Does weak breadth under a market at its high predict a worse stock market? A backtest
A backtest of weak market breadth and lagging credit at an index high: the question, the bar written first, the verdict and its limits. Coil Edge Research.
Hypothetical results from a simulation on historical market data; no money was traded; costs and slippage as assumed in the Method; past and simulated results don't predict future results. Not investment advice.
The question. When a stock index sits near its high while few of its stocks are above their long-term averages (or while high-yield bonds lag investment-grade bonds), do stocks return less over the next 20 and 60 sessions, or fall further? This tests a widely shared claim.
The verdict. Not enough evidence either way. The registered test could not be read: its setup, the index within 2% of its high with breadth in the bottom tenth of its own history, did not occur on any of the 1,226 near-high sessions from 2017 to mid-2026. Looser versions and the credit version had only five to thirteen episodes and pointed in no stable direction. If anything, weaker breadth at a high was followed by slightly better returns, not significantly. That is no evidence for the claim, and too few episodes to say more.
Backtest (hypothetical), run in Backtrader 1.9.78.123, data to 2026-06-01
Method, in one paragraph. 9.4 years of daily bars (2017-01-03 to 2026-06-01) for S&P 500 members, SPY, QQQ, HYG and LQD. A hypothetical buy at the next open and sale 20 or 60 sessions later, net of assumed costs, compared with near-high sessions that had healthy breadth and with a placebo of random near-high sessions.
What this does not show.
- Five, seven and thirteen episodes, with two to four in the out-of-sample years, are too few to say the effect is zero or that it is real.
- It measures buying the index after a signal. It is not a test of any trading rule.
- The breadth data ends 2026-06-01, so today's reading is not computed.
- A simulation is not a record of trades.
The full result tables, the control comparisons and the test statistics are in the paid text. A person pays from a crypto wallet (USDC on Base); an AI agent pays over x402. There is no card option for a single study.
Buy the full study ($0.25, pay with a crypto wallet)
Does weak breadth under a market at its high predict a worse stock market?
Not enough evidence either way. The registered setup never occurred in this backtest's window, and the looser versions had too few episodes. If anything, weaker breadth at a high was followed by slightly better returns, not significantly.
Coil Edge Research, 2026-10-22.