Just for fun (I already upvoted David's answer), I'll take the above lines to show in praxis where I cut a calculation which I do for CSE. (In RL, I hardly calculate at all.) For the record, 1...Bc4 is about -5 to -6, 1...Rxc3 2.Kd2 is about equal, 1...Rxc3 2.Rd7 Qa3+ is -2.
a) 1...Bc4. Immediate cut, since it matches the motive I already spotted - double deflection against Rd6. Maybe a glance at 2.Re5 Qxd6 3.Re8+ (White has no other counterattack) but it is obvious White suffers from the same problem, only worse.
b) 1...Rxc3 2.Rd7 Qa3+ 3.Kd2. Now 3...Rxc2+ is the obvious candidate, and even if you must now calculate a handful of consecutive moves, this is easy as White's hand is forced: 4.Nxc2 Qc3+ and immediate mate, or the likewise insufficient 4.Ke1 Rc1+ (if that comes out as not working, I can always switch to Qc1+) 5.Kf2 Rf1+ 6.Kg3 Rg1+ 7.Kh2 Rg2+ 8.Kh3 f2+. Game over, the new Q protects f7 if necessary. wK somewhere on first rank will suffer from Rcc1+. In this line, you should calculate until 8...f2+ and its consequences (9.Nb3 f1Q 10.Qxf7+ Qxf7).
c) 1...Rxc3 2.Rd7+ Qa3 3.Kb1. Ech, Rb8+ is impossible. Nevermind, 3...Rf8, cut. The Pf3 is still alive, especially as White must play Nb3 to survive. I let the position appear on board and ponder on when it occurs, as there are far too many candidates (and the computer wants the modest Rcc8.)
d) 1...Rxc3 2.Kd2. Too much chaos for sensible calculations. The computer suggests an equal endgame after 2...Qc7 3.Rd8+ with a lot of exchanges. Fine, I win all endgames anyway :-)
To sum up: There are variants where you can and should calculate five moves or even longer - that is if the game tree doesn't branch to hell and back. And, as David already stated, there are positions no human can calculate to the end, and positions that don't really need concrete calculation.