Trend direction: define bullish, bearish and neutral states

Choose a definition you can inspect

These methods classify different features of the same history. None is universally best for a particular asset class.

A three-state rule on one chart timeframe

This example calls a completed bar bullish only when close > baseline EMA and fast EMA > slow EMA. It calls it bearish only when both inequalities reverse. Every other available case is neutral. It uses no ATR, slope test or higher-timeframe request.

Pine Script v6: display only completed-bar states

The code enforces fast < slow < baseline lengths and waits at least baselineLength chart bars. That gate is a display policy, not proof that initialization effects have disappeared. The developing bar is blank, and the Data Window shows 1, -1 or 0 only for an available confirmed state.

//@version=6
indicator("Confirmed EMA Direction [Pineify]", overlay=true)
int fastLength = input.int(20, "Fast EMA length", minval=1)
int slowLength = input.int(50, "Slow EMA length", minval=2)
int baselineLength = input.int(200, "Baseline EMA length", minval=3)
if fastLength >= slowLength or slowLength >= baselineLength
    runtime.error("Require fast < slow < baseline lengths.")
float fastEma = ta.ema(close, fastLength)
float slowEma = ta.ema(close, slowLength)
float baseline = ta.ema(close, baselineLength)
bool ready = bar_index >= baselineLength - 1 and not na(baseline) and not na(fastEma) and not na(slowEma)
bool confirmed = ready and barstate.isconfirmed
int state = close > baseline and fastEma > slowEma ? 1 : close < baseline and fastEma < slowEma ? -1 : 0
color stateColor = state == 1 ? color.teal : state == -1 ? color.orange : color.gray
pFast = plot(confirmed ? fastEma : na, "Fast EMA", color=color.teal, style=plot.style_linebr)
pSlow = plot(confirmed ? slowEma : na, "Slow EMA", color=color.orange, style=plot.style_linebr)
plot(confirmed ? baseline : na, "Baseline EMA", color=color.blue, linewidth=2, style=plot.style_linebr)
fill(pFast, pSlow, color=confirmed ? color.new(stateColor, 80) : na, fillgaps=false)
plot(confirmed ? state : na, "State: 1 bullish, -1 bearish, 0 neutral", display=display.data_window)
bgcolor(confirmed ? color.new(stateColor, 93) : na)

Separate a direction filter from a trade

A filter can restrict another setup, but it does not define an order. For example, a study could require the bullish state before evaluating a breakout. Specify the breakout, entry fill, invalidation, quantity and costs before measuring results.

• A neutral state is not a short signal. A blank developing or insufficient-history bar is not neutral evidence.
• Check reversals after a long trend and repeated switches in a range. Slower lengths can reduce switching while increasing reaction delay; the trade-off needs measurement.
• The same 200-bar length spans different elapsed time on daily and five-minute charts. Keep symbol, session and adjustment settings fixed.
• If extending this to weekly data on daily bars, define completed-week timing separately. Three EMA lengths on a daily chart do not establish multi-timeframe agreement.

Verify the calculation before comparing results

Compare the three EMA values and Data Window state with the rule table on selected historical bars, then observe a developing bar through its close. This page supplies a specification and indicator code; it does not report a backtest.

• Test strict equality and conflicting comparisons as neutral cases.
• Use one frozen definition when comparing a strategy with and without the filter. Do not change the definition to remove losses after seeing them.
• A volume spike or a green cloud does not identify the buyer. Keep claims about participants separate from price-derived indicators.