Someone's range of holdings is often so wide that you sort of have to do it intuitively, but there are cases where you can narrow it down to a countable number of hands and make sharp plays based on that. I've said several times in other threads that I think that making the opponent's range as narrow as you can is one of the keys to going from a few % ROI heads-up to the double-digits.

Here's a hand I'm particularly proud of:

Full Tilt No-Limit Hold'em, 20+1 Tournament, 20/40 Blinds (2 handed) - Full-Tilt Converter Tool from FlopTurnRiver.com

Villain (SB) (t1260)
Hero (Button) (t1740)

Preflop: Hero is Button with A, 6
Hero raises to t120, Villain calls t80

Flop: (t240) 7, 5, 2 (2 players)
Villain bets t80, Hero calls t80

Turn: (t400) 7 (2 players)
Villain bets t160, Hero calls t160

River: (t720) 5 (2 players)
Villain bets t400

Hero, "I think you have a 2 and just got counterfeited."
Hero thinks.
Hero calls t400

Total pot: t1520

Results:
Hero had A, 6 (two pair, sevens and fives).
Villain had Q, 2 (two pair, sevens and fives).
Outcome: Hero won t1520


-----

Let me go through that hand street by street and explain why I made the call.

My preflop raise doesn't cut his range down very much. Takes out 72o, 82o, 83o, maybe 92o... that's about it. He was calling almost every time.

Him leading out on the flop is what really cuts his range down a lot. He might have air, but the bluffs this guy had made so far were more calculated than that... he wasn't bad, and I didn't think he'd lead out on this flop just because he thought I'd missed it.

I thought he'd probably check-raise an overpair, even a small one, and definitely check-raise or check-call with two pair, so most likely his range is any 2, 5 or 7, or a flush draw. Flush draw a little less likely, as I kind of felt he'd check-raise or check-call that as well.

Second barrel on the turn definitely feels like a made hand more than a flush draw. I still flat, even though it sort of gives away my hand, because my odds are good enough. He won't pay me for the flush, but he can't know my precise cards, so I might get a bit of value out of those by just calling a blocking/value bet if I make a pair.

River is verrrrry heeeeeeenteresting. His bet is too big, as it lets me remove 5 from his range - if that was his hand, he wouldn't want to bet so much that he'd feel compelled to call if I pushed. Maybe a donk would bet like that, but I didn't think this guy would. That leaves a 2 or a 7, but my instinctive guess was that a 2 was 50% more likely.

He folded 32-92o preflop, and wouldn't lead the flop with 52. A2 would probably check and hope the A kicker is good. That leaves:

32,42,62,72,82s = 5x3 = 15 hands - 1 (62 clubs, since I have 6c) = 14
T2-K2 = 4x12 = 48 hands

62 hands I beat.

Pessimistically, let's say he's playing 73-74s this way, and any bigger 7, but not 75.

73s-74s = 2x2 = 4 hands
76,78,79,7T,7J,7Q,7K,7A = 8x8 = 64 hands - 2 (Ac7x, since I have the Ac) = 62 hands

66 hands that beat me.

----

That makes it seem my call is more like 50/50, which is still good to make as I'm getting almost 3-1 odds. However, I'm not including the (smaller) possibility that he's bluffing a missed flush draw, which tips things a little more in my favour. I say small because his bet on the turn didn't feel like a draw, and because a busted flush draw is bluffing less often than a 2, simply because the 2 is more frustrated.

Anyway, doing this kind of analysis to check your hunches is good.