Rock Paper Scissors is a zero-sum game with a solved equilibrium. Every game theory course uses it as an example because the math is clean: three strategies, each beating one and losing to one, with a Nash Equilibrium at one-third probability for each throw. If both players are perfectly random, no one gains an edge. If either player deviates from random, the other can exploit it. That's the entire theory in three sentences.
The problem is that humans can't execute the theory. Ask someone to be random and they'll produce sequences with obvious structure — avoiding repetition, clustering certain throws, responding to outcomes in predictable ways. The Zhejiang University study from 2014 confirmed this with 72 players in controlled conditions: winners repeat, losers shift in a predictable cycle. The Nash Equilibrium holds in aggregate but collapses at the individual level into patterns that a paying-attention opponent can read and exploit.
This is why game theory is more useful as a frame than as a prescription. The equilibrium tells you the target — pure randomness is your ideal — but reaching that target requires active effort to suppress the responses your brain wants to make. After winning with Paper, there's a real psychological pull to throw Paper again. You have to override it consciously, and even then you're working against a tendency that years of reinforcement learning built into you.
The strategic layer above the equilibrium is exploitation. If your opponent is running win-stay lose-shift — the most documented human pattern in RPS — you can build a counter-strategy that gains an edge over multiple throws. The expected value shifts in your favor by a small but real amount. In a best-of-five, that margin probably won't be visible. In a longer session or across multiple tournament matches against the same player, it compounds.
The meta-game above exploitation is the part game theory doesn't model well: the interaction between two players who both know about conditional responses and are actively trying to predict each other's predictions. The recursion — I think you'll repeat, so I'll switch, so you should anticipate my switch and stay, so I should anticipate that and switch — has no stable terminus. At some point the computation cost exceeds the value, and the rational move is to commit to your highest-probability counter and accept the noise.
Tournament players tend to describe this as reading the room rather than running the math. You're not consciously calculating probabilities. You're picking up on cadence shifts, physical tension, the rhythm of your opponent's previous throws, whether they're varying their gambit or returning to a default. The game theory provides the structure. The actual competitive skill is the pattern recognition and the discipline to throw what your analysis suggests rather than what the moment makes you want to throw. Those two things are not the same, and the gap between them is where most tournament matches are decided.