"""Verifiche di correttezza + tabella completa 36 coppie x 2 direzioni.

Verifiche:
  V1 la direzione registrata coincide col segno di (A-B) all'incrocio
  V2 cret_1 coincide con ret_1
  V3 ricalcolo indipendente della media di un pattern
  V4 baseline drift ~0

Tabella: per ogni (coppia, dir) extra-rendimento 5/10/20g, quota che batte il baseline,
q-value a 10g, tasso di ritorno alla Mediana (rev10/rev20), giorni mediani al ritorno,
ed etichetta verbale "dopo 10g".

Uso:
  python full_table.py [SYMBOL] [YEAR_MIN]
Output:
  ../results/tabella_incroci_<SYM>_<YEAR>.csv (+ .md)
"""
import os, sys, numpy as np, pandas as pd
from _common import load_pair, add_regime, regime_edges, block_bootstrap_p, benjamini_hochberg, HZ, HERE

SYM = sys.argv[1] if len(sys.argv) > 1 else "EURUSD"
YEAR = int(sys.argv[2]) if len(sys.argv) > 2 else 1999
RESULTS = os.path.normpath(os.path.join(HERE, "..", "results"))

cr, ba = load_pair(SYM, YEAR)

# ---------------- VERIFICHE ----------------
sub = cr[cr.pair == "PRICExMED"]
chk = ((sub.price - sub.med) > 0).astype(int).replace(0, -1)
print(f"[V1] PRICExMED dir vs segno(price-med): {(np.sign(sub.dir)==np.sign(chk)).mean()*100:.1f}% (atteso ~100%)")
print(f"[V2] max|cret_1 - ret_1| = {(cr.cret_1-cr.ret_1).abs().max():.6f} (atteso 0)")
g0 = cr[(cr.pair=='MA121xMA7') & (cr.dir==1)]
print(f"[V3] MA121xMA7+ n={len(g0)} media cret_10 = {g0.cret_10.mean():.4f}%")
print(f"[V4] baseline cret_10 medio={ba.cret_10.mean():.4f}% sd={ba.cret_10.std():.3f}% (drift ~0)")

# ---------------- TABELLA ----------------
edges = regime_edges(ba)
cr = add_regime(cr, edges); ba = add_regime(ba, edges)
bmean = {h: ba.groupby("regime")[f"cret_{h}"].mean() for h in HZ}
bglob = {h: ba[f"cret_{h}"].mean() for h in HZ}
bsd   = {h: ba[f"cret_{h}"].std()  for h in HZ}

rows = []
for (pair, d), g in cr.groupby(["pair", "dir"]):
    rec = {"pair": pair, "dir": int(d), "n": len(g)}
    for h in HZ:
        ev = g[f"cret_{h}"].values
        bexp = g["regime"].map(bmean[h]).fillna(bglob[h]).values
        exc = ev - bexp
        rec[f"raw_{h}"] = np.nanmean(ev)
        rec[f"exc_{h}"] = np.nanmean(exc)
        rec[f"eff_{h}"] = np.nanmean(exc) / bsd[h]
        rec[f"pos_{h}"] = (exc > 0).mean() * 100
        rec[f"p_{h}"] = block_bootstrap_p(exc)
    rec["rev10"] = g["rev_10"].mean() * 100
    rec["rev20"] = g["rev_20"].mean() * 100
    rec["btr_med"] = g["bars_to_revert"].median()
    rows.append(rec)
res = pd.DataFrame(rows)
res["q10"] = benjamini_hochberg(res["p_10"].values)

def verb(r):
    if r["q10"] < 0.10 and r["exc_10"] > 0: return "SALE (extra +)"
    if r["q10"] < 0.10 and r["exc_10"] < 0: return "SCENDE (extra -)"
    return "neutro"
res["dopo_10g"] = res.apply(verb, axis=1)
res = res.sort_values(["q10", "exc_10"]).reset_index(drop=True)

os.makedirs(RESULTS, exist_ok=True)
csvp = os.path.join(RESULTS, f"tabella_incroci_{SYM}_{YEAR}.csv")
res.to_csv(csvp, index=False)

md = [f"# {SYM} (D1, {YEAR}+) — cosa succede dopo ogni incrocio (36 coppie x 2 dir)", "",
      "extra5/10/20 = extra-rendimento % vs baseline stesso regime; pos10 = % che batte il baseline;",
      "q10 = significativita' (BH, <0.10 robusto); rev10/20 = % ritorno alla Mediana; btr = giorni mediani.", "",
      "| coppia | dir | n | extra5 | extra10 | extra20 | pos10 | q10 | rev10 | rev20 | btr | dopo 10g |",
      "|---|---|---|---|---|---|---|---|---|---|---|---|"]
for _, x in res.iterrows():
    md.append(f"| {x['pair']} | {int(x['dir']):+d} | {int(x['n'])} | {x['exc_5']:+.3f} | "
              f"{x['exc_10']:+.3f} | {x['exc_20']:+.3f} | {x['pos_10']:.0f}% | {x['q10']:.3f} | "
              f"{x['rev10']:.0f}% | {x['rev20']:.0f}% | {x['btr_med']:.0f} | {x['dopo_10g']} |")
open(os.path.join(RESULTS, f"tabella_incroci_{SYM}_{YEAR}.md"), "w").write("\n".join(md))
print(f"\nsalvato: {csvp} (+ .md)  righe={len(res)}")