A day trade that follows the day's own tide

• Has the stock turned up? A line called the Supertrend follows the price at a distance set by how much it normally moves, and sits above the price in a fall and below it in a rise. When the line flips from above to below, the stock has turned.
• Are the buyers winning today? VWAP is the day's average price weighted by how much traded at each level. A price above it means more of today's business was done lower than here, which is to say buyers have had the upper hand since the open.
• Is there a trend at all? The ADX is a score of how strongly a price is moving in one direction, regardless of which. Below a threshold, the flip was a wobble, not a turn.

Three yeses and it buys. It sells when the Supertrend flips back, which is its main exit, with a stop below and a target above to bound the trade in between, and everything is closed by a quarter past three.

It lost. Not dramatically, and not because the idea is wrong about trend days, which it finds. It lost because the trades are small and the charges are not. An intraday trade in a large stock that goes well earns a fraction of a percent of the account. The brokerage and taxes on the way in and out take a meaningful slice of that, and on the trades that stop out they take the slice anyway. Over the window, about half of what the strategy lost was charges. The other half was the choppy days it could not tell apart from trend days until it was already in.

The idea finds trend days. The charges find every day. Over the test, the second was the larger effect.

Fewer trades or larger moves. The first is a stricter filter, a higher ADX bar that lets through fewer and better days, at the price of many days with nothing to do. The second means instruments that move more within a day than a large-cap share does. Either is a design change, and we would rather list nothing in this family than loosen a threshold until the test turns green, because a threshold tuned to the past is the most reliable way to produce a strategy that only ever worked there.

An earlier version of this strategy had a different problem, worth mentioning because it is the kind of thing a test catches. It treated 'the Supertrend is up' as the entry instead of 'the Supertrend has just flipped up', and so it bought again on every five-minute bar the trend held, paying brokerage on each. More than one order a bar across a dozen stocks. The rule now enters on the flip itself and nowhere else. The result still did not pass, but it fails for an honest reason now.