1.4 Momentum

Basic Prerequisites: Econometrics & Time Series (serial correlation, stationarity, forecasting) and Probability & Measure Theory (moments, skewness, correlation). (these are the folder-level prerequisites for pages 02–06; page 01 states its own, smaller, entry requirements)

1. Intuition & Practical Objective

Momentum is the empirical tendency for assets that have recently gone up to keep going up and assets that have recently gone down to keep going down, over intermediate horizons of roughly 1 to 12 months. It is one of the most persistent and pervasive anomalies in finance: documented in US equities since Jegadeesh & Titman (1993), across over 40 countries and 200+ years of data (Asness, Frazzini, Israel & Moskowitz 2014), and across asset classes - equities, bonds, currencies, and commodities (Asness, Moskowitz & Pedersen 2013). The US momentum premium is roughly an 8.3% per-year average spread between recent winners and recent losers (1927–2013).

There are two mathematically distinct implementations, and confusing them is a common error:

Cross-sectional momentum (relative, XSMOM): rank a universe by past return; go long winners, short losers (dollar-neutral). The bet is relative - an asset wins because it beat its peers.
Time-series momentum / trend (absolute, TSMOM): look at each asset alone; go long if its own past return is positive, short if negative, and scale each position inversely to its own volatility. This is the CTA / trend-following engine. The bet is absolute - an asset wins because it went up.

This folder is the topic-hub for momentum in Kwant-Atlas. It (a) gives the fast formula lookup below - job #1 of a hub - and (b) routes to six sub-pages walking from raw intuition through cross-sectional momentum, time-series momentum, the value–momentum interaction, failure modes (momentum crashes, crowding, the 2009 reversal), and advanced extensions (volatility scaling, dynamic weighting).

The one-sentence essence. “Past winners beat past losers because a positive serial-correlation structure in returns (underreaction / slow information diffusion, reinforced by herding) makes 1–12 month past returns a genuinely informative predictor of future returns - but that same structure concentrates risk into a negatively-skewed, crash-prone payoff that volatility scaling and value diversification can partly tame.”

2. Mathematical Ground Truth & Derivations

Quick-Reference Lookup (job #1). All formulas below are transcribed from the verified corpus papers (Jegadeesh–Titman 2001; Moskowitz–Ooi–Pedersen 2012; Asness–Moskowitz–Pedersen 2013; Daniel–Moskowitz 2016; Barroso–Santa-Clara 2015). Numbers in the check column were re-executed and reproduced exactly from the working Python in §3 and the sub-pages.

Notation: $r_{i, t}$ monthly return of asset $i$ at $t$ ; $R_{i}^{(L)} = \prod_{k = 1}^{L} (1 + r_{i, t - k}) - 1$ trailing $L$ -month cumulative return; $rank_{i}$ the 1-indexed rank of $R_{i}^{(12)}$ ; $σ_{i, t}$ ex-ante annualized vol; $σ_{tgt}$ target vol; $Ω = E [(R_{t - 12, t} - 12 μ) (R_{t, t + 1} - μ)^{'}]$ the 12-month-vs-next-month cross-covariance matrix.

Quantity	Formula	Verified check
Cross-sectional signal (skip last month)	$R_{i}^{(12 - 1)} = \prod_{k = 2}^{12} (1 + r_{i, t - k}) - 1$	-
Rank weights (dollar-neutral)	$w_{i} = \frac{rank _{i} - \frac{N + 1}{2}}{\sum _{j} ∥ rank _{j} - \frac{N + 1}{2} ∥}$	$\sum_{i} w_{i} = 0.000$ , $\sum_{i} ∥ w_{i} ∥ = 1.0000$
WML (winners minus losers)	$WML_{t} = \sum_{i} w_{i} r_{i, t}$	skip- $12 - 1$ : SR $+ 1.75$ ; include-last-mo: SR $- 0.07$
TSMOM position	$pos_{i, t} = sign (R_{i}^{(12)}) \cdot \frac{σ _{tgt}}{σ _{i, t - 1}}$	realized vol $39.2%$ vs target $40%$
EWMA ex-ante vol (MOP)	$σ_{i, t}^{2} = 261 \sum_{k} (1 - δ) δ^{k} (r_{i, t - 1 - k} - \overset{r}{ˉ})^{2}$ , center-of-mass $\frac{δ}{1 - δ} = 60$ d	-
Diversified TSMOM return	$r_{t, t + 1}^{TSMOM} = \frac{1}{S _{t}} \sum_{s} sign (r_{t - 12, t}^{s}) \frac{40%}{σ _{t - 1}^{s}} r_{t, t + 1}^{s}$	ann. vol $11.7%$ (paper: $12%$ ), SR $+ 1.95$
XSMOM expected return	$E [r^{XS}] = \frac{tr ( Ω )}{N} - \frac{1 ^{'} Ω 1}{N ^{2}} + 12 σ_{m}^{2}$	3 channels: own-AC, cross-serial, mean-dispersion
TSMOM expected return	$E [r^{TS}] = \frac{tr ( Ω )}{N} + \frac{12 μ ^{'} μ}{N}$	time-series channel dominates
TSMOM on XSMOM	$β (TSMOM, XSMOM) = 0.66$ ( $t = 15.2$ , $R^{2} = 44%$ )	related but not the same (alpha $+ 76$ bp/mo)
Combo (value+momentum)	$r^{COMBO} = 0.5 r^{VALUE} + 0.5 r^{MOM}$	combo SR $+ 0.67$ vs max(0.34) alone
Optimal dynamic weight (DM)	$w_{t} \propto \frac{μ ^ _{t}}{σ ^ _{t}^{2}}$ (conditional vol $\propto$ conditional Sharpe)	doubles static Sharpe

Critical caveat. Momentum’s average Sharpe hides a strongly negative skewness: the US WML monthly log-return skewness is $- 4.70$ (1927–2013). The whole failure-mode story - crashes in panic states, the option-like payoff of losers, crowding - lives in the higher moments, not the mean. See §4 and 05 · Failure Modes.

3. Computational Implementation - the formula engine

Standard library only. Reproduces the two cleanest verified numbers from §2: the rank-weight dollar-neutrality of the cross-sectional portfolio, and the value–momentum combination variance identity.

4. Failure Modes & First-Principles Breakdowns

Hub signposts - the full analysis lives in 05 · Failure Modes. In one line each:

Momentum crashes (negative skewness). After market declines the short leg (past losers) becomes high-beta and option-like; a sharp rebound makes them “crash up” far faster than winners, so WML loses violently (July–Aug 1932: $- 74.4%$ and $- 61.0%$ ; March–May 2009: losers $+ 163%$ vs winners $+ 8%$ ). Skewness $≪ 0$ .
The momentum turning-point whipsaw. In range-bound, mean-reverting regimes the strategy buys tops and shorts bottoms - a long, grinding drawdown (“death by a thousand papercuts”) distinct from the fast crash.
Crowding & capacity. Momentum is a factor everyone knows; capital inflow into the trade compresses returns, raises turnover/impact costs, and creates herding that can itself trigger a reversal (the 2009 crowding reversal).
Short-term reversal contamination. The most recent month reverses; using it in the signal destroys the premium - which is why the convention skips it.

5. References

Jegadeesh, Narasimhan & Titman, Sheridan: Profitability of Momentum Strategies: An Evaluation of Alternative Explanations, J. Finance 56(2), 699–720 (2001). Verified corpus refs/13; extends the classic 1993 JFE paper that first documented US equity momentum (1965–1989).
Moskowitz, Tobias J., Ooi, Yao Hua & Pedersen, Lasse Heje: Time Series Momentum, J. Financial Economics 104(2), 228–250 (2012). *Verified corpus refs/14
Asness, Clifford S., Moskowitz, Tobias J. & Pedersen, Lasse Heje: Value and Momentum Everywhere, J. Finance 68(3), 929–985 (2013). *Verified corpus refs/15
Daniel, Kent & Moskowitz, Tobias J.: Momentum Crashes, J. Financial Economics 122(2), 221–247 (2016). *Verified corpus refs/16
Barroso, Pedro & Santa-Clara, Pedro: Momentum Has Its Moments, J. Financial Economics 116(1), 111–120 (2015). *Verified corpus refs/17
Asness, Clifford S., Frazzini, Andrea, Israel, Ronen & Moskowitz, Tobias J.: Fact, Fiction, and Momentum Investing, J. Portfolio Management 40th Ann. (2014). *Verified corpus refs/18

6. Connected Graph Bridges

Foundational base: Econometrics & Time Series · Probability & Measure Theory · Statistics & Inference
Sibling topics (this pillar): Fundamental Multi-Factor Models (momentum as a factor) · Statistical Arbitrage & Pairs (mean-reversion - momentum’s mirror image) · Backtesting Hygiene (multiple-testing caveat)
Risk & portfolio: ES) · Risk Parity (vol targeting) · Mean–Variance (combo)
Sub-pages (in-folder): 01 From Zero · 02 Cross-Sectional Momentum · 03 Time-Series Momentum · 04 Value–Momentum Interaction · 05 Failure Modes & Practice · 06 Advanced Extensions

Beginner: start at 01 · Practitioner: start at 05

Kwant Atlas

Explorer

1. Intuition & Practical Objective

2. Mathematical Ground Truth & Derivations

3. Computational Implementation - the formula engine

4. Failure Modes & First-Principles Breakdowns

5. References

6. Connected Graph Bridges

1.4.1 Momentum from Zero

1.4.2 Cross-Sectional Momentum

1.4.3 Time-Series Momentum

1.4.4 The Value–Momentum Interaction

1.4.5 Failure Modes & Practice

1.4.6 Advanced Extensions