The recovery
inside the channel.

The move starts with a return.

Price can stretch away from its recent average and keep going. This strategy waits for a different event: a completed candle returning inside a volatility channel. For a long entry, its close crosses back above the lower band. For a short, it crosses back below the upper band.

The centre is an exponential moving average, or EMA. The outer bands use Average True Range (ATR), a measure of the size of recent price movements. Wider movement makes a wider channel. The long recipe uses 11 candles and 1.7 × ATR; the short recipe uses 19 candles and 1.5 × ATR. This is the recovery version of Keltner, not a rule that buys every upper-band breakout.

For the general channel concept: TradingView’s Keltner Channels guide. The exact entry and exit rules here come from the saved Strategy Lab configuration.

    bool kc = kind == "keltner" or kind == "keltner-revert"
    float mid = sl_average(src, length, kc ? "ema" : "sma")
    float atr = sl_average(tr, length, "rma")
    float spread = kc ? atr : ta.stdev(src, length, true)
    float hi = mid + factor * spread
    float lo = mid - factor * spread
    bool revert = kind == "bb-revert" or kind == "keltner-revert"
    goLong := sl_up(src, revert ? lo : hi)
    goShort := sl_down(src, revert ? hi : lo)

One market. Four sets of rules.

Before looking for an entry, the strategy asks which conditions apply. Its CHOP(50) threshold is 55: below it, use the trend recipes; at or above it, use the range recipes. CHOP measures how directional or choppy the recent path has been. It does not choose up or down. About CHOP

The labels are routing rules, not a judgement made by a human looking at the chart. A trade keeps the recipe that opened it, even if CHOP later crosses the threshold. Only one position can be open; conflicting long and short entry signals are skipped.

float regime = sl_chop(trueRange, regimeLength, n)
bool ranging = regime >= regimeThreshold
bool longSignal = active and not na(regime) and (ranging ? r3L : r1L)
bool shortSignal = active and not na(regime) and (ranging ? r4S : r2S)

A recovery. A confirmation. A fill.

At the close of the 11:15 UTC candle, price crossed back above the lower Keltner band. CHOP(50) was 47.82, selecting the trend recipes. The long recipe’s extra CHOP(33) check was 42.55, below its threshold of 48. Both conditions agreed.

1. 11:30 · Enter long

   The simulation buys at the next candle’s open, adjusted for slippage: 2,455.41 USDT. Its 4% fixed stop starts at 2,357.19 USDT.

2. 14:00 candle · Target reached

   The 8% price target is touched during this candle. After exit slippage, the fill is 2,651.31 USDT. It did not need the Bollinger exit or the 21-candle holding limit.

3. The account effect

   Net profit is $91.82, taking the compounded account from $1,169.82 to $1,261.64. The trade returned 7.85% on its allocation after costs; the advertised +27.09% belongs to all 17 trades combined.

    else if active and (not useSavedEnd or time_close < endTime) and bar_index + 1 - lastExit > cooldown and longSignal != shortSignal and strategy.equity > 0
        route := longSignal ? longRoute : shortRoute
        strategy.entry(longSignal ? "L" : "S", longSignal ? strategy.long : strategy.short, comment = sl_routeName(route), alert_message = longSignal ? "entry_long" : "entry_short", disable_alert = not enableTradeAlerts)

What this clean record can tell us.

All 17 trades closed with a net profit in this replay. None was forcibly closed at the end, and none touched both a stop and target in the same candle. The largest equity decline observed by this simulator was 0.86%. That does not mean each stop was 0.86%, or that a future drawdown is capped there.

The account started with $1,000 and used 100% of available equity at 1× exposure, one position at a time. Profits compounded. Fees were 0.06% per side and adverse slippage 0.02% per side. The replay deducted $22.22 in fees; its estimated slippage cost was $7.41, already reflected in the fills.

There is a limit to what five days of selected history can establish. Search favoured this result on the same data, and this snapshot does not retain a verified total number of candidates tried. There is no unseen-data or forward-test result attached. Funding, spread variation, market impact and real fill availability are outside this simulation. Candle-based drawdown can miss the path inside a candle.

The useful next question is how the unchanged rules behave on later, unseen candles with realistic costs. This page keeps the original sample fixed, so it remains a reference when that work is done.