Classic strategies in the library

Most published quant ideas belong to a small set of families. Quant Buffet encodes them as backtest.templates and as custom code across the library. Recognising the family lets you read any strategy article in a couple of minutes, because you already know its economic story and its characteristic failure mode.

Template cheat sheet

Family 1 — Trend following

Origin. Mebane Faber's 2007 paper on tactical asset allocation popularised holding an asset only while it trades above its 10-month (roughly 200-day) average. Why it might work. Investors react to news gradually and institutions de-risk slowly, so declines cluster. Characteristic failure. Choppy, range-bound markets produce whipsaws — the rule sells after a dip and rebuys higher, repeatedly.

from backtest.templates import make_sma_trend

ASSETS = ["SPY", "EFA", "TLT", "GLD", "BIL"]

def make_on_day(prices):
    # Rebalances monthly; BIL almost always sits above its own average,
    # so it behaves as the defensive sleeve.
    return make_sma_trend(prices, ASSETS, {"sma_days": 200})

Family 2 — Cross-sectional momentum

Origin. Jegadeesh and Titman (1993) showed that past 3–12 month winners kept outperforming past losers. Why it might work. Underreaction to information, plus flows that chase performance. Characteristic failure. Momentum crashes at violent turning points: in the spring of 2009, as the market bottomed and rebounded, momentum portfolios were positioned in defensive winners and suffered severe losses in weeks.

from backtest.templates import make_momentum_rotation

ASSETS = ["XLK", "XLF", "XLE", "XLV", "XLI", "XLP", "XLU", "XLY", "BIL"]

def make_on_day(prices):
    # Hold the 3 strongest sectors, refreshed on each month boundary
    return make_momentum_rotation(prices, ASSETS, {"lookback": 126, "top_n": 3})

Family 3 — Dual momentum

Origin. Gary Antonacci (2014) combined *relative* strength (which asset is best?) with *absolute* strength (is it beating cash at all?). Why it dominates this library. It answers both questions a real allocator has, in two lines of logic. Characteristic failure. It concentrates hard: the template holds exactly one position at a time, so a single bad month is felt in full.

from backtest.templates import make_dual_momentum

ASSETS = ["SPY", "EFA", "EEM", "TLT", "GLD", "BIL"]

def make_on_day(prices):
    # Holds exactly one asset: the strongest risky sleeve, or cash if none is positive
    return make_dual_momentum(prices, ASSETS, {"lookback": 252, "cash_symbol": "BIL"})

Family 4 — Mean reversion

Origin. De Bondt and Thaler (1985) documented long-horizon reversal; short-horizon versions became a staple of the 2000s. Why it might work. Forced selling and liquidity provision get paid. Characteristic failure. It is short volatility in disguise — it buys falling assets, so a genuine crisis is exactly when it hurts most, and its high turnover (Lesson 5) eats the edge.

from backtest.templates import make_mean_reversion

ASSETS = ["SPY", "QQQ", "IWM", "BIL"]

def make_on_day(prices):
    # Buy when the 10-day z-score is stretched to the downside
    return make_mean_reversion(
        prices, ASSETS, {"lookback": 10, "entry_z": -1.0, "exit_z": 0.0}
    )

Family 5 — Risk-based allocation

Origin. Bridgewater's All Weather and Harry Browne's Permanent Portfolio; later formalised as risk parity. Why it might work. Equal *dollars* is not equal *risk* — a 60/40 portfolio is roughly 90% equity risk. Sizing by inverse volatility balances contributions instead. Characteristic failure. 2022 again: when every sleeve falls together and rates rise, a levered or duration-heavy risk-parity book has nowhere to hide.

from backtest.templates import make_risk_parity, make_vol_target

ASSETS = ["SPY", "TLT", "GLD", "DBC"]

def make_on_day(prices):
    # Inverse-volatility weights from a 63-day window (about one quarter)
    # Swap in make_vol_target to scale total exposure to a volatility budget instead.
    return make_risk_parity(prices, ASSETS, {"vol_lookback": 63})

One table to read any strategy page

Case study: run this in the IDE right now

Open any dual-momentum strategy, then make exactly these three edits, one at a time, recording Sharpe and max drawdown after each run. This is the whole research loop in five minutes.

1. Baseline — run the code unchanged and write down CAGR, Sharpe, and max drawdown.
2. Lookback — change lookback from 252 to 126. A faster signal reacts sooner but whipsaws more, so watch turnover as well as return.
3. Concentration — swap make_dual_momentum for make_momentum_rotation with top_n: 3. Expect a lower CAGR but a shallower drawdown; you traded upside for stability.
4. Universe — swap ASSETS for the MULTI_ASSET book. More sleeves usually smooth the curve and dilute the best year.
5. Stress window — set start to 2007-01-01 and judge the strategy only on 2008 and 2022.

Where to go next

• Strategy Library — filter by momentum, mean reversion, or risk parity and read three articles in the same family to see how authors differ.
• API docs — the full reference for ASSETS, make_on_day, PortfolioEngine, and compute_metrics.
• Platform export — take any strategy to QuantConnect, Backtrader, Zipline, VectorBT, or Freqtrade when you outgrow the lab.
• Backtest Usage in your profile — track how many runs and simulated trades your research has actually produced.