BUILD: COURSE 1 | LESSON 3

Momentum: RSI, MACD, stochastic

Learning objectives

Explain what RSI, MACD and the stochastic oscillator actually measure, and read their standard settings critically
Distinguish the trap of "overbought/oversold" in trends from the legitimate uses: regime-appropriate signals and divergence
Combine one momentum tool with trend structure instead of stacking three oscillators that say the same thing

Momentum: the speedometer, not the satnav

Momentum indicators measure how fast price has been moving and whether that speed is changing. They do not know where price is going. Think of them as a speedometer: useful for noticing the car is decelerating, useless for telling you where the driver lives.

That framing kills the most expensive beginner belief up front: "overbought" does not mean "about to fall". In a strong trend, momentum indicators go to extremes and stay there — that's what a strong trend is. Some of the best trends of the decade spent weeks "overbought". Selling something because RSI hit 70 is, in a trending market, selling something because it is going up.

With that warning installed, let's take the three tools one at a time.

RSI: measuring the balance of recent gains and losses

The Relative Strength Index (Wilder, 1978) compares average gains to average losses over the last N periods (default 14):

RSI = 100 − 100 / (1 + RS), where RS = average gain ÷ average loss over N periods

Intuition: if EURUSD's up days and down days over the last 14 days were equal in size, RS = 1 and RSI = 50. If up moves have been twice the size of down moves, RS = 2 and RSI ≈ 67. All gains and no losses pushes RSI toward 100; all losses toward 0. RSI is just a 0–100 scoreboard of who has been winning recently and by how much.

Standard reading: above 70 = overbought, below 30 = oversold. The trade-offs: RSI(14) on a daily chart moves slowly and flags genuine extremes; RSI(7) or RSI(2) is far twitchier — RSI(2) is used in some mean-reversion research precisely because it fires constantly. Longer settings mean fewer, later, more meaningful signals. Same lag-versus-noise dial as everything else.

The regime rule that makes RSI usable:

In a range, fading extremes has a genuine statistical basis: buy sub-30 readings near range support, sell 70+ near range resistance.
In a trend, extremes are confirmation, not warning. Trend traders instead buy pullbacks — RSI dipping to 40–50 in an uptrend — or use Wilder's own trick: in strong uptrends RSI tends to floor near 40 rather than 30.

Divergence is RSI's higher-value use. Price makes a higher high; RSI makes a lower high — the second push was slower than the first. On GBPUSD: price prints 1.2780 with RSI 72, pulls back, then grinds to 1.2810 with RSI only 64. The new high came on fading force — a warning that pairs naturally with the "failure to make a higher high" logic from Lesson 1. Honest caveat: divergences fail routinely — strong trends print divergence after divergence while marching on ("divergences can stay divergent longer than you can stay solvent"). Tested alone, divergence hit rates are modest — think "somewhat better than a coin flip in the right context", not a secret weapon. It's a warning light, tradeable only with structure confirmation and a stop.

MACD: trend and momentum in one window

The Moving Average Convergence Divergence is built from Lesson 2's parts:

MACD line = EMA(12) − EMA(26) · Signal line = EMA(9) of the MACD line · Histogram = MACD line − signal line

Intuition for the default 12, 26, 9: the MACD line is the gap between a fast and a slow EMA. When price accelerates upward, the 12-EMA pulls away from the 26-EMA and the MACD line rises. When the trend tires, the gap narrows — the two averages converge — and MACD rolls over before the slow average itself turns. The signal line smooths the MACD line, and the histogram shows whether the gap between them is widening or shrinking — momentum's momentum.

Three standard readings, in rising order of noise:

Zero-line side (slowest, most reliable): MACD above zero means the 12-EMA is above the 26-EMA — an uptrend regime filter.
Signal-line crossovers: earlier but frequent; in ranges they whipsaw exactly like the MA crosses they secretly are.
Histogram shrinking: the earliest and noisiest tell that a move is decelerating.

Worked example on USDJPY daily: price runs 152.00 → 157.00; MACD line climbs from +0.20 to +0.85. Price then pushes to a marginal new high at 157.40, but MACD tops at +0.70 and the histogram has been shrinking for six sessions. That's deceleration and divergence — grounds to tighten stops on longs, not grounds to short a pair making new highs. If structure then breaks (price takes out the last higher low at, say, 155.60), the momentum evidence and the structural evidence agree, and now there's a trade with a defined invalidation.

MACD's limitation is inherited from its parents: it's EMAs all the way down, so it lags, and it has no fixed scale — "+0.85" means nothing across instruments, which is why overbought/oversold language doesn't apply to it.

Stochastic: where did we close within the recent range?

The stochastic oscillator (Lane) asks a different question: where is the current close relative to the high–low range of the last N periods?

%K = 100 × (Close − Lowest Low(N)) ÷ (Highest High(N) − Lowest Low(N)) · %D = 3-period SMA of %K

With the default 14, 3, 3 (14 lookback, %K smoothed by 3, %D a 3-SMA of that — "slow stochastic"): a reading of 80 means you're closing in the top 20% of the two-week range; 20 means the bottom 20%. Overbought/oversold thresholds are 80/20.

Because it's range-anchored, stochastic is the natural range-market tool: in a sideways EURUSD between 1.0700 and 1.0800, %K crossing down through %D above 80 as price tags 1.0790 is a coherent fade setup — you're selling the top of a defined range with a stop above it. In trends, stochastic is the worst of the three to fade: it pins above 80 for the entire move. It's also the fastest and noisiest of the trio — hence the smoothing, and hence why many traders use it only on higher timeframes or only in confirmed ranges.

Using them together (which mostly means: don't)

RSI, MACD and stochastic are three flavours of the same ingredient — recent price change. Stacking all three and calling agreement "confluence" is counting one witness three times. A cleaner recipe:

One regime tool (structure from Lesson 1, or the 200-MA / MACD zero line) to say trend or range.
One momentum tool matched to the regime: RSI pullbacks or MACD in trends; stochastic at the edges of ranges.
Structure for entries and stops. The oscillator never sets your risk; the chart does.

And the standing disclaimers, because this is the anti-hype academy: every formula above is public, decades old, and coded into every algo on the planet — none of them is an edge by itself. Their published standalone hit rates hover near chance. What they genuinely do is describe momentum conditions faster and more consistently than your eyes, which improves the quality of decisions whose profitability still depends on regime, risk-reward and position sizing.

Key takeaways

Momentum tools measure speed, not destination; "overbought" in a strong trend is confirmation, not a sell signal.
RSI = 100 − 100/(1+RS): a 0–100 scoreboard of average gains vs losses (default 14; 70/30 thresholds). Fade extremes only in ranges; in trends, buy RSI pullbacks and watch divergence — knowing divergences often fail.
MACD (12, 26, 9) is the gap between two EMAs plus a signal line: zero-line for regime, crossovers for timing, histogram for deceleration — and it lags, because it's made of EMAs.
Stochastic (14, 3, 3) locates the close within the recent range (80/20); strongest at defined range edges, dangerous to fade in trends.
The three are correlated — pick one, match it to the regime, and let structure define entry, stop, and size. No oscillator signal works alone.