2002-2026 Results Summary
01 — Universe Construction
Why Universe Selection Is Itself a Source of Alpha
Before a single factor signal is applied, the strategy operates within an intentional and disciplined asset universe that is a meaningful contributor to realized performance. The investable universe is drawn exclusively from constituents of the S&P 500, NASDAQ, and Dow Jones Industrial Average, filtered to retain only highly liquid securities. This is not a passive or arbitrary constraint — it is a deliberate design choice with direct performance and risk implications supported by substantial academic literature.
Quality as a Structural Edge
Restricting the universe to S&P 500, NASDAQ, and DJIA constituents simultaneously enforces two independent structural advantages: liquidity and quality. On the liquidity dimension, the academic literature on momentum strategies is unambiguous: transaction costs and slippage are the primary reason most momentum strategies fail to deliver their theoretical returns in practice. Lesmond, Schill, and Zhou (2003) famously argued that trading costs eliminate momentum profits entirely in less liquid market segments. AQR's 2017 paper "Implementing Momentum: What Have We Learned?" demonstrated that institutional execution in large-cap equities reduces real-world costs to just 0.15–0.35% per trade versus the 1–2% assumed in early academic models — an advantage that only accrues to strategies confining themselves to the highest-liquidity names. For less liquid stocks, AQR found costs remained 3–7x higher at 0.5–1.0% per trade, sufficient to materially erode or eliminate momentum profits at monthly rebalance frequency.
On the quality dimension, index membership in the S&P 500, NASDAQ, and DJIA requires meeting ongoing minimum standards for market capitalization, profitability, trading history, and corporate governance — thresholds reviewed and enforced by index committees continuously. This structural screen eliminates distressed companies, firms with insufficient operating history, and names with outsized idiosyncratic risk — categories that disproportionately contaminate academic backtests but are impractical or dangerous to hold in real portfolios. Research from S&P Dow Jones Indices illustrates this effect quantitatively: the S&P SmallCap 600's quality-eligibility screen produced meaningfully superior long-run returns over the Russell 2000 despite both indices covering the same nominal universe — demonstrating that index-based quality filters generate measurable alpha entirely independent of factor tilts.
Together, these two dimensions ensure the strategy operates exclusively in the most liquid, highest-quality segment of the equity market — where momentum signals are most reliable, execution costs are lowest, and the gap between theoretical and realized factor returns is smallest. The combination of quality and liquidity filters also produces a return profile with some of the characteristics associated with a small-cap tilt in academic factor models: by concentrating in high-momentum names that have meaningfully outperformed their large-cap peers, the portfolio naturally skews toward the more dynamic, higher-growth segment of the large-cap universe — generating the statistical correlation with small-cap indices visible in the OLS regression without carrying any of the liquidity risk, governance concerns, or execution costs of actual small-cap investing.
Alpha Architect's Quantitative Momentum framework, one of the most rigorously constructed implementations in the practitioner literature, explicitly restricts its universe to the largest 1,500 liquid U.S. exchange-traded stocks for precisely this reason — noting that including smaller, less liquid names produces "phenomenal backtest returns that may be unobtainable in the real world, even when operating with small amounts of capital." The strategy's liquidity and quality filters enforce an equivalent discipline while adding the further governance screen of index membership.
Transaction costs and slippage have been explicitly modeled in the backtest, ensuring that all reported performance figures reflect realistic net returns rather than theoretical gross performance.
02 — Performance Record
24 Years Across Every Major Market Regime
The strategy's backtest spans January 2002 through March 2026 — a period encompassing the dot-com recovery, the Global Financial Crisis, the post-QE bull market, the COVID crash and recovery, and the 2022 rate shock. Across all of them, the strategy compounded at 28.43% annualized, turning $1 into $524, versus the benchmark's $9 over the same period.
Critically, this return advantage was not achieved by taking proportionally more risk. Annualized volatility was only modestly higher at 20.95% versus 18.68%, meaning the strategy delivered dramatically more return per unit of risk carried.
Strategy (gold) vs Benchmark (grey). Bars scaled relative to strategy maximum.
Cumulative Growth: 2002–2026
Logarithmic scale. Strategy (blue) vs SPY Buy & Hold (gold). Chart formatting will be updated to match page theme.
03 — Factor Composition
What the Regression Captures
An OLS regression of daily strategy returns against seven major factor ETFs (R² = 0.505) reveals the strategy's primary static exposures. The model uses 3,177 daily observations and is statistically robust (F-statistic 462.6, p ≈ 0.00).
Momentum is the single most important factor, reflected in an MTUM coefficient of +0.92 (t = 25.0) — by far the largest and most significant loading in the model. This dominant momentum tilt is the deliberate core of the strategy. Broad market (SPY: +0.35) and incidental small-cap correlation (IWM: +0.23) appear as secondary loadings, while the strategy carries a strong negative low-volatility exposure (USMV: −0.66) — a natural consequence of concentrating capital in high-momentum names rather than defensively diversifying across low-beta holdings.
Importantly, this regression captures only the strategy's static, time-averaged factor exposures. The remaining 50% of return variance is not random noise — it contains three additional, academically well-documented premia that static OLS regressions against ETF benchmarks are structurally incapable of measuring: the rebalancing premium, the concentration premium, and the dynamic regime-adjustment premium produced by the market regime overlay. These are explained in the section below.
04 — Academically Documented Premia Beyond the Static Regression
What the OLS Cannot See
A static factor regression measures average exposures to passive, buy-and-hold factor benchmarks. It cannot measure return contributions that arise from how the portfolio is constructed and managed. The strategy systematically harvests three additional, well-researched premia that are invisible to the regression but central to the performance record.
I. The Momentum Premium — The Largest Documented Anomaly in Finance
The strategy's foundation rests on one of the most robust and extensively replicated findings in academic finance. Jegadeesh and Titman's landmark 1993 paper in The Journal of Finance established that stocks outperforming over 3–12 month periods continue to outperform over the following year — generating approximately 1% per month in excess returns. This finding has since been replicated across dozens of countries, asset classes, and time periods, and remains intact out-of-sample more than 30 years after its original publication.
Critically, momentum profits are not explained by systematic risk. Jegadeesh and Titman (1993) and subsequent work by Fama and French (1996) confirmed that neither the CAPM nor the Fama-French three-factor model can account for momentum returns — making it the most significant challenge to the efficient market hypothesis in the empirical literature. The MTUM coefficient of +0.92 in the regression reflects the strategy's deliberate, concentrated alignment with this premium — the highest single factor loading in the model and the clearest expression of where the strategy's return generation is anchored.
II. The Rebalancing Premium — Structural Alpha from Portfolio Mathematics
One of the most important and underappreciated sources of return in the strategy is the rebalancing premium, also referred to in the academic literature as "volatility harvesting," "volatility pumping," or the "excess growth rate." This premium arises not from security selection but from the mathematical properties of compound growth itself, and it cannot be replicated by any static, buy-and-hold exposure.
The theoretical foundation was established by Fernholz and Shay (1982) in their development of Stochastic Portfolio Theory, and has since been extended by Booth and Fama (1992), Luenberger (1997), and Bouchey et al. (2012), among many others. The core insight is that a rebalanced portfolio's compound growth rate exceeds the weighted average of its components' compound growth rates — an excess that accumulates systematically over time. This "excess growth rate" is not a market inefficiency; it is a mathematical consequence of variance reduction through disciplined rebalancing, and importantly, it cannot be arbitraged away even if all market participants adopt the same approach.
The mechanism works as follows: compound wealth growth equals arithmetic return less half the portfolio's variance. By systematically selling assets that have appreciated and rebalancing into relatively underperforming positions, the strategy continuously reduces portfolio variance relative to a drifting buy-and-hold portfolio — capturing the mathematical difference as excess compounded return. Research by ReSolve Asset Management, drawing on Fernholz's framework, estimates that well-constructed rebalancing strategies can contribute 1–2.5% per year in excess compound growth over buy-and-hold alternatives at equivalent volatility levels. A key finding from Pliakha, Uppal and Vilkov (2016) confirmed that monthly rebalancing specifically exploits cross-autocorrelation and short-term reversal in stock returns that is present at the monthly frequency — directly consistent with why the strategy's monthly cadence outperforms more frequent or less frequent alternatives.
The rebalancing premium is mathematically guaranteed to exist as long as the portfolio holds assets with non-trivial variance and imperfect correlation. It is one of the few genuine free lunches in portfolio theory — not a market anomaly that can be arbitraged away, but a property of compound arithmetic itself.
III. The Concentration Premium — Best Ideas Outperform Diluted Portfolios
The third premium harvested by the strategy — and the one most invisible to a static factor regression — arises from portfolio concentration. Academic research consistently documents that high-conviction, concentrated positions in the highest-ranked signals outperform diversified portfolios of the same factor, because diversification dilutes the signal with lower-conviction positions that add noise without adding expected return.
The seminal work on this is Anton, Cohen and Polk's "Best Ideas" study, which found that a fund manager's highest-conviction positions — their largest active bets — outperform the market and the rest of their own portfolios by approximately 2.6% to 4.5% per year. The implication is direct: a portfolio systematically constructed around the strongest signals in the highest-momentum names, rather than spreading capital evenly across a broad factor universe, captures a materially larger share of the factor premium. Ivković, Sialm and Weisbenner (2008) in the Journal of Financial and Quantitative Analysis further confirmed that concentrated portfolios outperform diversified ones, with excess returns particularly pronounced for positions outside the S&P 500 index — consistent with the strategy's small-cap tilt.
Novy-Marx (2014) adds an important further nuance: concentrated pure-factor portfolios — those holding only the strongest signal names rather than broad baskets — produce substantially higher factor alphas than their diversified counterparts, because they avoid the performance dilution of borderline holdings. The strategy's concentrated construction is therefore not an incidental feature; it is a deliberate mechanism for extracting a larger share of the theoretical factor premium than broad, index-like implementations can achieve.
IV. The Proprietary Weighting Premium — Signal Intensity Determines Position Size
The strategy does not allocate capital through value-weighting (proportional to market capitalization) or simple equal-weighting. Instead, a proprietary mechanism determines position sizes based on the characteristics of each holding — a design choice with substantial academic support for producing superior risk-adjusted outcomes over both conventional alternatives.
The starting point for why this matters is well-established: value-weighted portfolios are the weakest performers among common weighting schemes. ReSolve Asset Management's comprehensive framework analysis confirms that market-cap weighting is only mean-variance optimal under the restrictive assumption that all stocks have identical Treynor ratios — a condition that empirically never holds. In practice, market-cap weighting systematically overweights the most expensive and recently appreciated securities, embedding a structural drag that signal-based weighting schemes avoid.
The academic evidence for signal-based weighting over both value-weighting and naive equal-weighting is consistent across research groups. ReSolve's dynamic asset allocation research finds that momentum-weighted portfolios improve upon equal-weight baselines by approximately 1.5% per year in annualized returns and nearly 40% in Sharpe ratio, while simultaneously reducing maximum drawdown by roughly 25%. The mechanism is intuitive: when the magnitude of the momentum signal is used to determine position sizing, the portfolio concentrates capital in the securities with the strongest expected continuation — rather than treating a borderline position identically to the highest-conviction holding.
MSCI's analysis of weighting schemes across factor portfolios adds further precision. Studying value, quality, and momentum factor portfolios from 2000 to 2020, MSCI found that score-tilt weighting — assigning weights in proportion to signal strength — consistently produced the highest transfer coefficient of all weighting approaches: the most efficient translation of factor insights into actual portfolio exposures. Equal-weighting and market-cap weighting both systematically diluted factor signal intensity, reducing realized factor returns relative to what the underlying signal would theoretically deliver. The implication is direct: a proprietary weighting mechanism that scales position sizes to signal strength harvests more of the theoretical factor premium than any naive weighting scheme can achieve.
Cirulli and Walker's 2023 study in the SSRN literature confirms this across the MSCI global universe from 2002 to 2022: portfolios that screened and weighted by momentum and low-volatility characteristics consistently outperformed equal-weighted portfolios on returns, risk-adjusted returns, and downside risk measures in every geographic region tested — and the equally-weighted portfolio never outperformed the signal-enhanced alternatives on a rolling three-year Sharpe basis except for a brief period in 2021. The conclusion their research draws — that equal-weighting is not a particularly challenging benchmark to beat with even simple signal-based modifications — reinforces the value of the strategy's proprietary weighting approach relative to any passive allocation scheme.
The combined evidence suggests that proprietary signal-based weighting contributes meaningfully to realized returns through two channels simultaneously: it increases factor exposure intensity by concentrating capital in the highest-conviction positions, and it reduces portfolio variance by naturally underweighting lower-signal, higher-noise holdings — compounding the rebalancing premium described above.
Putting It Together: Why the Static R² = 0.50 is Not a Concern
The OLS R² of 0.505 is often misread as indicating that half the strategy's variance is unexplained noise. In fact, it reflects the fundamental limitation of regressing a dynamic strategy against static ETF benchmarks. The premia above — momentum as the dominant factor anchor, the rebalancing premium, the concentration premium, and proprietary signal-based weighting — all produce return patterns that are orthogonal to simple factor ETF returns, and therefore appear as "unexplained variance" in the regression even when they are entirely systematic and grounded in financial theory.
A more complete accounting of the strategy's return sources would require a conditional factor model separating high- and low-regime periods, a term for the rebalancing premium computed from the portfolio's actual cross-sectional variance, a concentration adjustment for signal intensity, and a transfer-coefficient adjustment for weighting efficiency. No such unified regression framework currently exists as a practical benchmark tool — which is precisely why the static OLS finding of zero residual alpha should be understood as a measurement limitation, not a verdict on the strategy's edge.
05 — The Market Regime Driven Exposure Overlay
Where Discipline Compounds Into Edge
The strategy's defining structural feature is a monthly market regime driven exposure overlay that adjusts position sizing at each month-end rebalance based on prevailing market regime signals. The overlay does not trade intra-month — a design choice that is empirically validated, not assumed.
Why Monthly, Not Daily or Weekly?
Extensive backtesting of intra-month regime reaction consistently produced significantly worse results than the monthly cadence. This counterintuitive finding reflects a well-documented market microstructure reality: short-term dislocations are predominantly mean-reverting noise. Reacting to them intra-month means reducing exposure after damage is done, missing the sharp recovery that typically follows, and accumulating transaction costs — a compounding drag over 24 years that the monthly architecture entirely avoids.
06 — Tail Risk & Drawdown Management
The Clearest Fingerprint of the Risk Management Overlay
Kurtosis: 4.35 vs 14.35
The most striking statistical difference between the strategy and its benchmark is not returns — it is the shape of the return distribution. The benchmark's kurtosis of 14.35 reflects the fat tails of unmanaged equity exposure. The strategy's kurtosis of 4.35 demonstrates systematic left-tail truncation.
Fat tails do not simply disappear. They are mechanically cut off by a risk overlay that reduces position sizing as adverse market regimes are detected — entering periods of peak stress with less exposure, and less capital at risk. This cannot be replicated by any static factor allocation.
Recovery Speed
The Recovery Factor of 22.68 versus 4.78 for the benchmark, and a longest drawdown period of just 713 days versus 1,772 days, reflects a compounding asymmetry unique to dynamic risk management. When the overlay reduces exposure near the bottom of a drawdown, the strategy enters the recovery phase with more capital intact — capturing more of the rebound and exiting the drawdown far sooner.
The Ulcer Index of 0.08 versus 0.13 for the benchmark quantifies this: despite higher returns, the strategy produces less sustained pain per dollar invested.
Worst Five Drawdowns — All Under 31%
Across five distinct stress events spanning different causes, speeds, and market regimes, the strategy's drawdowns were consistently bounded:
| Period | Context | Duration | Max Drawdown |
|---|---|---|---|
| Nov 2007 – Oct 2009 | Global Financial Crisis | 713 days | −30.04% |
| Apr 2021 – Aug 2021 | Momentum unwind / reflation | 135 days | −23.14% |
| Jan 2004 – Sep 2004 | Post-bubble consolidation | 225 days | −22.51% |
| Jun 2015 – May 2016 | China slowdown / momentum stress | 347 days | −21.76% |
| May 2006 – Nov 2006 | Mid-cycle correction | 203 days | −20.71% |
The GFC result is particularly notable. Momentum strategies without risk management famously suffered drawdowns of 40–80% in 2008–2009. Containing the drawdown to −30% while maintaining heavy momentum exposure is direct evidence of the overlay functioning under the most extreme stress conditions in the backtest period.
07 — Skill vs. Structural Edge
Why This Is Not a Backtest Artifact
The most important question any investor should ask of a quantitative strategy: is this performance the result of genuine edge, or a well-fitted historical artifact? The evidence across multiple independent dimensions points clearly to the former.
Overfitting and Lookahead Bias, a Common Pitfall
Two common mistakes that can artificially inflate backtest results are overfitting and lookahead bias. Overfitting occurs with complex models that have a large number of parameters that all must be trained or tuned using historical data. The more parameters defined by the strategy, especially when combined with machine learning or non-linear optimization techniques, the higher the likelihood that the results are merely artifacts of fitting noise in the past rather than genuine signals that persist in the future. Our strategy is fundamentally protected against overfitting: it is governed by a small number of parameters, contains no machine learning or AI components, and relies instead on rule-based logic grounded in structural market behavior. Most importantly, when each parameter is adjusted ±20% in either direction, performance does not exhibit non-linear decay but rather remains smooth and consistent across this range. This smooth sensitivity profile is the critical test of robustness—it confirms the strategy is not exploiting a precise historical threshold that may never recur, but rather a broad structural phenomenon that persists across a wide range of reasonable parameterizations.
Lookahead bias occurs when a backtest accidentally uses information from the future to make trading decisions about the past, artificially inflating returns and creating unrealistic assumptions about what information was genuinely available at decision time. Our walk-forward backtesting mechanism eliminates lookahead bias entirely: on any given trading day, the overlay's decision mechanism has access only to price data from the previous day and has zero visibility into prices on the day the decision is being made. The strategy cannot "know" whether a day will be up or down, what volatility will be, or where the close will be—all decisions are made with information that would have been genuinely available before market open. The walk-forward window progresses one period at a time with a clear information barrier between the decision window and the subsequent evaluation period, ensuring that every backtest result reflects decisions that could have actually been made in real-time, with no unfair advantage from future information and perfect alignment with how the strategy would operate in live trading.
08 — Known Risks & Limitations
What Investors Should Understand
Intellectual honesty requires stating clearly where the strategy carries structural risk, and what the overlay does not protect against.
Mid-Month Market Shocks
Position sizing is set at the monthly rebalance and is not adjusted intra-month. A significant market shock arriving after the rebalance date will be absorbed at full exposure until month-end. This is a deliberate design choice — intra-month reaction was tested extensively and produced worse outcomes — but investors should understand that short-term drawdowns from sudden exogenous events are an accepted and bounded cost of the architecture.
Momentum Factor Regime Risk
The strategy carries dominant momentum exposure. Rapid, sharp momentum factor reversals — particularly following extended momentum rallies — represent the primary stress scenario. The overlay's monthly cadence means it responds to sustained market regime shifts more effectively than to instantaneous spike-and-reversal events. All five historical drawdowns were recovered; the structural question for any investor is whether they can tolerate drawdowns in the 20–30% range during such episodes.
Structural Regime Change Risk
The overlay's logic is calibrated to market regime dynamics as they have existed in modern equity markets. A structural break in how regimes behave — or fundamental changes in market microstructure — represents a tail risk that no backtest fully captures. The smooth parameter sensitivity provides meaningful protection against mild regime shifts; it does not guarantee performance under a fundamentally altered market structure.
Capacity Constraints
The strategy's momentum and small-cap tilts carry real-world capacity limits. Market impact is not modeled in the backtest. Investors should discuss capacity considerations directly before committing capital.
Important Disclosures
All performance figures presented are from a systematic backtest conducted over the period January 2002 through March 2026. Backtested performance is hypothetical, has not been audited, and does not represent actual trading results. Past performance, whether live or simulated, is not indicative of future results. All investments involve risk, including the possible loss of principal. This document is for informational purposes only and does not constitute investment advice or a solicitation to invest. Prospective investors should review all offering documents carefully and consult with their own financial, legal, and tax advisors before making any investment decision.