Multi-timeframe strategy: time confluence.

Imagine two traders with the same moving-average crossover strategy. One trades looking only at the M5 and takes trades in every direction. The other only accepts the M5 signals that are in line with the H1 trend. The second one will trade less — and better. That's the essence of multi-timeframe: use the higher chart as a compass and the lower one as a trigger. It's one of the most effective ways to filter out bad trades.

01The single-timeframe mistake

Anyone trading off a single chart falls into a trap: a buy signal on the M5 may be perfectly aligned in the short term, but be just a bounce inside a downtrend on the H1. You buy, the bounce ends, and the larger trend swallows you. A single timeframe leaves you blind to context.

02The concept: higher trend, lower entry

The structure is always the same — two (sometimes three) timeframes with distinct roles:

03Which timeframes to combine

The practical rule is a ratio of about 4 to 6 times between them — large enough to give different context, close enough to make sense. Common combinations:

• Day trading: M15 (trend) + M3 or M5 (entry)
• Wider intraday: H1 (trend) + M15 (entry)
• Swing: Daily (trend) + H4 (entry)

Some traders use three: a top one (macro direction), a middle one (operational trend) and a low one (trigger). More than three becomes paralysis — too much information, too little decision.

04Coding the confluence

The bot needs to pull two timeframes and cross-reference the information. Here it is on both platforms:

//@version=5
strategy("Multi-Timeframe", overlay=true)

// trend on the HIGHER timeframe (e.g. 60 min)
htf = input.timeframe("60", "Trend timeframe")
ma200_htf = request.security(syminfo.tickerid, htf, ta.sma(close, 200))
close_htf = request.security(syminfo.tickerid, htf, close)
uptrend = close_htf > ma200_htf

// trigger on the CURRENT timeframe (the chart's, lower)
fast_ma = ta.ema(close, 9)
slow_ma = ta.ema(close, 21)
buy_trigger = ta.crossover(fast_ma, slow_ma)

// CONFLUENCE: only buy if trigger AND higher trend agree
if buy_trigger and uptrend
    strategy.entry("Buy", strategy.long)

# multi_timeframe.py
def trend_direction(df_htf):
    # trend on the higher chart: price vs the 200 average
    ma200 = df_htf["close"].rolling(200).mean().iloc[-1]
    return "UP" if df_htf["close"].iloc[-1] > ma200 else "DOWN"

def entry_trigger(df_ltf):
    # trigger on the lower chart: EMA crossover
    f = df_ltf["close"].ewm(span=9).mean()
    s = df_ltf["close"].ewm(span=21).mean()
    crossed_up = f.iloc[-2] < s.iloc[-2] and f.iloc[-1] > s.iloc[-1]
    return "BUY" if crossed_up else "WAIT"

# confluence
df_h1  = fetch_candles("ES", timeframe_H1)
df_m15 = fetch_candles("ES", timeframe_M15)

if trend_direction(df_h1) == "UP" and entry_trigger(df_m15) == "BUY":
    print("Confluence confirmed — buy")

05Cautions in automation

• Repaint: in Pine, beware of request.security using the still-forming candle of the higher timeframe — it can "repaint" and mislead the backtest. Use the value of the already-closed candle ([1]) for reliable data.
• Fewer trades is the goal: multi-timeframe will sharply reduce the number of trades. That's a feature, not a bug — you're filtering out the bad ones. Don't "loosen" the filter just to trade more.
• Synchronization: make sure both timeframes are of the same asset and updated at the same moment. Lagged data between them generates false signals.

06Why it works

Multi-timeframe works because it attacks the biggest enemy of most strategies: trading against the dominant trend. Statistically, trades in the direction of the higher trend have better expectancy. By filtering out everything that goes against the "tide," you cut a large slice of the losing trades — while keeping the same entry strategy. It's one of the few adjustments that improves results without adding fragile complexity.

07Frequently asked questions