Two days into Battle of the Bots, the thing I found most interesting wasn't which agent was winning. It was how fast the two of them stopped looking like variations on the same system and started looking like genuinely different operators.
Teresa, the momentum agent, behaved the way I expected momentum strategies to behave — quick to enter on strength, quicker to cut losses, visibly reactive to short-term price action. Omar, the long-hold agent, did the opposite almost immediately: fewer trades, wider tolerance for drawdown, a kind of patience that looked, if I'm honest, a little unsettling to watch in real time with real money attached.
Neither behavior was a surprise on paper — that's what the strategies were designed to do. What surprised me was how quickly the design turned into something that felt like temperament. Watching two systems I'd built make opposite decisions off the same market data, in real time, made the abstractions in the strategy design suddenly feel a lot more concrete.
The other early observation: the interesting information wasn't in either agent's individual P&L. It was in the gap between them — where they agreed, where they violently disagreed, and what that disagreement implied about the actual market conditions underneath both of their decisions.
This is still Day 2 thinking, and Day 2 thinking is allowed to be wrong. But it's already reshaped how I think about autonomous systems generally: a single agent's behavior tells you about the agent. Two agents' divergence tells you about the environment.
Current Hypothesis
Two agents with opposite, internally consistent strategies will diverge in behavior almost immediately — and the divergence itself is more informative than either agent's individual performance.
Not a conclusion — a working idea, revisited as evidence comes in.