DL-Broadcast-Tool/match_engine.py

# match_engine.py (Modernized & GUI-Compatible)

import os
import json
import math

BASE_DIR = os.path.dirname(os.path.abspath(__file__))
CAREER_DB_PATH = os.path.join(BASE_DIR, "final_db.json")
CAREER_STATS_PATH = os.path.join(BASE_DIR, "career_stats.json")


def load_career_stats_db():
    if not os.path.exists(CAREER_STATS_PATH):
        return None
    with open(CAREER_STATS_PATH, "r", encoding="utf-8") as f:
        return json.load(f)


# ---------------------------------------------------------
# Load DB
# ---------------------------------------------------------

def load_career_db():
    if not os.path.exists(CAREER_DB_PATH):
        return None
    with open(CAREER_DB_PATH, "r", encoding="utf-8") as f:
        return json.load(f)


def _get_player_entry(db, pid):
    pid = str(pid)
    return db["players"].get(pid)
    

def attach_career_tags_to_match_player(match_player, career_entry):
    """
    Injects career tag blocks into a match_player object so that
    storylines, predictions, and UI panels can access:
        - consistency
        - anchor
        - clutch
        - breaker
    """
    for tag in ("consistency", "anchor", "clutch", "breaker"):
        if tag in career_entry:
            match_player[tag] = career_entry[tag]


# ---------------------------------------------------------
# Universal Match Ranking (Career Tags)
# ---------------------------------------------------------

def rank_match_players_by_tag(match_players, tag_name):
    """
    Rank match players using career tag percentiles.
    """

    db = load_career_db()
    if db is None:
        return []

    ranked = []

    for p in match_players:
        pid = p.get("id") or p.get("PlayerUUID") or p.get("uuid")
        if not pid:
            continue

        entry = _get_player_entry(db, pid)
        if not entry:
            continue
            
        attach_career_tags_to_match_player(p, entry)    

        tag = entry.get(tag_name, {})
        pct = tag.get("pct", 0)
        raw = tag.get("raw", 0.0)

        ranked.append({
            "id": pid,
            "name": entry.get("name", "Unknown"),
            "team": p.get("team", "Unknown"),
            "pct": pct,
            "raw": raw,
            "tier": tag.get("tier", "D"),
            "summary": tag.get("summary", ""),
        })

    ranked.sort(key=lambda x: x["pct"], reverse=True)

    for i, r in enumerate(ranked, 1):
        r["rank"] = i

    return ranked


# ---------------------------------------------------------
# Match-Based Consistency (using match stats only)
# ---------------------------------------------------------

def rank_match_players_consistency(match_players):
    ranked = []

    for p in match_players:
        tag = p.get("consistency", {})
        raw = tag.get("raw", 0.0)

        ranked.append({
            "name": p.get("name", "Unknown"),
            "team": p.get("team", ""),
            "score_raw": raw,
            "score_display": f"{raw:.2f}",
        })

    ranked.sort(key=lambda x: x["score_raw"], reverse=True)

    for i, r in enumerate(ranked, 1):
        r["rank"] = i

    return ranked


# ---------------------------------------------------------
# Match-Based Clutch (using match stats only)
# ---------------------------------------------------------

def rank_match_players_clutch(match_players):
    """
    Uses the same attached-tag pattern as Slayer/Sharpshooter/Payload.
    The previous version of this function only ever returned name/team/
    score_raw/score_display — it never included "pct" at all, so no
    amount of fixing the underlying data could have made a percentile
    show up here; the field simply wasn't being read/returned.
    """
    return _add_score_fields(_rank_by_attached_tag(match_players, "clutch"))


# ---------------------------------------------------------
# Utility: Split into two columns by team
# ---------------------------------------------------------

def split_two_columns(ranked_players, team_a_name, team_b_name):
    left = []
    right = []
    for p in ranked_players:
        if p["team"] == team_a_name:
            left.append(p)
        elif p["team"] == team_b_name:
            right.append(p)
    return left, right


def _rank_by_attached_tag(match_players, tag_name):
    """
    Ranks using whatever's already attached to each player object
    (p[tag_name]) — the same career_stats.json-sourced data
    on_generate_slots already attached — instead of
    rank_match_players_by_tag()'s independent final_db.json lookup by
    raw UUID. That lookup was a completely separate, stale data source
    from career_stats.json, keyed differently, and any player missing
    from final_db.json was silently dropped from the ranking entirely.

    "has_data" marks whether this player actually HAS a real tag
    entry, vs an empty/missing one defaulting to 0. Without this
    distinction, a player with no career data (e.g. not yet present in
    the local match-history cache used to build career_stats.json)
    looks identical to a player who genuinely scored the worst possible
    result — which silently drags their team's average down in every
    prediction, exactly as if they were a liability rather than simply
    unmeasured.
    """
    ranked = []
    for p in match_players:
        tag = p.get(tag_name) or {}
        ranked.append({
            "id": p.get("id"),
            "name": p.get("name", "Unknown"),
            "team": p.get("team", "Unknown"),
            "pct": tag.get("pct", 0),
            "raw": tag.get("raw", 0.0),
            "tier": tag.get("tier", "D"),
            "summary": tag.get("summary", ""),
            "has_data": tag.get("has_data", bool(tag)),
        })
    ranked.sort(key=lambda x: x["pct"], reverse=True)
    for i, r in enumerate(ranked, 1):
        r["rank"] = i
    return ranked


# ---------------------------------------------------------
# Legacy-Compatible Wrappers (GUI expects these)
# ---------------------------------------------------------

def _add_score_fields(ranked):
    """Adds BOTH score_raw and score_display required by GUI."""
    for r in ranked:
        r["score_raw"] = r["raw"]
        r["score_display"] = f"{r['raw']:.2f}"
    return ranked


def rank_match_players_slayer(match_players):
    return _add_score_fields(_rank_by_attached_tag(match_players, "slayer"))


def rank_match_players_sharpshooter(match_players):
    return _add_score_fields(_rank_by_attached_tag(match_players, "sharpshooter"))


def rank_match_players_objpl(match_players):
    return _add_score_fields(_rank_by_attached_tag(match_players, "objective_payload"))

def rank_match_players_objdom(match_players):
    ranked = []

    for p in match_players:
        # CHANGE: Look for "objective_domination" instead of "_components"
        dom_data = p.get("objective_domination", {})
        raw = dom_data.get("raw", 0.0)

        ranked.append({
            "id": p.get("id"),
            "name": p.get("name"),
            "team": p.get("team"),
            "score_raw": raw,
            "score_display": f"{raw:.2f}",
            "has_data": bool(dom_data),
        })

    ranked.sort(key=lambda x: x["score_raw"], reverse=True)

    for i, r in enumerate(ranked, 1):
        r["rank"] = i

    return ranked
    
    
def rank_match_players_consistency_career(match_players):
    """
    Ranks by each player's career-level Consistency tag, read directly
    from career_stats.json BY CANONICAL ID — not the old final_db.json,
    a separate stale file keyed by raw UUID that many current players
    (and anyone resolved/merged since it was last built) simply aren't
    in, silently scoring them 0.
    """
    career_db = load_career_stats_db()
    players_db = career_db.get("players", {}) if career_db else {}

    ranked = []
    for p in match_players:
        cid = p.get("canonical")
        pdata = players_db.get(cid, {}) if cid else {}
        tag = pdata.get("consistency", {})
        raw = tag.get("raw", 0.0)
        ranked.append({
            "id": p.get("id"),
            "name": p.get("name", "Unknown"),
            "team": p.get("team", "Unknown"),
            "score_raw": raw,
            "score_display": f"{raw:.2f}",
            "pct": tag.get("pct"),
            "tier": tag.get("tier"),
            "has_data": bool(tag),
        })

    ranked.sort(key=lambda x: x["score_raw"], reverse=True)
    for i, r in enumerate(ranked, 1):
        r["rank"] = i
    return ranked

# ---------------------------------------------------------
# Legacy Slayer Prediction Wrapper
# ---------------------------------------------------------

def compute_slayer_prediction(team_a_players, team_b_players):
    """
    Legacy wrapper for GUI compatibility.

    IMPORTANT: team_a_players/team_b_players here are the RANKED ROW
    dicts produced by rank_match_players_slayer() (id/name/team/pct/
    raw/tier/...), not the original match player objects — "raw" is a
    TOP-LEVEL key on these rows, not nested under a "slayer" key. This
    previously read p.get("slayer", {}).get("raw", 0.0), which doesn't
    exist on these rows and always silently returned 0.0 for every
    player on both teams — making the win prediction always compute to
    a flat, meaningless 50/50 regardless of the real data, while the
    Relative Advantage shown alongside it (computed correctly elsewhere
    from the same rows' real score_raw) told a completely different,
    accurate story. Reading "raw" directly fixes that mismatch.
    """

    A = [p.get("raw", 0.0) for p in team_a_players]
    B = [p.get("raw", 0.0) for p in team_b_players]

    teamA_avg = sum(A) / max(1, len(A))
    teamB_avg = sum(B) / max(1, len(B))

    if max(teamA_avg, teamB_avg) == 0:
        slayer_edge = 0
    else:
        slayer_edge = (teamB_avg - teamA_avg) / max(teamA_avg, teamB_avg)

    k = 1.1
    pB = 1 / (1 + math.exp(-k * slayer_edge))
    pA = 1 - pB

    return {
        "teamA_avg": teamA_avg,
        "teamB_avg": teamB_avg,
        "teamA_win": round(pA * 100),
        "teamB_win": round(pB * 100),
    }