The Quick-and-Dirty First Pass

Most research ideas are wrong, or at least not good enough to trade, and the fastest way to find that out is not to build the idea properly first. A quick-and-dirty first pass means running the roughest version of a test that can plausibly answer "does this have any signal at all" — a crude signal computed with a spreadsheet-level implementation, tested on one liquid instrument, with no transaction costs, no careful point-in-time data, and no parameter tuning. If the result isn't at least directionally interesting under these forgiving conditions, it almost never becomes interesting once you add back all the realism that real trading requires.

This is a deliberate triage step, not laziness: careful, production-quality backtests are expensive to build, and running that full machinery on every idea a researcher has in a week would leave no time for anything else. The first pass exists purely to sort ideas into "worth the real work" and "move on," and researchers should expect to throw the quick version's code away rather than polish it into the real backtest.

The failure mode is skipping straight to a full, rigorous backtest for every idea — technically more correct, but far slower, so fewer ideas ever get tested per unit of researcher time.