Strategy style
Trend-following futures strategies: what the evidence says
A trend-following strategy enters in the direction a market is already moving and stays in while the move continues, accepting many small losses for a few large wins. It works only if moves persist more often than chance allows, and that is something the data can measure at every horizon.
Does trend persist?
It depends on the horizon, and the record has measured two of them. On daily bars, return autocorrelation at lags one to ten is indistinguishable from zero on every major contract tested: the last ten days' direction carries no measurable information about the next. About three quarters of the index contracts' long-run drift also accrues overnight, so a long-biased multi-day trend system collects market exposure that a backtest without a benchmark reports as skill.
Within the day the picture changes by market. Crude oil trends intraday, though not from one bar to the next, and the index contracts trend where natural gas tends to revert. That is why a trend rule has to let a winner run: a screen can hold one to the bell, or across sessions when the rule says so. A position held overnight or over a weekend can open well beyond its stop. The screen fills that stop at the opening price, and a real order can fill worse, so a loss can be larger than the stop you set.
Three habits the record says to drop
Adding to winners. Pyramiding lost to constant size at matched risk in all 42 schemes tested, because each added contract earns less per unit of risk than the first while moving almost perfectly with it.
Letting the optimizer choose the stop. A walk-forward that scores average R keeps choosing the tightest stop on the grid, because R is defined by the stop; it finds a smaller denominator, not a better stop. Score in dollars at constant risk instead.
Trusting one era. A book whose profit sits almost entirely in its last few years is a hypothesis about those years, not a strategy. Every result on the ledger is split by era before it is believed.
Testing a trend-following idea
The screener tests trend ideas within the day and across days: a moving-average cross, a channel break or an opening-range or prior-day break, on bars from one minute to a week, entered at the market or on a stop, and held until its stop, its target, a trailing stop, a time limit or a condition ends it, overnight and over weekends when the rule says so. A position held overnight or over a weekend can open well beyond its stop. The screen fills that stop at the opening price, and a real order can fill worse, so a loss can be larger than the stop you set.
From the research record
- SettledDaily-bar momentum on index futuresReturn persistence at the daily horizon reads zero, and most of the index drift accrues overnight.
- SettledAdding contracts to a position that is already winningEvery add earns less per unit of risk than the first lot and is correlated with it. Across forty-two schemes, flat sizing came out ahead.
- SettledLetting walk-forward optimization choose your stopIt gravitates to the tightest stop whenever the objective restates the parameter, so score in dollars at constant risk instead.
- SettledA strategy that only works on the 30-minute chartHold the entry instant and the stop distance fixed and an 18x span of bar sizes pays the same, which makes the strategy far more robust than it first appeared.
- SettledAn edge that only exists in the last three yearsNinety-eight percent of a six-stream book's lifetime profit sat after 2022, which is why every result is now split by era before it ships.
Trend following questions
- Does trend following work on futures?
- At the daily horizon the record finds no persistence to follow: return autocorrelation at lags one to ten is indistinguishable from zero on the major contracts. Within the day some markets trend and some revert, and the answer depends on the market, the time of day and the costs. Test the specific rule rather than the style.
- Why do trend-following strategies have low win rates?
- Because they cut losers quickly and let winners run: many trades stop out small while a few catch the move. A win rate below one half is normal for the style. What matters is that the average winner outweighs the average loser by enough to clear the costs.
- Should I add to a winning trend trade?
- On the research record, no. Forty-two pyramiding schemes were tested against constant size at matched risk, and constant size won every one: the expected follow-through after a position is already a quarter of an ATR in profit is smaller than the entry's own edge, and it keeps falling the further the move has run.
- What is the best timeframe for trend following?
- Timeframe comparisons are usually confounded. One strategy's result swung 79% across bar sizes until the entry instant and the stop in points were held fixed; after that, an eighteen-fold span of bar sizes paid about the same. Choose the timeframe for the mechanism, then check the result is not an artefact of the bar.
Screen a trend-following idea
Describe it in a sentence. The desk turns it into an exact specification and screens it through seven gates written down before the run, from real per-fill costs to a drift null and an era split. The verdict usually comes back in 15 to 25 seconds.