Quants and AI eating alpha

A lot of the easy intraday money used to come from humans being slow. Now the competition is code reading five-minute bars, classifying market “regimes,” and firing orders before a discretionary trader has finished the thought. A Markov model is just a probability machine for “what state are we in now, and what state usually comes next.” In trading, that means labeling a market as trending up, trending down, or drifting sideways, then betting on the next tiny move instead of waiting for a big one. That works best in the most liquid names, because liquid markets give bots a constant stream of fresh prices and enough counterparties to get in and out fast. QuantConnect’s own intraday example uses five-minute returns on the 10 largest United States stocks for exactly that reason. Once enough firms do this, the old edge gets thinner. The Securities and Exchange Commission said in its report to Congress that algorithmic trading can improve liquidity and execution, but it also changes market quality, volatility, and arbitrage dynamics as more activity gets automated. That is why the new social-media clips of bots flashing daily gains matter less as proof of genius than as proof of crowding. If many systems are chasing the same one- or two-tick mispricing, the market starts repricing those micro-opportunities away, the same way ride-share surge pricing disappears when too many drivers rush to one block. The plumbing of the market is already built for this race. The Bank for International Settlements said fast-paced electronic markets have become more automated, more fragmented across venues, and more dependent on data and execution technology, which is exactly the terrain where short-horizon bots thrive. Regulators are reacting to the speed, not just the profits. The European Securities and Markets Authority published a supervisory briefing on February 26, 2026 that tells firms to tighten governance, testing, outsourcing controls, and pre-trade controls for algorithmic trading systems, including systems that use artificial intelligence. The United States watchdogs are reacting to the marketing too. On January 25, 2024, the Securities and Exchange Commission, the Financial Industry Regulatory Authority, and the North American Securities Administrators Association warned that scammers were using “artificial intelligence” claims to sell unregistered platforms and impossible return promises. So the real story is not that a bot can post a flashy green screenshot. The real story is that short-term alpha is being compressed into smaller, faster, more competitive scraps, and anyone selling “guaranteed” machine profits into that arms race is colliding with both math and regulators.