you raise all-in, you are risking 2500 to win the current pot of 1950. Suppose the first guy will call only with AA-99, AKs, AK, and AQs. Suppose that his opening range was something like AA-55, AKs-A9s, KQs-KJs, AK-AT, KQ. So there's a 56/148 = 37.8% chance he'll call. Suppose the second guy cold-called with the same hands the opener raised with (for simplicity). He is also never trapping, and will only call with 99-77, AQs, AQ (all the better hands, he would have re-raised with). His chances of having a calling hand are only 23%. So (1-.378)*(1-.23)=47.9% of the time, you win the pot without a showdown! Since if you have a trashy hand like 74s, you are less than a 3:1 dog for the money in the pot, this should be a huge ev play under these circumstances.
Now, I've assumed some really tight players here. I think people will call much thinner than this in real life. So I don't know if the play is profitable, even in a vacuum (i.e. assuming there is not the problem of people catching on to this play), because I don't know enough about what hands people raise with, cold-call with, and call jams with.
I suspect that, at least in a vacuum, the play has value in certain spots. But what I actually started thinking about yesterday (because there were so many open-raises and cold-calls at my table) was not the expectation of the play, but rather the set of hands with which you would make it, if you knew you wanted to make it.
Suppose I can break down hands into just three groups. The first group is comprised of the set of hands that are very strong and I am happy about getting all-in with. The second group is comprised of hands that I wouldn't be happy about getting all-in with, but are good enough hands to call and see a flop. The third group is comprised of hands that are not strong enough to call and see a flop. Then it seems that you move in with hands from the first group and part of the third group, and call with those from the second group.
But which hands from the third group do you select? What kind of surprises me from when I see people making this move, is that people seem to think that the worse the hand in this third group, the better it is to move in with. That can't be right. If the move has positive expectation with 62o, then it's almost certainly even more positive with T7s. (Btw, I don't necessarily think that Harrington thinks this; he didn't choose to be dealt 62o, and had he been dealt T7s instead he probably would have played it the same.)
So here's my question. Is the set of hands you select for the squeeze play selected solely from the top of the distribution of hands that you are not willing to call and see a flop with, or does it look different from that?