If you sorted a strategy library by Profit Factor and picked the top eight, would that give you the best eight-strategy portfolio available? It’s a natural instinct, and it’s wrong more often than it feels like it should be. Individual strategy quality and portfolio contribution are two different questions, and collapsing them into one is probably the single most common mistake in systematic portfolio construction.
Two different questions, easy to confuse
The first question — is this strategy good? — is answered by looking at a strategy on its own: its Profit Factor, its Expectancy, its Return-to-Drawdown, its sample size, its Stability. These are honest, useful measures of standalone quality.
The second question — does this strategy make the portfolio better? — can only be answered by looking at the strategy in relation to everything else already in the combination. A strategy can score well on every individual measure and still be a poor addition, because the question isn’t whether it’s good; it’s whether it adds something the portfolio doesn’t already have.
Where the two diverge: overlap and duplicate behaviour
The most common way a strong standalone strategy turns into a weak portfolio addition is overlap. If a portfolio already contains three trend-following strategies on correlated instruments, adding a fourth excellent trend-following strategy on a related instrument doesn’t diversify the portfolio — it concentrates it further. All four are likely to win together, lose together, and drawdown together, because they are, in effect, four versions of a similar bet.
This is easy to miss because nothing about the fourth strategy’s own metrics reveals the overlap. Its backtest can look completely clean. The problem only becomes visible once you look at how it interacts with the strategies already in the combination — which requires evaluating the portfolio, not the strategy.
Concentration hides behind aggregate numbers
A related trap: a portfolio can look well-diversified in aggregate — spread across several symbols, several timeframes — while a small number of highly-correlated strategies quietly drive most of its risk and most of its return. Adding “the best available strategy” repeatedly, without checking its relationship to what’s already there, tends to produce exactly this pattern: broad-looking coverage sitting on top of a genuinely narrow behavioural base.
Portfolio-level metrics tell a different story than individual ones
This is why portfolio-level metrics are calculated from the portfolio’s own combined chronological trade stream, rather than from averaging its member strategies’ individual numbers. Two candidate eight-strategy portfolios, built from strategies with identical average individual Profit Factors, can have meaningfully different combined drawdowns, combined Return/DD, and combined Stability — because those combined figures depend on how the trades actually line up in time, not on the average quality of the ingredients.
Why Portfolio Drawdown Can Matter More Than Individual Strategy Drawdown explores this specific effect in more depth. The short version: a portfolio’s real risk profile is an emergent property of the combination, not a simple sum of its parts.
“Best eight strategies” is not “best eight-strategy portfolio”
Put plainly: ranking a strategy library and taking the top N by any single individual metric answers a strategy-quality question, not a portfolio-construction question. It can accidentally produce a strong portfolio if the top strategies happen to behave differently enough from one another — but that outcome is coincidental, not designed. A deliberately constructed portfolio instead asks, for every candidate addition: given what’s already here, does this strategy improve the combined result, or does it mostly repeat what’s already represented?
What to check instead
- Individual quality as a first filter — a strategy still needs to be credible on its own before it’s worth considering.
- How the candidate’s trades line up in time with the strategies already in the portfolio.
- Whether the candidate meaningfully changes the portfolio’s diversification across market, direction, timeframe and behaviour, not just its symbol count.
- The portfolio’s own combined metrics before and after adding the candidate — not the candidate’s standalone metrics in isolation.
- How often the candidate’s underlying strategy already appears elsewhere across your retained combinations, as one signal of how distinctive it actually is.
Where this fits in the SIPS workflow
This is the central idea SIPS is built around. Portfolio Comparison presents every retained candidate using genuine combined-portfolio metrics rather than averaged member figures, and every portfolio carries a Bank Reuse reading showing how often its member strategies also appear elsewhere in your retained bank. Portfolio Analysis then drills into one specific portfolio’s own diversification and contribution in detail. None of this replaces judgement — it gives you the combined evidence the judgement should be based on.
The practical takeaway
Strong individual metrics are necessary, but they are a filter, not a selection process. The question that actually builds a good portfolio is a comparative one — not “is this strategy good?” but “given everything else already here, does adding this strategy make the combination better?” Those two questions can have very different answers, and it’s the second one that determines what the account actually experiences.