Money Flow Quadrants: Real vs Fragile Rallies
What this does
Tells you which sectors are genuinely attracting capital, and which are only rising because nobody is selling. Those two look identical if you only watch price.
The rule it will not break
Trading value is not capital inflow. Every trade has a buyer and a seller in equal size. If trading value doubles, the money in the market has not doubled.
Net flow is observable in exactly three places: investor-type net buy/sell data, ETF creations and redemptions, and margin balance changes. Everything else is attention, not money.
Most commentary conflates the two. This skill writes "attention is concentrating" and "capital is flowing in" as separate claims, always.
The method
Flow concentration = sector share of trading value / sector share of market cap
Momentum = 5-day average / 20-day average
Then cross concentration against price:
| Price up | Price down | |
|---|---|---|
| Concentration up | ① Real inflow | ② Capitulation |
| Concentration down | ③ Fragile — no new money | ④ Abandoned |
Quadrant three is where the edge is. Most people watch price alone, so they cannot distinguish a rally with buyers from a rally that is simply missing sellers. The second one breaks on any real supply.
Sample output
■ Asset class
Risk off. Yields up, gold bid, volatility rising.
■ Region
Investor-type: cash -286bn, futures -165bn. Both directions aligned,
so this is position reduction, not hedging. First week — not yet a trend.
■ Sectors
Pharma concentration 1.31 (20d +18%) price +1.8% → ① real inflow
Electronics concentration 1.62 (20d +24%) price -2.8% → ② capitulation
Insurance concentration 0.78 (20d -11%) price +1.3% → ③ fragile
Steel concentration 0.61 (20d -14%) price -0.4% → ④ abandoned
■ The rotation
Out of semis, into defensives. But total trading value FELL on a heavy down
day — if money were rotating between sectors, both sides would show volume
and the total would rise. It fell, so the money left the market rather than
moving inside it. The defensive bid is quadrant three, not quadrant one.
Four layers
Asset class → region → sector → single name. Always descend in that order. A correct sector call fails if the region above it is bleeding.
Persistence test
- Same direction three consecutive weeks
- Concentration momentum above 1.2 for two weeks or more
- Consistent with the layers above
And exclude the days where supply and demand are distorted mechanically: expiry, index rebalancing, month and quarter-end, ex-dividend, earnings dates.
Coverage
Japan: investor-type net flows from the exchange (cash and futures read together), sector statistics, margin balances, short-selling ratios.
Global: sector and regional ETF flows, fund net issuance, futures positioning, fund manager survey positioning.
Reading cash and futures together matters. When they move in opposite directions it is hedging. When they align, the position itself is changing — a much heavier signal, and one rarely mentioned.
FAQ
How often should this run?
Weekly. Investor-type data updates weekly, and the persistence test needs three weeks to mean anything.
Does it recommend trades?
No. It describes where capital is going and what that implies. Decisions are yours.
What if data is missing?
It says so explicitly and states how the gap affects the conclusion. Nothing is filled in by guessing.
