WaveTrend Oscillator

A community-famous momentum oscillator — essentially a smoothed, EMA-normalized cousin of the CCI — with two lines and overbought/oversold zones near ±60.

What it is

The WaveTrend Oscillator is one of the most-copied community scripts on TradingView, published by the user LazyBear, and it circulates almost entirely without formula documentation — which is why we reverse-engineer it here. Under the hood it is a smoothed variant of the Commodity Channel Index idea: it measures how far the typical price (hlc3) sits from its own exponential average, scales that distance by a smoothed measure of typical deviation (using the same 0.015 constant CCI uses), and then smooths the result again into two lines — a faster WT1 and a slower WT2 (a 4-period SMA of WT1). The output oscillates around zero, with ±60 commonly drawn as extreme zones, though the scale is not strictly bounded. Because every stage is an average of past prices, WaveTrend is a lagging, descriptive tool: it summarizes recent momentum and says nothing certain about future price. Popularity is not validation — a heavily-used script is not a more predictive one. This entry is educational only, not financial advice, and all trading involves risk of loss.

How it works

The calculation runs in five deterministic steps. First, take the typical price: ap = (high + low + close) / 3. Second, compute an EMA of it over the channel length (default 10): esa = ta.ema(ap, 10). Third, measure typical deviation the same way: d = ta.ema(math.abs(ap − esa), 10) — an EMA of the absolute distance between price and its average. Fourth, form the normalized channel index: ci = (ap − esa) / (0.015 × d). This is structurally the CCI formula with EMAs substituted for the SMA and mean absolute deviation, which is why WaveTrend behaves like a smoother CCI. Fifth, smooth ci with a longer EMA (default 21) to get WT1, and take a 4-period SMA of WT1 to get WT2 as a signal line. Crossings of WT1 and WT2, and visits to the ±60 region, are what most published scripts trade off. Two implementation notes matter for correctness: the division blows up when d approaches zero (flat, zero-range bars — thin sessions, illiquid symbols), so robust code guards that case; and clones differ — some swap hlc3 for close, alter the 0.015 constant, or add extra smoothing, so two 'WaveTrends' on the same chart can legitimately disagree.

How traders read it

• Zero line: WT1 persistently above zero means typical price has been holding above its own EMA (upward momentum); persistently below zero, the reverse.
• ±60 zones: readings beyond roughly +60 or −60 are conventionally read as stretched momentum. The oscillator is not strictly bounded, so these are habits from the original script's plot lines, not mathematical limits.
• WT1/WT2 crossings: WT1 crossing above WT2 is read as short-term momentum turning up relative to its own average, and crossing below as turning down. These are lagging, descriptive events — not instructions to act.
• Crossings inside extreme zones: many users pay attention only to crosses occurring beyond ±60, treating mid-range crosses as noise; that filter reduces frequency but does not create predictive power.
• Divergence: price making a new extreme while WT1 does not is read by some as fading momentum. Like all divergence reading, it can persist for many bars without a reversal.
• Whatever the pattern, pair it with price structure and defined risk — WaveTrend describes what momentum has done, not what price will do.

Common settings

The canonical LazyBear defaults are channel length 10 (for the esa and d EMAs), average length 21 (for WT1), and a fixed 4-period SMA for WT2, computed on hlc3, with guide lines at ±60 and sometimes ±53. Shorter channel lengths make the oscillator jumpier; longer average lengths smooth WT1 at the cost of more lag. Because clones vary the source, constants, and smoothing, state which variant you use when comparing results — identical settings labels do not guarantee identical formulas.

Strengths

• Fully deterministic and reproducible once the formula is written down — every stage is a standard EMA/SMA operation that can be verified line by line.
• Smoother than raw CCI on the same data, so its swings are easier to read visually.
• The two-line structure gives a built-in reference (WT2) for reading changes in WT1 without picking arbitrary thresholds.
• Self-normalizing: dividing by smoothed deviation makes readings loosely comparable across instruments and timeframes, unlike raw price-difference oscillators.
• Works from ordinary OHLC data with no special feeds.

Pitfalls to watch

• No official documentation: the 'real' WaveTrend is whatever the original script computed, and the many clones diverge in source, constants, and extra smoothing — two charts labeled WaveTrend can disagree.
• Division-by-near-zero: when bars are flat, the deviation term d collapses toward zero and ci can spike enormously; implementations without a guard produce absurd readings on illiquid symbols.
• Lag at every stage: four separate smoothing operations mean WT1/WT2 crossings confirm momentum shifts well after they begin.
• Whipsaw in ranges: mid-zone crossings flip frequently in sideways markets, producing many conflicting readings.
• The ±60 levels are conventions, not statistics: nothing is calibrated about them, and momentum can stay 'extreme' for long stretches in strong trends.
• No predictive power: WaveTrend is arithmetic on past prices. Its popularity does not make it predictive, and no indicator removes the risk of loss.

Pine v6 example

//@version=6
indicator("WaveTrend Example", overlay = false)

chLen  = input.int(10, "Channel Length", minval = 1)
avgLen = input.int(21, "Average Length", minval = 1)
src    = input.source(hlc3, "Source")

esa = ta.ema(src, chLen)
d   = ta.ema(math.abs(src - esa), chLen)
ci  = d != 0 ? (src - esa) / (0.015 * d) : 0.0  // guard flat bars
wt1 = ta.ema(ci, avgLen)
wt2 = ta.sma(wt1, 4)

plot(wt1, "WT1", color = color.blue)
plot(wt2, "WT2", color = color.orange)
hline(60,  "Upper", color = color.gray)
hline(0,   "Zero",  color = color.new(color.gray, 50))
hline(-60, "Lower", color = color.gray)

crossUp = ta.crossover(wt1, wt2) and barstate.isconfirmed
plotshape(crossUp, "WT1 x WT2 up", style = shape.triangleup, location = location.bottom, color = color.teal, size = size.tiny)

Pro tip: If you compare a WaveTrend port against the original, test on a symbol with occasional flat bars — that is where unguarded implementations diverge wildly because of the near-zero deviation term. And always gate any alert or marker on barstate.isconfirmed, since WT1/WT2 can cross and uncross repeatedly within a live bar. Educational context only, not financial advice: WaveTrend describes past momentum, predicts nothing, and trading always involves risk of loss.

Educational only — not financial advice, not a recommendation to trade. No indicator is predictive; trading involves substantial risk of loss.