Where trading ideas come from (video)

Most people start hunting for a trading setup and only later ask what market mistake it is supposed to exploit. The better order is the reverse: start with a repeatable market behavior, then ask what rule could capture it. (youtube.com) Algorithmic trading is just computers placing and managing orders by rule, and the United States Securities and Exchange Commission says those rules now sit inside market structure, data, communications, liquidity, and execution. A trading idea has to survive all of those layers before it is worth real money. (sec.gov) One place ideas come from is market microstructure, which is the study of how an exchange actually works minute by minute. That field looks at quotes, spreads, price discovery, intraday behavior, and transaction costs, which is why tiny details like queue position can matter more than a chart pattern. (sciencedirect.com) Another place is cross-sectional trading, which means ranking many assets against each other instead of guessing whether one market goes up or down. A standard cross-sectional process sorts stocks by a characteristic known at the time, builds portfolios from the ranks, and checks whether the spread between the top and bottom groups persists without look-ahead bias. (amoreira2.github.io) A third source is event-driven trading, where the idea is not “this stock looks cheap” but “this corporate event has a priced probability that may be wrong.” GMO describes that universe as opportunities created by mergers, spin-offs, restructurings, litigation, and regulatory decisions where the outcome becomes known relatively soon. (gmo.com) A fourth source is structural friction, which is the set of costs and constraints that stop markets from behaving like frictionless textbook machines. Nature’s research summary lists fixed fees, proportional expenses, liquidity effects, and no-trade regions, all of which can turn a good-looking backtest into a losing live strategy if you ignore them. (nature.com) That is why the useful question is rarely “what indicator should I add,” and more often “what mechanism forces this pattern to exist.” If the answer is “I saw it once on a chart,” the idea usually dies when costs, crowding, or a regime change shows up. (youtube.com) The video’s funnel follows the same order good research teams use in practice: observation first, mechanism second, data third, then test design, failure modes, and execution. That sequence stops people from fitting a rule to noise the way a student might memorize exam answers without understanding the subject. (youtube.com) Interviewers in systematic trading often probe exactly that chain. They want to hear where the signal came from, why the market should pay for it, what dataset would measure it, how you would avoid look-ahead bias, and what real-world cost could erase it. (youtube.com) The point is not to collect more setups. The point is to build an idea factory that can keep producing hypotheses from exchange mechanics, relative rankings, corporate events, and trading frictions long after one specific trade stops working. (youtube.com)