Breadth Divergence
Now: 39.5% above the 50-day, 2.09% at new lows, 1.9% at new highs.
▶ View the live, interactive version with full charts →
A drop in the share of stocks above their 50-day means two different things, and the new-low count on the same basket says which. Measured on 535 names back to 2007: a 10-point breadth fall in five sessions is followed by a month positive 62% of the time against a 67% baseline — worse than nothing on its own. Split it by new lows and the halves separate, 66%/79% when new lows stay under 1% of the basket against 52%/64% when they clear 3%. The split survives holding the breadth level constant, so it is not a level proxy. Also reports the two findings that cut the other way: the signal is asymmetric, and drops that leave breadth mid-range resolve worse than ones that wash it out.
- 4954 sessions since 2007-01 on 527 common stocks.
- Now: 39.5% above the 50-day, 2.09% at new lows, 1.9% at new highs.
- A breadth drop with quiet new lows resolves above baseline; the same drop with new lows confirming resolves well below it.
When to run this study
Run after any sharp breadth deterioration — the study exists to answer "is this a pullback or a breakdown", and the new-low count is what settles it. Also worth a look when breadth is fine but new lows are creeping up, which is the configuration that has historically preceded trouble. Not an entry trigger: there is no position to size off a breadth reading.
Historical results
Sharp breadth drops, split by the new-low count
| Condition | Events | 1M | 1M win | 3M | 3M win |
|---|---|---|---|---|---|
| ALL SESSIONS (baseline) | 4933 | 1.62 | 67 | 4.15 | 73 |
| Sharp breadth drop, unconditional | 233 | 1.47 | 62 | 4.90 | 72 |
| Drop + new lows quiet (<1%) | 140 | 1.86 | 65 | 5.27 | 78 |
| Drop + new lows confirming (>=3%) | 86 | 0.86 | 53 | 4.13 | 65 |
| Drop but breadth still >=55% | 114 | 1.14 | 61 | 4.10 | 69 |
| Drop washing breadth under 40% | 104 | 2.54 | 67 | 4.56 | 76 |