Abstract. Net return is decomposed into four forces by analogy with powered flight: thrust, drag, gravity and steering. Table 1 defines them. Backtest telemetry from 11 strategies with complete cost accounting is mapped onto the four, and Table 2 reports the measurements.
1. The decomposition
Powered flight balances four forces, and net return admits the same accounting. A positive gross edge is neither necessary nor sufficient for a positive net result, and drag ends more strategies than weak signals do.
| Flight force | Portfolio force | Where it is measured here |
|---|---|---|
| Thrust | Gross edge, before costs | backtest gross return |
| Drag | Friction: costs × turnover, slippage, borrow | 5bps per unit turnover + 25bps/yr borrow |
| Gravity | Volatility drag: growth ≈ return − σ²/2 | the $100k→$93k plaques |
| Steering (control inputs) | Trading: every course change burns fuel | annual turnover × cost per trade |
2. Measured forces
Cost drag comes first. Only about a third of the sample produced positive gross thrust, and among those the friction ratio, cost drag as a share of gross thrust, took nearly half the edge before any return reached the account. Stall speed, the cost per unit of turnover at which net return falls to zero, sits above the model’s generous assumption and inside the range of realistic institutional costs for less liquid names. Those strategies fly under the assumption, not under real friction.
Every rebalance pays the friction toll, so steering authority has to be budgeted like fuel. The binding constraint is the signal half-life check: a signal that decays faster than its rebalance period pays turnover to chase noise rather than a persistent edge, and that check is a standard at stage 4 of the pipeline.
Geometric growth is approximately the arithmetic return minus half the variance, so a positive average return can still compound to a loss, and one strategy in the sample did. Leverage scales thrust and volatility drag together, the mechanism behind structural failures under load such as Long-Term Capital Management (LTCM), documented separately in the Case Studies section of this site.
| Measure | Value |
|---|---|
| Strategies with complete cost accounting | 11 |
| Share with positive gross thrust | about a third |
| Median cost drag, share of gross thrust | 42% |
| Median stall speed, breakeven per unit turnover | about 12 bps |
| Cost model assumption | 5 bps per unit turnover |
| Realistic institutional cost, less liquid names | 10 to 20 bps |
| Borrow charge in the cost model | 25 bps/yr |
| Median annual turnover | approximately 10× |
| Highest turnover in the sample | approximately 19× |
| Volatility drag | growth ≈ return − σ²/2 |
| Positive average return, compounded | $100k to $93k over twenty years |
| Passive benchmark profile | thrust near zero, drag single digits of bps, turnover near zero, high return to cost |
3. Verdict
A passive index fund is the limiting case, profiled in Table 2: compounding does the flying. Most active strategies here carried large gross thrust and equally large drag and turnover, and most did not beat that benchmark once their own trading costs were charged. The bar is explicit. Gross thrust net of drag and volatility drag must exceed the passive benchmark, at realistic friction and out of sample.