How the scan works

The rule is the same one a chartist applies by eye, written down so it runs the same way every time.

• Find the swing points. A bar is a swing high when its close is above the closes on both sides of it (5 bars each side on 60-minute and daily charts, 3 bars each side on weekly, 2 on monthly). Swing lows are the mirror image.
• Compare the last two. For each new swing high we look back (up to about 60 bars) for the previous swing high that price has now cleared — the new peak must be the highest close since that earlier one, so a lower high inside a bigger move never counts. If price made a higher high but the indicator made a lower high, that is a bearish divergence. For swing lows: price lower low, indicator higher low = bullish divergence.
• Two indicators, checked separately. RSI(14) and the MACD line (12, 26, 9). A divergence that shows on both carries more weight than one that shows on just one.
• Confirmed vs. forming. A swing high only becomes official once price has pulled back for the required number of bars. Until then the reading is labelled forming: price is at a new high right now but momentum is already lower than at the previous peak. Forming readings can disappear if price keeps pushing and momentum catches up.
• Active until price settles it. A confirmed bearish divergence stays in force until price either closes above the second peak (a new high cancels it, though a fresh divergence may form at that new peak) or closes below the lowest point between the two peaks (the pattern has played out). Bullish is the mirror image. Each card shows both levels. Because the breakdown is the moment the warning matters most, a played-out divergence stays on the scan as playing out for a short while afterwards (12 hourly bars, 10 daily, 6 weekly, 3 monthly), and the card notes when the latest MACD crossover agrees with it. A divergence that does nothing for a long stretch (about 60 daily bars, 52 weeks, 24 months, 70 hourly bars) drops off as expired. Settled divergences still draw on the charts.
• Monthly bars cover the last ten years. Yahoo's longer monthly history for the indices has months missing, which would corrupt the indicators, so the monthly scan uses the ten-year feed that comes back complete. The first three years of it are spent warming up RSI and MACD, so the monthly chart shows roughly the last seven.

Buy / sell signals follow a separate, simpler rule:

• Buy when the MACD line crosses above its signal line after spending at least 10 closed bars below it. Sell when it crosses below after at least 10 closed bars above it. Crosses that come sooner are ignored, which filters out the quick back-and-forth wobbles when the two lines run close together.
• RSI has to have turned first. For a sell, RSI must already be below its high of the recent window (20 daily bars, 12 weekly, 24 hourly, 10 monthly) when the cross comes; for a buy, above its low. RSI does not have to have been overbought or oversold — in practice almost every 10-bar cross passes this test, because by then RSI has long since turned. What the RSI adds is context: each card and chart shows the reading the turn came from, and a turn from at or beyond 70 / 30 is marked as the stronger version of the signal.
• Setup armed. When RSI has already turned but the MACD has not crossed yet, the card says so and tells you how many more bars the lines need before a cross would count.
• Confirmed on the next bar. A cross only counts once the bar it happened on has closed, so the marker sits on the bar after the cross: the next hour, the next day, or the first bar of the next week or month. A qualifying cross on the still-open bar is shown as pending and becomes a signal only if the bar closes that way.
• Fresh vs. old. A signal card is highlighted while the signal is recent (same windows as above); older signals stay listed in grey so you can see where the last one was.

How to read it. Divergences are a warning, not a timing tool. Bull markets can carry a bearish divergence for weeks. The pattern is strongest when it shows on the daily and weekly chart at the same time, on more than one index, and on both indicators. A 60-minute divergence alone is short-term noise and usually resolves within a few sessions. A monthly divergence is the rarest and slowest of all: it can take a year to play out, but it is the chart the biggest turns have tended to show up on, and one that lines up with the weekly deserves the most respect of anything on this page.