In evolutionary game theory, stronger strategies should dominate. A strategy that extracts more payoff from interactions should outcompete one that extracts less. Wang et al. (arXiv: 2603.20998) find a parameter regime where the opposite happens.
Their model has three strategies in a donation game with bias: unconditional cooperators, defectors, and tit-for-tat players who condition their behavior on the type of partner. The bias parameters control how much each strategy favors or penalizes different opponents. Under certain parameter conditions, a “hidden T phase” emerges where the objectively weaker strategy achieves dominance.
The mechanism is non-transitive. Tit-for-tat suppresses its own performance against cooperators while simultaneously eliminating the cyclic dominance pattern that would normally prevent any single strategy from dominating. The strongest strategy weakens itself in a way that clears the field for the weakest. Spatial structure is essential — the effect requires local interactions on a lattice, not well-mixed populations.
The through-claim: competitive suppression can be self-defeating. The strategy that is best at eliminating rivals may, in doing so, create the conditions for its own displacement. Tit-for-tat doesn't lose in direct competition; it loses by winning too specifically, eliminating the very opponents whose presence kept the weakest strategy in check. The competitive landscape reshapes itself around the winner's success, and the reshaped landscape favors someone else.
Wang, Li, Wang, Zhu & Szolnoki, 2603.20998. Evolutionary game theory / cooperation / spatial dynamics / non-transitive competition.