VCP probe 03 - rsi smooth divergence

// This Pine Script® code is licensed under Apache-2.0
// SPDX-License-Identifier: Apache-2.0
// © PineForge contributors 2026
//
// VCP probe 03 — smoothed-RSI divergence detection
//
// Purpose: isolate the momentum half of the VCP integration probe family.
// That family smooths RSI through an EMA(3) and then derives both
// ``momBullish`` / ``momBearish`` regime flags AND a divergence pair
// ``rsiDivBull`` / ``rsiDivBear`` keyed off ``close[5]`` and
// ``rsiSmooth[5]``. That 5-bar history-reference of an EMA-of-RSI is the
// most-likely place for a numerical drift to creep in — even a single
// epsilon error in the EMA(3) seed shifts every divergence event by a bar.
//
// This probe routes a trivial entry/exit through ONLY the smoothed-RSI
// divergence pair, with the EXACT VCP expressions and the VCP defaults
// (``rsiLength = 14``, EMA smoothing length 3, divergence lookback 5,
// ``rsiOverbought = 70`` / ``rsiOversold = 30`` not used here — only
// the 60/40 divergence thresholds). Any TV/engine drift here is
// attributable to ``ta.ema(ta.rsi(close, 14), 3)`` precision and the
// 5-bar history references the divergence formula uses, NOT to other
// VCP filters.
//
// Trade shape: long-only, pyramiding=1. Enter long on bullish RSI
// divergence (smoothed RSI < 40 with rising RSI vs falling 5-bar
// price reference, EXACT VCP wording flipped to a ``close > close[5]``
// constructive divergence). Close the long on bearish RSI divergence
// (smoothed RSI > 60 with falling RSI vs rising 5-bar price reference,
// VCP's exact ``rsiDivBear``).
//
// TV setup: 15m chart, ETH-USDT-USDT, same window as
// data/ohlcv_ETH-USDT-USDT_15m.csv. Export the trade list to trades.csv
// in this folder.
//@version=6
strategy("VCP probe 03 - rsi smooth divergence", shorttitle="VCP_p03",
     overlay=true, initial_capital=1000000,
     commission_type=strategy.commission.percent, commission_value=0,
     slippage=0, default_qty_type=strategy.fixed, default_qty_value=1,
     pyramiding=1, process_orders_on_close=false)

// ---- inputs (mirror VCP defaults) ----------------------------------
int i_rsi = input.int(14, "RSI Length", minval=5, maxval=30)

// ---- smoothed RSI — IDENTICAL formula to VCP -----------------------
float rsiValue  = ta.rsi(close, i_rsi)
float rsiSmooth = ta.ema(rsiValue, 3)

// ---- divergence flags — IDENTICAL expressions to VCP --------------
bool rsiDivBull = rsiSmooth < 40 and close > close[5] and rsiSmooth > rsiSmooth[5]
bool rsiDivBear = rsiSmooth > 60 and close < close[5] and rsiSmooth < rsiSmooth[5]

// ---- trivial entry / exit ------------------------------------------
if rsiDivBull and strategy.position_size == 0
    strategy.entry("L", strategy.long, qty=1, comment="rsi div bull")

if rsiDivBear and strategy.position_size > 0
    strategy.close("L", comment="rsi div bear exit")