TSMOM
Time-series momentum: measure trend persistence across multiple assets
Risk
Medium
Holding Period
1 month (rebalanced monthly)
Best For
Capturing strong market trends across multiple assets
How it works
Academic research has documented that assets with positive excess returns over the past 12 months tend to exhibit positive returns over the following month, and vice versa. TSMOM applies this observation mechanically across a diversified asset universe.
Mathematical Foundation
signal = sign(R_{t−21,t−252}) size ∝ target_vol / σ_EWMASignal Generation Logic
- 1Calculate the 12-month (252-day) total return, skipping the most recent 21 days to avoid short-term reversal effects
- 2The sign of this return determines direction: positive → long (+1), negative → short (−1)
- 3Calculate R² of a linear regression on log-prices over the past 63 days to measure trend consistency
- 4Estimate realised volatility using EWMA with a 30-day half-life
- 5Scale position size inversely to volatility so each asset contributes equal risk toward the 10% annual target
- 6Rebalance monthly; apply trailing drawdown stop at 20% from peak
Parameters Explained
lookbackCore momentum lookback in trading days (~12 months). This is the primary signal window. Shorter periods are noisier; longer periods react more slowly.
Default
252skip_recentDays to exclude from the end of the lookback (~1 month). Skips the short-term reversal anomaly documented in academic literature.
Default
21target_volTarget annualised portfolio volatility. Each asset is sized so that its individual contribution equals this target. Higher values mean larger positions and more risk.
Default
10%ewma_halflifeHalf-life of the exponentially weighted moving average for volatility estimation. 30 days means recent volatility is weighted roughly twice as much as volatility from 30 days ago.
Default
30max_drawdown_pctMaximum trailing drawdown from the position's peak before forced exit. Limits losses on individual momentum positions that suddenly reverse.
Default
20%When It Works
In markets with clear and sustained directional trends, especially when trends are driven by macroeconomic fundamentals (interest rate cycles, commodity supply shocks, risk-on/risk-off flows). Historically works well across equities, fixed income, currencies, and commodities.
When It Fails
During sharp trend reversals (e.g., 2009 recovery, 2020 COVID crash and bounce). Also underperforms in range-bound, choppy markets where trend signals are frequent but unreliable. Monthly rebalancing means slow response to sudden regime shifts.
Risks & Limitations
- Whipsaws: frequent false signals in choppy markets generate transaction costs without gains
- Large drawdowns during trend reversals — momentum strategies are known to have fat-tail risk
- Crowding risk: many institutional funds run TSMOM, so reversals can be amplified and fast
- Transaction costs from monthly rebalancing across many assets can be significant
- Short-selling constraints on some assets may prevent negative-momentum positions
Implementation
Based on Moskowitz, Ooi & Pedersen (2012) Journal of Financial Economics paper 'Time Series Momentum'. Uses EWMA volatility estimation (not simple rolling std) for more responsive risk scaling. Trend-strength R² filter can optionally screen out weak signals.
Model parameters
Lookback
12-month return measurement window
Skip Recent
Skip last month to avoid short-term reversal
Target Vol
Annualised volatility target for sizing
EWMA Half-life
Decay rate for volatility estimate
Max Drawdown
Trailing drawdown exit threshold
Academic background
Academic Basis
Based on Moskowitz, Ooi & Pedersen (2012), 'Time Series Momentum', Journal of Financial Economics
Backtest this strategy
Run the exact model on your selected assets and date range. See trade-by-trade performance.
Backtest This →