Asset classes & ETF proxies

An asset class is a category of investments that responds to the same economic forces. Real portfolios mix them so that no single force can sink everything at once. In Quant Buffet you trade ETF proxies from backtest.universes — liquid, cheap to simulate, and comparable across the whole library.

Why ETFs instead of single stocks?

• Liquidity — SPY turns over tens of billions of dollars a day, so a slippage model of a couple of basis points is realistic.
• Diversification — one ticker gives you hundreds of underlying names.
• Survivorship — the index sheds failing members for you; the ETF keeps trading.
• Consistency — one data source for the entire whitelist.
• Teaching focus — you learn the *strategy logic*, not ticker-specific noise.

What each class actually does under stress

Real example: 2022 broke the classic hedge

For two decades, "stocks fall so bonds rise" was close enough to true — SPY and TLT spent most of 2000–2020 negatively correlated, which is exactly why 60/40 portfolios felt safe. Then inflation arrived. In calendar 2022 the S&P 500 lost roughly 18% *and* long Treasuries lost roughly 31%, giving a 60/40 investor something like -17% with no shelter anywhere. Meanwhile energy was the best sector in the market: XLE returned roughly +64% while XLK fell about -28%.

Two durable lessons come out of that year. First, correlation is a regime property, not a constant — a strategy that assumes bonds will always hedge equities is assuming a monetary regime. Second, **diversification across *drivers* beats diversification across tickers**: holding SPY, QQQ and XLK is one bet wearing three names, while holding SPY, TLT, GLD and DBC is four.

Real example: the year gold earned its place

In 2008 SPY returned roughly -37% while GLD returned roughly +5%. That single year is why gold appears in so many RISK_ON_OFF books. But 2011–2015 is the counterweight: gold fell about 45% from its peak while equities rose steadily. An asset that helps in crises and hurts in expansions is doing its job — you just have to size it as insurance rather than as a return engine.

Named books in code

from backtest.universes import US_SECTORS_FULL, BONDS, MULTI_ASSET, RISK_ON_OFF

# Books are plain Python lists — inspect, slice, or combine them
# RISK_ON_OFF   = ["SPY", "QQQ", "TLT", "IEF", "GLD", "BIL"]
# MULTI_ASSET   = ["SPY", "EFA", "EEM", "VNQ", "DBC", "GLD", "TLT", "IEF", "HYG"]

# Sector rotation plus a safety sleeve (lab limit: 15 symbols)
ASSETS = ["XLK", "XLF", "XLE", "XLV", "TLT", "GLD", "BIL"]

Costs are part of the asset, not a footnote