RSI, ADX, MACD: what they actually measure.

A doctor checking your vitals reads four numbers — pulse, blood pressure, oxygen, temperature. None of them tell you what's wrong by themselves. Together they triangulate. Misread any single one in isolation and you'll panic about a high pulse during cardio or a low temperature in a cold room. Indicators are the same. RSI is a speedometer for recent momentum. ADX is a wind-strength meter for trend. MACD is two moving averages in a trench coat. Each measures something specific. None of them tell you what to do. Most retail trading content treats them as triggers. They're not. They're confirmations — and most of the time they're confirming something the structure already told you weeks earlier.

What an oscillator actually is

An "oscillator" is a bounded numerical summary of recent price action. Bounded matters: the value lives between fixed limits (RSI between 0 and 100, ADX between 0 and 100, MACD swings around zero with no fixed bounds but generally small). The bounds make the value comparable across stocks and across time — a 70 RSI on NVDA today and a 70 RSI on AAPL three years ago are comparable in a way that raw prices aren't. That's the genuine virtue. The problem starts when retail traders treat the bounded value as a signal rather than a summary. The number summarizes the past. It does not predict the future.

The honest frame: an oscillator value is the answer to a specific question about where price has been, computed in a specific way. The question matters; the computation matters. If you use the value without knowing the question, you're using astrology with extra steps.

RSI: a speedometer for momentum

Relative Strength Index, 14-day default. The math:

avg_gain  = mean of UP closes over last 14 days
avg_loss  = mean of DOWN closes over last 14 days (absolute value)
RS        = avg_gain / avg_loss
RSI       = 100 − (100 / (1 + RS))

Translation: RSI is "what fraction of the recent move was up." If every one of the last 14 days closed higher, RSI is 100. If every one closed lower, RSI is 0. If the recent up-days and down-days roughly balance, RSI is around 50. The 70 line is "the recent move has been mostly up"; the 30 line is "the recent move has been mostly down." That's the honest reading. Not "overbought" or "oversold" — those words imply a future direction, which the math does not contain.

The retail trap is the 70 = sell, 30 = buy heuristic. It works in mean-reverting choppy stocks. It fails catastrophically in trends. NVDA spent stretches of late 2024 with daily RSI above 70 for weeks; the move kept going. Anyone who shorted at "overbought" got run over. RSI 70 in a strong uptrend means "the trend is healthy and momentum is real." RSI 70 in a chop means "expect mean-reversion." The number is the same. The interpretation depends on the structure. Always.

Where RSI is genuinely useful: divergence. Price makes a higher high; RSI makes a lower high. The momentum that drove the prior peak isn't there anymore. That's a real signal — not a trigger, but a flag. The reverse for downtrends. The framework reads RSI divergence as one input to the Trap & Structure Coach, weighted as a confirmation alongside structure and volume.

ADX: how strong is the wind

Average Directional Index, 14-day default. The math is uglier than RSI's — directional movement plus and minus indices smoothed and combined into a single 0-100 reading. The conceptual output is what matters:

ADX measures TREND STRENGTH, not direction.
ADX < 20: choppy market, no trend
ADX 20-25: weak trend forming
ADX 25-50: trending market
ADX > 50: very strong trend (rare, usually unsustainable)

The thing to internalize is that ADX is direction-blind. A reading of 35 doesn't tell you up or down. It tells you "whichever direction the trend is going, it's trending well." You need separate evidence for direction (the swing structure from lesson 8 does this cleanly).

What ADX is genuinely useful for: filtering chop. Most retail trading systems work great in trends and lose money in chop. ADX is the most reliable filter for "is there a trend at all." Below 20, mean-reversion strategies are licensed and trend strategies will get sawed up. Above 25, trend strategies are licensed and mean-reversion strategies will keep getting steamrolled by continuations. The framework uses ADX as one of the trend-state inputs to refuse trades in chop, alongside the structural read.

What ADX is not useful for: timing entries within a trend. By the time ADX climbs to 30+, the trend has been visible in the swings for weeks. ADX confirms; it doesn't enter.

MACD: two moving averages in a trench coat

Moving Average Convergence Divergence, 12/26/9 default. The math:

MACD line   = 12-day EMA of close − 26-day EMA of close
signal line = 9-day EMA of MACD line
histogram   = MACD line − signal line

Translation: it's the difference between a fast-smoothed price line and a slow-smoothed price line, with a third smoothing layered on top to "confirm" crossovers. Two moving averages in a trench coat, plus a third that tells you when the first two are agreeing. The whole stack is downstream of price by definition — every smoothing layer adds lag.

The retail trap is the "MACD cross" as a buy or sell signal. The cross happens after the underlying trend change has been visible in the swings for weeks (see lesson 8 — moving-average crossovers are the cousin of cargo-cult candle patterns). By the time the MACD line crosses the signal line, you're entering on news that the chart already digested.

Where MACD is mildly useful: divergence, like RSI. Price makes a higher high; MACD makes a lower high. Momentum is fading even as price extends. The framework reads MACD divergence as a low-weight confirmation, never as a trigger. Anyone who builds a trading system around MACD crossovers is building around lag.

The indicator stack trap

Open most retail trading content and you'll see charts with eight indicators stacked: RSI, MACD, Stochastic, CCI, Williams %R, OBV, MFI, and a couple of moving averages. The implicit claim is that more measurements = more confidence. The actual math is the opposite. RSI, Stochastic, CCI, and Williams %R all measure momentum, all derived from the same recent price action, all heavily correlated. Stacking them on a chart gives you four versions of the same signal, weighted by your eyeball in inscrutable ways. It feels like confluence; it's redundancy.

The discipline is one indicator per question. Trend strength? ADX. Momentum divergence? RSI or MACD, not both. Volume confirmation? OBV or volume itself. Stack one of each and you have four genuinely different reads. Stack four momentum oscillators and you have one read with extra steps.

Confirmation, not trigger

The single most important shift in how retail traders should read indicators is treating them as confirmation rather than trigger. A trigger is "this number crossed this line, place the order." A confirmation is "the structure says X; the indicator agrees, weakly disagrees, or strongly disagrees." The trigger framing makes the indicator the entry decision. The confirmation framing makes the indicator one input among several, with the structural read upstream of all of them.

Lesson 8 made the case that swing structure is upstream of moving-average signals — by the time the MA crosses, the structure has already changed. The same logic applies to oscillators. The structural read precedes every indicator reading the chart can compute. The indicators say "yes, this is real" or "this looks weaker than the swing structure suggests" or "the structure has changed but I haven't reacted yet." All useful. None of them is the entry.

What the framework does with each

• RSI — read for divergence at structural levels. RSI 70 alone means nothing. RSI 70 with bearish divergence at confluence resistance is a meaningful read. The framework's Trap & Structure Coach uses divergence as a weighted confirmation alongside the candle and the level itself.
• ADX — used as a coarse trend-state filter. ADX below 20 routes the system into mean-reversion mode (or, more often, refuses entries entirely because chop is the enemy of swing trading). ADX above 25 licenses trend continuation reads. Direction comes from the swing structure.
• MACD — read for divergence only. MACD crossovers are not used as triggers. The framework treats MACD as the slowest, laggiest of the three; it makes a poor primary signal and a fine background sanity check.

What about the rest

Stochastic, CCI, Williams %R, MFI, OBV, Bollinger Bands, Ichimoku Cloud, and the long tail of named indicators each measure something specific (most of them measure variations on momentum or volume), each have their adherents, and each get oversold by content creators as primary triggers. The disciplined frame is the same for all of them: what specific question does this indicator answer? If you can't state the question in plain English, you don't know what the indicator measures and you shouldn't be reading it. If you can, ask whether the question is one the structure already answered.

Most of the time, the answer is yes. Indicators are downstream of structure. Reading the structure first and then sanity-checking with one indicator per genuinely-different question is the cleanest workflow. Reading eight indicators and trying to weigh them by feel is how retail traders convince themselves that their gut is signal.

The real lesson

Don't make decisions from a numerator. The number is a summary. The summary answered a question. The question was specific. Know the question. Know the limits. Know that "RSI 70" doesn't mean "sell" any more than "60 mph" means "you're going too fast" — it depends on the road. The road is the structure. The indicator is the speedometer. Both are useful. Neither drives the car.

Related: Lesson 7 — Reading candles · Lesson 8 — HH/HL trend structure · Trap & Structure Coach (docs)