The Alpha Break-Even Cost: Where Trading Edges Go to Die

A 12-1 momentum rotation strategy across SPY, TLT, GLD, and BIL returned 9.17% annualized over 19 years with zero trading costs. At 35 basis points per trade — roughly the round-trip cost of a retail ETF order with slippage — its entire edge over a static buy-and-hold portfolio disappeared. At 50 bps, it trailed buy-and-hold by nearly a full point of CAGR and destroyed 49% of the wealth the zero-cost backtest promised.

That is the alpha break-even cost: the precise per-trade cost at which a strategy's edge vanishes. Below it, you beat doing nothing. Above it, you would have been better off never trading at all. This is not a hypothetical. We ran the same momentum strategy at eight cost levels, from 0 to 50 basis points, and watched the edge dissolve in real time.

What We Tested

The strategy is a standard 12-1 momentum rotation: each month, hold the single best-performing asset from the prior 12 months (skipping the most recent 21 days) among SPY (US equities), TLT (long-duration Treasuries), GLD (gold), and BIL (Treasury bills). Rebalanced every 21 trading days. The universe spans four genuinely different risk premia — equity, duration, commodity, and cash — so momentum has something to rotate into when equities sell off.

We ran the backtest on split/dividend-adjusted daily data from June 2007 through August 2026 (19.2 years), at cost levels of 0, 5, 10, 20, 30, 35, 40, and 50 basis points per trade. One basis point is 0.01%; 50 bps means a $10,000 trade costs $50 in round-trip friction. For context, a commission-free ETF trade at a retail broker might cost 1-2 bps in spread, but realistic slippage on larger orders or less liquid ETFs easily reaches 10-20 bps, and 50 bps is what you'd see on a smaller ETF with poor liquidity or a strategy trading at scale.

The Numbers: Cost Eats the Edge

For reference, a static equal-weight buy-and-hold portfolio of the same four assets returned 7.19% CAGR with a 0.96 Sharpe and -17.4% max drawdown. The buy-and-hold Sharpe is higher because it never rotates — it just holds. Lower volatility, lower return, no trading costs to worry about.

The chart tells the story. At zero costs, the momentum strategy grows $1 to $5.40 — well above the buy-and-hold's $3.79. But as costs rise, the gap closes. At 35 bps, the momentum curve crosses below the buy-and-hold line. At 50 bps, the momentum portfolio ends at $3.25 versus buy-and-hold's $3.79. The strategy that looked brilliant at zero cost is now losing to doing nothing.

The Mechanism: Why Cost Drag Is Linear and Inescapable

The cost drag on a strategy is not mysterious. It is arithmetic:

Annual cost drag ≈ annual turnover × cost per trade

This strategy turns over 5.26 times per year — meaning 526% of the portfolio's value is traded annually. That sounds high, but it is what happens when you rotate 100% of the portfolio into a single asset each month and the signal switches. At 10 bps per trade, the annual cost drag is 5.26 × 10 = 52.6 bps, or about 0.53 percentage points of CAGR. At 50 bps, it is 5.26 × 50 = 263 bps, or 2.63 points of CAGR.

The backtested numbers confirm this almost exactly. The actual CAGR drag from 0 to 50 bps is 2.84 points — slightly more than the 2.63 the simple multiplication predicts, because costs also reduce the compounding base over time. The drag is linear in cost per trade but compounds in wealth terms.

Here is the critical insight: the 2.84-point CAGR drag does not mean you lose 2.84% of your money. It means you lose 49% of your wealth. Over 19 years, $1 grows to $5.40 at zero costs and $3.25 at 50 bps. The $2.15 difference is 48.8% of the zero-cost gain. Compounding amplifies a constant CAGR drag into a massive wealth gap because the drag operates on an ever-growing base.

This is why backtests that ignore costs are not just optimistic — they are structurally misleading. A 2-point CAGR difference looks small in a table. It is not small in dollars.

The Alpha Break-Even: 35 Basis Points

The most useful number in this entire exercise is the break-even cost: the per-trade cost at which the momentum strategy's CAGR equals the buy-and-hold's 7.19%. That crossover happens at approximately 35 basis points.

Below 35 bps, the momentum edge is real — you are being paid for the friction you endure. Above 35 bps, every additional basis point of cost is pure value destruction. The strategy's signal is still "working" in the sense that it is identifying the right asset to hold, but the cost of acting on that signal exceeds the benefit.

This is the number every systematic trader should compute before deploying a strategy: not "what is my Sharpe?" but "what is my break-even cost?" If your realistic per-trade cost is close to the break-even, you are one bad fill or one widening spread away from underperforming a static portfolio.

Why Low-Turnover Strategies Don't Care

Compare the momentum strategy to a static equal-weight portfolio rebalanced quarterly. That portfolio turns over just 0.05 times per year — barely any trading. At 50 bps per trade, the annual cost drag is 0.05 × 50 = 2.5 bps, which is 0.025 points of CAGR. The actual CAGR drops from 7.19% to 7.16%. Costs are irrelevant.

This is the tradeoff nobody talks about clearly: turnover is the multiplier on cost. A strategy that trades 100x more often pays 100x more in costs, even if the per-trade cost is identical. The question is not "is 10 bps expensive?" but "is 10 bps expensive for a strategy that trades 5x per year?"

The momentum strategy trades 100x more than the quarterly rebalance and pays 100x more in costs. The edge it generates must clear that bar every single year. In the zero-cost backtest, it clears it comfortably. In the 50-cost backtest, it does not clear it at all.

What the Literature Says

Garleanu and Pedersen, in their study of dynamic trading with transaction costs (Journal of Finance, 2013, vol. 68(6)), showed that the optimal response to trading costs is not to stop trading but to trade partially toward the target — "aim in front of the Markowitz portfolio." The insight is that costs create a friction that makes the full rebalance suboptimal. You do not jump to your target weights; you move partway, accepting some tracking error to save on friction.

Novy-Marx and Velikov, in their taxonomy of anomalies and their trading costs (Review of Financial Studies, 2016), tested a large set of factor strategies after accounting for transaction costs. Their finding: most strategies with monthly turnover above 50% lost statistical significance after costs. The strategies that survived were the ones with lower turnover or the ones that used a buy/hold spread — holding positions you would not actively buy into, to avoid the round-trip cost.

Carhart's classic study of mutual fund persistence (Journal of Finance, 1997) found that after accounting for transaction costs and fees, mutual fund "persistence" — the tendency of past winners to keep winning — largely disappeared. The edge was real in gross terms. It vanished in net terms. The same pattern shows up in our backtest: the momentum signal works, but the costs of acting on it consume the edge at a surprisingly low threshold.

Honest Caveats

• Survivorship bias. SPY, TLT, GLD, and BIL are all currently-listed ETFs. We are testing a universe of survivors. A strategy that rotated among a broader set of assets would include some that delisted or merged, which would lower returns. This inflates our numbers modestly.
• One period is not proof. This is a single 19-year sample that includes the 2008 crash, a decade-long bull market, a pandemic whipsaw, and a 2022 bond rout. A different period would produce different break-even costs. The mechanism (turnover × cost = drag) is structural, but the specific 35 bps threshold is sample-dependent.
• Costs are modeled, not experienced. We charge a flat cost per trade. Real-world costs vary by liquidity, order size, time of day, and market conditions. A strategy trading at scale faces market impact that grows with order size — a cost our model does not capture. The true break-even cost for a large portfolio is lower than 35 bps.
• No look-ahead bias. The engine uses the 12-month lookback with a 1-month skip, so the signal at each rebalance date is computed from data available before that date. The momentum signal is point-in-time correct.

Where This Meets Our Work

This is why RiskHarvest uses a portfolio-level rebalance trigger rather than a fixed calendar. Instead of rebalancing every month regardless of whether anything changed — which generates turnover and costs for its own sake — we rebalance when the portfolio drifts beyond a threshold that meaningfully changes the risk profile. The trigger is set at the portfolio level, not the individual asset level, because the question is not "has SPY moved 5%?" but "has the overall risk exposure drifted enough to justify the cost of correcting it?"

The break-even cost framework also informs our premia-decomposition test: before adding any holding to a portfolio, we ask whether it contributes a genuinely different risk premium or whether it just adds turnover. A holding that requires frequent trading to maintain but exposes you to the same equity risk premium you already own is a cost generator, not a diversifier. The 35 bps break-even is not a universal constant — it depends on the strategy's turnover and edge — but the method of computing it is. Every strategy we evaluate gets a break-even cost calculation before it enters the portfolio.

The One Thing to Remember

Every backtest has a break-even cost — the per-trade friction at which its edge disappears. If you do not know what yours is, you do not know whether your strategy works. The momentum strategy in this test looked excellent at zero costs and failed at 50. The difference between "alpha" and "noise" was 35 basis points. That is not a wide margin, and it is the margin that matters.

Educational only. Not financial advice. Past performance does not guarantee future results.

Sources

• Garleanu, N. & Pedersen, L.H. (2013). "Dynamic Trading with Predictable Returns and Transaction Costs." Journal of Finance, 68(6), 2309–2340. Journal of Finance · NBER Working Paper w15205
• Novy-Marx, R. & Velikov, M. (2016). "A Taxonomy of Anomalies and Their Trading Costs." Review of Financial Studies, 29(1), 104–147. NBER Working Paper w20721
• Carhart, M.M. (1997). "On Persistence in Mutual Fund Performance." Journal of Finance, 52(1), 57–82. Wiley Online Library