Simple Systematic Momentum with Rules

What it is

This is a fully specified momentum system for a portfolio of stocks: rank a liquid universe by past returns, hold the top slice, rebalance on a schedule, and use a market filter to reduce exposure in bear markets. Every rule is written so that two people running it would hold the same portfolio. It is included as a worked example of what "systematic" means, and because momentum is the anomaly with the most out-of-sample support. It is not included because it is easy to live with; its drawdowns are large and its worst periods are long.

The logic

Stocks that have outperformed over the past 6 to 12 months have, on average, continued to outperform over the next 1 to 12 months, in most markets and most decades studied. The explanations are behavioural (under-reaction to news, then herding) and institutional (slow capital flows, benchmark-hugging). The most recent month is excluded from the ranking because short-term returns tend to reverse.

The other side is value investors selling winners, contrarians, and index funds that must sell stocks leaving an index. Momentum's counterparties are not naive; the premium exists because holding momentum through its crashes is genuinely unpleasant, and most participants will not.

Setup rules

Universe: the largest 1,000 US stocks by market cap (or the constituents of a broad index), with average daily dollar volume above $10 million and a price above $5. Recomputed at each rebalance from data available at the time, including stocks later delisted.
Ranking: total return over months 2 through 12 (skip the most recent month). Optionally divide by the stock's realised volatility over the same window to favour smoother trends.
Selection: hold the top 10 percent (100 stocks) or, for a smaller account, the top 30 by rank, equal weight.
Rebalance: monthly, on the first trading day, at the open. A stock is sold if it drops below the top 20 percent (a buffer to cut turnover) and replaced with the highest-ranked stock not held.
Market filter: when the broad index closes the month below its 10-month moving average, hold only the top-ranked half of the normal count and hold the rest in treasury bills; or, in the stricter version, move fully to bills.
No discretion: no overrides for news, valuations or "obviously overextended" names.

Entry, stop, target

Entries and exits are the rebalance. There are no per-stock stops; the monthly rank is the exit, and individual positions can and do fall 30 to 50 percent between rebalances. That is the nature of the system, and a per-stock stop has historically reduced returns more than it reduced risk in most tests, because momentum stocks are volatile.

Item	Value	Notes
Universe	Top 1,000 by market cap, liquid	Survivorship-free
Ranking	12-1 month return	Volatility-scaled optional
Holdings	Top 10 percent, equal weight	Or top 30 for small accounts
Rebalance	Monthly	Buffer to cut turnover
Annual turnover	Roughly 100 to 200 percent	Costs and taxes follow
Market filter	Index below 10-month MA	Reduce or exit
Historical worst drawdown	Deeper than the index in momentum crashes	Past tendency, not a forecast

There is no per-trade R:R; the relevant statistics are annual return over the index, tracking error, turnover and max-drawdown, all of which should be stated from your own backtest, not from a paper.

Position sizing and risk

Equal weight across 30 to 100 names caps single-stock risk at 1 to 3 percent of the portfolio per name. Portfolio risk is the market filter and the allocation to the system relative to other holdings, covered in /learn/risk-management. /tools/position-size is useful for converting the equal-weight target into share counts for each rebalance. Never add leverage to a momentum system; its crashes arrive precisely when leverage is most damaging.

What breaks it

Momentum crashes. At sharp market bottoms the past losers rally violently and the past winners lag; a long-only momentum portfolio can underperform the index by 20 to 30 percentage points in a few months. The market filter reduces this; it does not remove it.
Costs and taxes. High turnover in a taxable account converts much of the premium into short-term gains; after costs the net edge of a retail implementation is materially smaller than the academic figure.
Crowding. Momentum is a core factor for many large funds; when they de-risk together, momentum stocks fall together.
Edge decay. The premium has narrowed since the 1990s; it has not disappeared in most tests, but the reward-to-pain ratio has worsened.
Implementation drift. The trader skips a rebalance, keeps a favourite name, or adds a "quality" screen after a bad month. Each tweak makes the live system different from the tested one.

How to test it

Get 25 or more years of survivorship-free daily data with delistings. Simulate the rules exactly, with the universe recomputed monthly from point-in-time data, costs of 0.2 percent per side plus a market-impact estimate, and a one-day lag between signal and execution. Report annual return, sharpe-ratio, max-drawdown, worst 12-month relative return, turnover and the number of months in the filter. Then vary the lookback (6, 9, 12 months), the holdings count and the filter; the results should be broadly similar, and if only one combination works, discard the system. Then run a walk-forward-testing protocol. Paper-trade the monthly rebalance for a year before committing capital; the operational load of rebalancing 30 to 100 names is part of the test.

Variations

Sector or ETF momentum for a simpler, lower-turnover version; see relative-strength-rotation.
Long-short momentum shorting the bottom decile; academically cleaner, practically much harder (borrow, squeezes).
Momentum plus quality screening out the most leveraged or unprofitable names; modest evidence, more parameters.