src/advanced_indicators.py

"""
Advanced Indicators — Absorption Bubbles + Pro Scalper logic.

Implements TradingView-style indicators using OHLCV data:
1. Absorption Bubbles: Detects volume absorption on wicks (normalized volume vs rolling std dev)
2. Pro Scalper: Kalman-filtered Supertrend + VWMA bands for overbought/oversold zones

These work with yfinance OHLCV data. For TradingView screener data (no candle history),
a simplified version is computed from available fields.
"""

import numpy as np
import pandas as pd
from typing import Optional


# ──────────────────────────────────────────────
# Absorption Bubbles
# ──────────────────────────────────────────────

def compute_absorption_bubbles(
    df: pd.DataFrame,
    lookback: int = 20,
    threshold: float = 2.0,
) -> dict:
    """Detect absorption events on candle wicks.

    Logic (based on TradingView 'Absorption Bubbles' by profitprotrading):
    - Normalize volume against a rolling standard deviation over `lookback` bars
    - Identify candles where wick volume (estimated) exceeds `threshold` std devs
    - Green bubbles on upper wicks = selling absorption (sellers absorb buying pressure)
    - Red bubbles on lower wicks = buying absorption (buyers absorb selling pressure)

    Returns:
        dict with absorption analysis results
    """
    if df is None or len(df) < lookback + 5:
        return {"absorption_detected": False, "events": []}

    close = df["Close"].astype(float).values
    open_ = df["Open"].astype(float).values
    high = df["High"].astype(float).values
    low = df["Low"].astype(float).values
    volume = df["Volume"].astype(float).values

    n = len(df)

    # Calculate wick ratios (proxy for wick volume distribution)
    body = np.abs(close - open_)
    full_range = high - low
    full_range = np.where(full_range == 0, 1e-10, full_range)  # avoid div by zero

    upper_wick = high - np.maximum(close, open_)
    lower_wick = np.minimum(close, open_) - low

    # Wick volume estimation: distribute volume proportionally to wick size
    upper_wick_vol = volume * (upper_wick / full_range)
    lower_wick_vol = volume * (lower_wick / full_range)

    # Rolling mean and std of volume
    vol_series = pd.Series(volume)
    vol_mean = vol_series.rolling(lookback).mean().values
    vol_std = vol_series.rolling(lookback).std().values
    vol_std = np.where(vol_std == 0, 1e-10, vol_std)

    # Normalized wick volumes (z-scores)
    upper_z = (upper_wick_vol - vol_mean) / vol_std
    lower_z = (lower_wick_vol - vol_mean) / vol_std

    # Detect absorption events
    events = []
    recent_buying_absorption = []
    recent_selling_absorption = []

    # Look at last 5 bars for recent events
    for i in range(max(lookback, n - 5), n):
        if np.isnan(upper_z[i]) or np.isnan(lower_z[i]):
            continue

        # Selling absorption: large upper wick volume (sellers absorb buyers pushing up)
        if upper_z[i] > threshold:
            events.append({
                "bar_index": i,
                "type": "selling_absorption",
                "strength": round(float(upper_z[i]), 2),
                "price_level": round(float(high[i]), 4),
            })
            recent_selling_absorption.append(float(upper_z[i]))

        # Buying absorption: large lower wick volume (buyers absorb sellers pushing down)
        if lower_z[i] > threshold:
            events.append({
                "bar_index": i,
                "type": "buying_absorption",
                "strength": round(float(lower_z[i]), 2),
                "price_level": round(float(low[i]), 4),
            })
            recent_buying_absorption.append(float(lower_z[i]))

    # Also check body absorption (aggressive directional volume in body)
    body_ratio = body / full_range
    body_vol = volume * body_ratio
    body_z = (body_vol - vol_mean) / vol_std

    last_bar = n - 1
    body_absorption = None
    if not np.isnan(body_z[last_bar]) and body_z[last_bar] > threshold:
        if close[last_bar] > open_[last_bar]:
            body_absorption = "aggressive_buying"
        else:
            body_absorption = "aggressive_selling"

    # Summary for current bar
    current_upper_z = float(upper_z[last_bar]) if not np.isnan(upper_z[last_bar]) else 0
    current_lower_z = float(lower_z[last_bar]) if not np.isnan(lower_z[last_bar]) else 0

    return {
        "absorption_detected": len(events) > 0,
        "events": events,
        "current_bar": {
            "upper_wick_z": round(current_upper_z, 2),
            "lower_wick_z": round(current_lower_z, 2),
            "selling_absorption": current_upper_z > threshold,
            "buying_absorption": current_lower_z > threshold,
            "body_absorption": body_absorption,
        },
        "recent_buying_strength": round(max(recent_buying_absorption), 2) if recent_buying_absorption else 0,
        "recent_selling_strength": round(max(recent_selling_absorption), 2) if recent_selling_absorption else 0,
        "signal_bias": _absorption_bias(recent_buying_absorption, recent_selling_absorption),
    }


def _absorption_bias(buying: list, selling: list) -> str:
    """Determine directional bias from absorption events.

    Buying absorption (red bubbles on lower wicks) = support forming = bullish
    Selling absorption (green bubbles on upper wicks) = resistance forming = bearish
    """
    buy_strength = sum(buying) if buying else 0
    sell_strength = sum(selling) if selling else 0

    if buy_strength > sell_strength and buy_strength > 0:
        return "bullish"  # Buyers absorbing = support = price likely to go up
    elif sell_strength > buy_strength and sell_strength > 0:
        return "bearish"  # Sellers absorbing = resistance = price likely to go down
    else:
        return "neutral"


# ──────────────────────────────────────────────
# Pro Scalper (Kalman-filtered Supertrend + VWMA Bands)
# ──────────────────────────────────────────────

def compute_pro_scalper(
    df: pd.DataFrame,
    atr_period: int = 10,
    atr_multiplier: float = 3.0,
    kalman_gain: float = 0.7,
    vwma_period: int = 20,
    vwma_multiplier: float = 2.0,
) -> dict:
    """Compute Pro Scalper signals.

    Logic (based on TradingView 'Pro Scalper' by profitprotrading):
    - Kalman-filtered Supertrend for buy/sell signals
    - VWMA bands for overbought/oversold zones
    - Reversal signals when price hits OB/OS zones

    Returns:
        dict with scalper signals and zones
    """
    if df is None or len(df) < max(atr_period, vwma_period) + 10:
        return {"signal": "neutral", "confidence": 0}

    close = df["Close"].astype(float).values
    high = df["High"].astype(float).values
    low = df["Low"].astype(float).values
    volume = df["Volume"].astype(float).values
    n = len(df)

    # ── Kalman-filtered Supertrend ──
    # Step 1: Compute ATR
    tr = np.maximum(
        high[1:] - low[1:],
        np.maximum(
            np.abs(high[1:] - close[:-1]),
            np.abs(low[1:] - close[:-1])
        )
    )
    tr = np.insert(tr, 0, high[0] - low[0])
    atr = pd.Series(tr).rolling(atr_period).mean().values

    # Step 2: Kalman filter on price (smoothed source)
    kalman_price = np.zeros(n)
    kalman_price[0] = close[0]
    for i in range(1, n):
        kalman_price[i] = kalman_price[i - 1] + kalman_gain * (close[i] - kalman_price[i - 1])

    # Step 3: Supertrend using Kalman-filtered price
    hl2 = (high + low) / 2.0
    upper_band = np.zeros(n)
    lower_band = np.zeros(n)
    supertrend = np.zeros(n)
    direction = np.ones(n)  # 1 = bullish, -1 = bearish

    for i in range(atr_period, n):
        if np.isnan(atr[i]):
            continue

        # Use Kalman-filtered midpoint for band calculation
        mid = (kalman_price[i] + hl2[i]) / 2.0
        upper_band[i] = mid + atr_multiplier * atr[i]
        lower_band[i] = mid - atr_multiplier * atr[i]

        # Supertrend logic
        if i > atr_period:
            # Adjust bands (don't let them move against the trend)
            if lower_band[i] < lower_band[i - 1] and close[i - 1] > lower_band[i - 1]:
                lower_band[i] = lower_band[i - 1]
            if upper_band[i] > upper_band[i - 1] and close[i - 1] < upper_band[i - 1]:
                upper_band[i] = upper_band[i - 1]

            # Direction
            if direction[i - 1] == 1:
                if close[i] < lower_band[i]:
                    direction[i] = -1
                    supertrend[i] = upper_band[i]
                else:
                    direction[i] = 1
                    supertrend[i] = lower_band[i]
            else:
                if close[i] > upper_band[i]:
                    direction[i] = 1
                    supertrend[i] = lower_band[i]
                else:
                    direction[i] = -1
                    supertrend[i] = upper_band[i]

    # Detect buy/sell signal (direction change)
    scalper_signal = "neutral"
    signal_bar = -1
    for i in range(n - 1, max(n - 4, atr_period), -1):
        if direction[i] == 1 and direction[i - 1] == -1:
            scalper_signal = "buy"
            signal_bar = i
            break
        elif direction[i] == -1 and direction[i - 1] == 1:
            scalper_signal = "sell"
            signal_bar = i
            break

    # Current trend direction
    current_direction = "bullish" if direction[n - 1] == 1 else "bearish"

    # ── VWMA Bands (Overbought/Oversold Zones) ──
    vwma = _compute_vwma(close, volume, vwma_period)
    vwma_std = pd.Series(close).rolling(vwma_period).std().values

    upper_vwma = vwma + vwma_multiplier * vwma_std
    lower_vwma = vwma - vwma_multiplier * vwma_std

    # Check if current price is in OB/OS zone
    last_price = close[n - 1]
    last_upper = upper_vwma[n - 1] if not np.isnan(upper_vwma[n - 1]) else float('inf')
    last_lower = lower_vwma[n - 1] if not np.isnan(lower_vwma[n - 1]) else float('-inf')

    zone = "neutral"
    if last_price >= last_upper:
        zone = "overbought"
    elif last_price <= last_lower:
        zone = "oversold"

    # Reversal signal: price was in OB/OS and is now reversing
    reversal = None
    if n >= 3:
        prev_price = close[n - 2]
        if prev_price >= upper_vwma[n - 2] if not np.isnan(upper_vwma[n - 2]) else False:
            if last_price < prev_price:
                reversal = "bearish_reversal"
        if prev_price <= lower_vwma[n - 2] if not np.isnan(lower_vwma[n - 2]) else False:
            if last_price > prev_price:
                reversal = "bullish_reversal"

    # Confidence based on signal recency and zone alignment
    confidence = _scalper_confidence(scalper_signal, zone, reversal, signal_bar, n)

    return {
        "signal": scalper_signal,
        "direction": current_direction,
        "zone": zone,
        "reversal": reversal,
        "supertrend_level": round(float(supertrend[n - 1]), 4) if supertrend[n - 1] != 0 else None,
        "vwma_upper": round(float(last_upper), 4) if not np.isnan(last_upper) else None,
        "vwma_lower": round(float(last_lower), 4) if not np.isnan(last_lower) else None,
        "confidence": confidence,
        "bars_since_signal": n - 1 - signal_bar if signal_bar > 0 else None,
    }


def _compute_vwma(close: np.ndarray, volume: np.ndarray, period: int) -> np.ndarray:
    """Compute Volume-Weighted Moving Average."""
    vwma = np.full(len(close), np.nan)
    for i in range(period - 1, len(close)):
        window_vol = volume[i - period + 1:i + 1]
        window_close = close[i - period + 1:i + 1]
        total_vol = window_vol.sum()
        if total_vol > 0:
            vwma[i] = (window_close * window_vol).sum() / total_vol
        else:
            vwma[i] = window_close.mean()
    return vwma


def _scalper_confidence(signal: str, zone: str, reversal: Optional[str], signal_bar: int, n: int) -> float:
    """Calculate confidence score for scalper signal (0-1)."""
    conf = 0.5

    # Signal present
    if signal == "buy":
        conf += 0.2
        if zone == "oversold":
            conf += 0.15  # Buy in oversold = high confidence
        elif zone == "overbought":
            conf -= 0.2  # Buy in overbought = risky
        if reversal == "bullish_reversal":
            conf += 0.15
    elif signal == "sell":
        conf += 0.2
        if zone == "overbought":
            conf += 0.15
        elif zone == "oversold":
            conf -= 0.2
        if reversal == "bearish_reversal":
            conf += 0.15
    else:
        conf = 0.3  # No signal

    # Recency bonus (signal within last 2 bars)
    if signal_bar > 0 and (n - 1 - signal_bar) <= 2:
        conf += 0.1

    return round(max(0.0, min(1.0, conf)), 2)


# ──────────────────────────────────────────────
# Simplified versions for TradingView screener data
# (no candle history, only current indicator values)
# ──────────────────────────────────────────────

def estimate_absorption_from_indicators(indicators: dict) -> dict:
    """Estimate absorption-like signals from TradingView screener data.

    Without full OHLCV history, we approximate using:
    - Volume ratio (high volume = potential absorption)
    - RSI extremes (oversold/overbought = absorption zones)
    - ADX (low ADX + high volume = absorption/consolidation)
    - Price vs range position
    """
    vol_ratio = indicators.get("volume_ratio")
    rsi = indicators.get("rsi")
    adx = indicators.get("adx")
    change = indicators.get("change_pct", 0)
    trend = indicators.get("trend", "unknown")

    absorption_score = 0.0
    bias = "neutral"
    events = []

    # High volume with small price change = absorption
    if vol_ratio and vol_ratio > 1.5 and abs(change or 0) < 1.5:
        absorption_score += 0.3
        events.append("high_volume_low_movement")
        # Determine bias from trend
        if "bullish" in trend:
            bias = "bullish"
        elif "bearish" in trend:
            bias = "bearish"

    # RSI at extremes with volume = absorption zone
    if rsi is not None:
        if rsi < 35 and (vol_ratio or 0) > 1.0:
            absorption_score += 0.3
            bias = "bullish"
            events.append("oversold_zone")
        elif rsi > 65 and (vol_ratio or 0) > 1.0:
            absorption_score += 0.2
            bias = "bearish"
            events.append("overbought_zone")

    # Low ADX (no trend) + volume = range absorption
    if adx is not None and adx < 25 and vol_ratio and vol_ratio > 1.0:
        absorption_score += 0.2
        events.append("range_absorption")

    # Volume spike
    if vol_ratio and vol_ratio > 2.0:
        absorption_score += 0.2
        events.append("volume_spike")

    # Moderate volume with strong trend = trend absorption (continuation)
    if vol_ratio and vol_ratio > 0.8 and "strong" in trend:
        absorption_score += 0.15
        if "bullish" in trend:
            bias = "bullish"
            events.append("trend_continuation_support")
        elif "bearish" in trend:
            bias = "bearish"
            events.append("trend_continuation_resistance")

    # If no specific bias detected, infer from trend
    if bias == "neutral" and absorption_score > 0:
        if "bullish" in trend:
            bias = "bullish"
        elif "bearish" in trend:
            bias = "bearish"

    return {
        "absorption_detected": absorption_score > 0.2,
        "absorption_score": round(min(1.0, absorption_score), 2),
        "signal_bias": bias,
        "events": events,
    }


def estimate_scalper_from_indicators(indicators: dict) -> dict:
    """Estimate Pro Scalper-like signals from TradingView screener data.

    Approximates Kalman Supertrend using:
    - EMA alignment for trend direction
    - Stochastic for OB/OS zones
    - TradingView's own recommendation score
    """
    ema9 = indicators.get("ema9")
    ema21 = indicators.get("ema21")
    ema50 = indicators.get("ema50")
    price = indicators.get("current_price")
    stoch_k = indicators.get("stoch_k")
    stoch_d = indicators.get("stoch_d")
    rsi = indicators.get("rsi")
    tv_rec = indicators.get("tv_recommend")
    adx = indicators.get("adx")

    signal = "neutral"
    direction = "neutral"
    zone = "neutral"
    reversal = None
    confidence = 0.3

    # Direction from EMA alignment (proxy for Supertrend direction)
    if price and ema9 and ema21:
        if price > ema9 > ema21:
            direction = "bullish"
            signal = "buy"
            confidence = 0.6
        elif price < ema9 < ema21:
            direction = "bearish"
            signal = "sell"
            confidence = 0.6

    # OB/OS zone from Stochastic (proxy for VWMA bands)
    if stoch_k is not None:
        if stoch_k > 80:
            zone = "overbought"
            if signal == "buy":
                confidence -= 0.15  # Buying in OB is risky
        elif stoch_k < 20:
            zone = "oversold"
            if signal == "sell":
                confidence -= 0.15  # Selling in OS is risky

    # Reversal detection
    if stoch_k is not None and stoch_d is not None:
        if stoch_k < 20 and stoch_k > stoch_d:
            reversal = "bullish_reversal"
            if signal != "sell":
                confidence += 0.1
        elif stoch_k > 80 and stoch_k < stoch_d:
            reversal = "bearish_reversal"
            if signal != "buy":
                confidence += 0.1

    # TV recommendation alignment
    if tv_rec is not None:
        if tv_rec > 0.3 and signal == "buy":
            confidence += 0.1
        elif tv_rec < -0.3 and signal == "sell":
            confidence += 0.1
        elif tv_rec > 0.3 and signal == "neutral":
            signal = "buy"
            confidence = 0.5
        elif tv_rec < -0.3 and signal == "neutral":
            signal = "sell"
            confidence = 0.5

    # ADX strength bonus
    if adx is not None and adx > 25 and signal != "neutral":
        confidence += 0.1

    return {
        "signal": signal,
        "direction": direction,
        "zone": zone,
        "reversal": reversal,
        "confidence": round(max(0.0, min(1.0, confidence)), 2),
    }