"""Long-history (2010+) validation for TrendRider V9 vs V7.

Reuses the Yahoo fetch of research/lab/trend_rider_long.py (ETF_UNIVERSE +
SPY/SHY, auto-adjusted, per-ticker with retries). Runs each variant through
main.strategy_positions + main.simulate_live_book with the same live-book
parameters as the earlier long-run study ($10k, 5bp, $2/order, 10% band).

    PYTHONPATH=. uv run python -m research.tr9_long [--refresh]
"""
from __future__ import annotations

import os
import pickle
import sys

import pandas as pd

import trader  # noqa: F401  (registry side effects)
from main import simulate_live_book, strategy_positions
from research.lab.trend_rider_long import fetch
from strategies.permanent import ETF_UNIVERSE
from trader import RESIZE_BAND, STRATEGY_REGISTRY

import research.tr9_lab as lab

OUT = os.path.join("data", "research", "tr9_long.pkl")

VARIANTS = {
    # baseline registry names
    "trend_rider_v7": lambda: STRATEGY_REGISTRY["trend_rider_v7"](),
    "trend_rider_v7_vt24": lambda: STRATEGY_REGISTRY["trend_rider_v7_vt24"](),
    "trend_rider_v7_vt32": lambda: STRATEGY_REGISTRY["trend_rider_v7_vt32"](),
    "trend_rider_v7_vt36": lambda: STRATEGY_REGISTRY["trend_rider_v7_vt36"](),
    "trend_rider_v9": lambda: STRATEGY_REGISTRY["trend_rider_v9"](),
    "trend_rider_v9_vt32": lambda: STRATEGY_REGISTRY["trend_rider_v9_vt32"](),
    "trend_rider_v10": lambda: STRATEGY_REGISTRY["trend_rider_v10"](),
    "trend_rider_v10_vt32": lambda: STRATEGY_REGISTRY["trend_rider_v10_vt32"](),
    "trend_rider_v10_vt36": lambda: STRATEGY_REGISTRY["trend_rider_v10_vt36"](),
    # V9 candidates
    "v9vt24_t20": lambda: lab.TrendRiderV9(target_vol=0.24, min_lev=0.5, trail_pct=0.20),
    "v9vt28_t20": lambda: lab.TrendRiderV9(trail_pct=0.20),
    "v9vt32_t20": lambda: lab.TrendRiderV9(target_vol=0.32, min_lev=0.7, trail_pct=0.20),
    "v9vt34_t20": lambda: lab.TrendRiderV9(target_vol=0.34, min_lev=0.72, trail_pct=0.20),
    "v9vt34_t20_h10": lambda: lab.TrendRiderV9(target_vol=0.34, min_lev=0.72, trail_pct=0.20,
                                               pt_min_hold=10),
    "v9vt34_t20_thr28_h10": lambda: lab.TrendRiderV9(target_vol=0.34, min_lev=0.72, trail_pct=0.20,
                                                     pt_min_hold=10, pt_threshold=0.28),
    "v9vt34_t20_thr28_b08_h10": lambda: lab.TrendRiderV9(
        target_vol=0.34, min_lev=0.72, trail_pct=0.20, pt_min_hold=10, pt_threshold=0.28,
        pt_band=0.08),
    "v9vt32_t20_thr28_h10": lambda: lab.TrendRiderV9(target_vol=0.32, min_lev=0.7, trail_pct=0.20,
                                                     pt_min_hold=10, pt_threshold=0.28),
    "v9vt36_t20": lambda: lab.TrendRiderV9(target_vol=0.36, min_lev=0.75, trail_pct=0.20),
    "v9vt36_t22": lambda: lab.TrendRiderV9(target_vol=0.36, min_lev=0.75, trail_pct=0.22),
    "v9vt28_notrail": lambda: lab.TrendRiderV9(),
}

BLOCKS = [("2010-03-01", "2013-02-28"), ("2013-03-01", "2016-09-26"),
          ("2016-09-27", "2021-09-26"), ("2021-09-27", "2026-09-25"),
          ("2010-03-01", "2026-09-25")]


def block_stats(eq: pd.Series, a: str, b: str) -> dict:
    r = eq.pct_change().dropna().loc[a:b]
    e = (1 + r).cumprod()
    return {"cagr": e.iloc[-1] ** (252 / len(r)) - 1,
            "sharpe": r.mean() / r.std() * 252 ** 0.5,
            "mdd": float(((e / e.cummax()) - 1).min())}


def main() -> None:
    refresh = "--refresh" in sys.argv
    if refresh or not os.path.exists(OUT):
        tickers = sorted(set(ETF_UNIVERSE) | {"SPY", "SHY"})
        close, opens = fetch(tickers)
        with open(OUT, "wb") as fh:
            pickle.dump({"close": close, "opens": opens}, fh)
    d = pickle.load(open(OUT, "rb"))
    close, opens = d["close"], d["opens"]
    print("panel", close.shape, close.index.min().date(), close.index.max().date())

    rows = []
    for name, factory in VARIANTS.items():
        held = strategy_positions(factory(), close, open_data=opens)
        eq, _ = simulate_live_book(held, close, 10_000, 0.0005, 2.0, 0, 0, 50.0, RESIZE_BAND)
        full = block_stats(eq, *BLOCKS[-1])
        per = {f"{a[:4]}-{b[:4]}": block_stats(eq, a, b) for a, b in BLOCKS[:-1]}
        rows.append({"name": name, **full, **{k: v["sharpe"] for k, v in per.items()}})
    df = pd.DataFrame(rows)
    pd.set_option("display.width", 220)
    print("\nfull-period 2010-2026:")
    print(df[["name", "cagr", "sharpe", "mdd"]].to_string(
        index=False, formatters={"cagr": "{:.1%}".format, "sharpe": "{:.2f}".format,
                                 "mdd": "{:.1%}".format}))
    print("\nblock Sharpes:")
    print(df[["name", "2010-2013", "2013-2016", "2016-2021", "2021-2026"]].to_string(
        index=False, formatters={c: "{:.2f}".format for c in
                                 ["2010-2013", "2013-2016", "2016-2021", "2021-2026"]}))


if __name__ == "__main__":
    main()