Frictions all the way down: the implementation gap, layer by layer

The gap nobody prices

A strategy’s reported Sharpe lives at the strategy layer: daily total-return series, frictionless weights. A retail account lives three layers down — it must pick an instrument, and it must hold a whole number of contracts in a finite account. Most write-ups quote the top layer and wave at the rest. A complete audit measures all three.

Three altitudes, same strategy

• Strategy (ETF total-return data). A long/short time-series-momentum sleeve nets an excess Sharpe of 0.57 with −0.04 correlation to SPY. This is the number a clean backtest reports.
• Instrument (CME futures). Replicating the sleeve in futures tracks the ETF version with strict-pair correlations of 0.90–0.97. The wrapper costs something, but it is small and measurable — not a hand-wave.
• Execution (integer contracts, ~$500K account). Rounding prescribed weights to whole contracts produces an annualized tracking error of 327 bp against the frozen model.

Where the gap actually comes from

The 327 bp does not decompose the way intuition expects. Only ~110 bp is rounding plus commissions — the part everyone worries about. The larger ~355 bp is instrument basis: micro contracts drifting against their full-size parents. The retail implementation story is decided by tracking error, and tracking error is dominated by basis, not by the rounding you can see on the order ticket.

The lesson

Report the implementation gap; don’t bury it. A single headline Sharpe is a claim about the top layer only — the deliverable is what survives the descent to an order ticket. Measuring each step, and naming which step costs the most, is the difference between a research result and a sales number. It also redirects effort: if basis dominates rounding, the lever that matters is instrument choice, not shaving a few contracts.