Portfolio building tips

The Portfolio Builder has many settings, and it is easy to get buried in them. Before you start, decide what portfolio you actually want: which assets it should draw from—broad and multi-asset, ETF-only, a regional tilt—roughly how many assets it should hold, and whether it should adapt quickly to markets or trade rarely and cheaply. With that picture in mind, every setting has a job: the asset selection expresses the universe, the goals and the number of assets express the holdings, and the reactivity controls express how much turnover you are willing to accept. Without it, the settings are just knobs.

The fastest route to a good custom portfolio is rarely a blank page. Copy the pfolio pre-built portfolio closest to your risk level and modify the copy—the copy action gives you an editable version with a sensible configuration already in place. You then have a natural yardstick: compare your variation against the original, and keep your changes only if they earn their place.

A portfolio can only be as good as its asset universe. Two habits pay off. Spread the universe across several asset classes—correlation is generally low between classes and high within one, so a multi-class universe is what makes real diversification available in the first place. And keep the universe comfortably larger than the number of assets the portfolio holds: if the universe barely exceeds it, the "selection" is no choice at all, and the goals and ranking have nothing to work with.

A backtest that trades for free flatters reactive configurations. Switch transaction costs into the returns so every rebalance is paid for, then watch what the reactivity controls do to the bill: frequent rebalancing, short lookbacks, and zero inertia make a portfolio that rebuilds itself eagerly—and expensively. A configuration that only wins with costs off has not won.

The great pitfall of backtesting is tuning a configuration until its one history looks perfect—a portfolio fitted to the past rather than prepared for the future. Three defences. Judge the portfolio over several date ranges, not just the full backtest—a crisis year tells you more than a bull run. Judge it on drawdowns as well as returns—the drawdown view is closer to what holding the portfolio would have felt like. And be suspicious of fragile excellence: if a small change to a lookback or a goal weight flips the result, the result was luck, not design.

When a portfolio is not what you hoped for, resist rebuilding everything at once. Change one lever—a goal, the optimisation method, a constraint—then update and read the summary again; the levers and what each one does are mapped in Controlling the asset allocation of a portfolio. For bigger experiments, copy the portfolio and run the variant beside the original in a comparison—two configurations side by side beat one configuration edited back and forth.