diff --git a/BOT/scoreboard.py b/BOT/scoreboard.py
index def2902..363c185 100644
--- a/BOT/scoreboard.py
+++ b/BOT/scoreboard.py
@@ -225,14 +225,20 @@ def _create_scoreboard_sync(model: dict[str, Any], output_path: str, bar_color:
         name_x = start_x if not flipped else start_x + section_width - name_w
         draw.text((name_x, header_y), team_name, font=fonts["team"], fill=header_fill)
 
-        # W/L slot — only drawn once real per-tournament standings exist (stub returns
-        # None today, so nothing renders under the team name).
-        standings = get_tss_standings(team.get("team_id"))
+        # W/L slot — per-tournament record under the team name. Coloured exactly like
+        # SREBOT: green W · grey dash · red L · grey dash · win-rate% on a red→green
+        # gradient. Placed below the team name's full glyph box (descender-inclusive)
+        # so names with p/g/y descenders don't clip into the record line.
+        standings = get_tss_standings(team.get("team_id"), model.get("tournament_id"))
         if standings:
-            wl_text = f"{standings['wins']}W - {standings['losses']}L"
-            wb = draw.textbbox((0, 0), wl_text, font=fonts["sub"])
-            wl_x = start_x if not flipped else start_x + section_width - (wb[2] - wb[0])
-            draw.text((wl_x, header_y + (nb[3] - nb[1]) + 18), wl_text, font=fonts["sub"], fill=GREY_FILL)
+            wl_segs = _winloss_segments(standings)
+            seg_w = _segments_width(draw, wl_segs, fonts["sub"])
+            wl_x = start_x if not flipped else start_x + section_width - seg_w
+            # Sit halfway between the original (descender-clipping) baseline and a
+            # full descender-clear drop — enough to clear p/g/y, no more.
+            name_full_h = fonts["team"].getbbox("ApgjyQ")[3]
+            wl_y = header_y + ((nb[3] - nb[1] + 18) + (name_full_h + 10)) // 2
+            _draw_segments(draw, wl_x, wl_y, wl_segs, fonts["sub"])
 
         # Stat columns (rating now lives inline next to the player name).
         col_labels = ["Air", "Ground", "Assists", "Deaths", "Caps"]
@@ -270,7 +276,7 @@ def _create_scoreboard_sync(model: dict[str, Any], output_path: str, bar_color:
                     pass
 
         # Player rows. Layout per row:
-        #   top line  — player name + inline (rating ±delta) at the column edge; stats
+        #   top line  — player name + inline (rating) at the column edge; stats
         #   lower line — vehicle icons next to their translated names
         body = fonts["body"]
         small = fonts["veh"]
@@ -287,7 +293,7 @@ def _create_scoreboard_sync(model: dict[str, Any], output_path: str, bar_color:
             veh_y = line_y + name_h + 12          # vehicle line (icons + names together)
             vname_y = veh_y + (veh_icon - small_h) // 2
             name_w = draw.textbbox((0, 0), name, font=body)[2]
-            rating_segs = _rating_segments(p.get("pvp_ratio"), p.get("pvp_ratio_delta"))
+            rating_segs = _rating_segments(p.get("pvp_ratio"))
             rating_w = _segments_width(draw, rating_segs, body)
 
             if not flipped:
@@ -342,16 +348,45 @@ def _create_scoreboard_sync(model: dict[str, Any], output_path: str, bar_color:
 RATING_FILL = (170, 200, 255, 255)
 
 
-def _rating_segments(rating, delta):
-    """Build colored (text, fill) segments for the inline ``(1468 +2)`` rating.
+# W/L record colours — matched exactly to SREBOT's scoreboard.
+WL_WIN_FILL = (0, 255, 0, 255)
+WL_LOSS_FILL = (255, 60, 60, 255)
+WL_DASH_FILL = (200, 200, 200, 255)
 
-    ``delta`` is a signed change vs the player's previous rating; it's ``None`` until
-    rating history is tracked (see wl.py), so for now only the current value shows.
+
+def _winrate_gradient(win_rate: float) -> tuple:
+    """Red→yellow→green gradient for a win-rate percentage (SREBOT's get_gradient_color).
+
+    0% = red (255,0,0), 50% = yellow (255,255,0), 100% = green (0,255,0).
     """
+    win_rate = max(0.0, min(100.0, win_rate))
+    if win_rate <= 50:
+        return (255, int(255 * (win_rate / 50)), 0, 255)
+    return (int(255 * (1 - (win_rate - 50) / 50)), 255, 0, 255)
+
+
+def _winloss_segments(standings: dict[str, int]) -> list[tuple[str, tuple]]:
+    """Colored ``3W - 2L - 60%`` segments for the per-tournament record line.
+
+    Win count + 'W' green, ' - ' grey, loss count + 'L' red, ' - ' grey, then the
+    win-rate% on the red→green gradient — mirroring SREBOT's _draw_winloss.
+    """
+    wins = int(standings.get("wins") or 0)
+    losses = int(standings.get("losses") or 0)
+    total = wins + losses
+    pct = (wins / total) * 100 if total else 0.0
+    return [
+        (f"{wins}W", WL_WIN_FILL),
+        (" - ", WL_DASH_FILL),
+        (f"{losses}L", WL_LOSS_FILL),
+        (" - ", WL_DASH_FILL),
+        (f"{pct:.0f}%", _winrate_gradient(pct)),
+    ]
+
+
+def _rating_segments(rating):
+    """Build colored (text, fill) segments for the inline ``(1468)`` rating."""
     rating_str = f"{int(round(rating))}" if isinstance(rating, (int, float)) else "—"
-    if isinstance(delta, (int, float)) and delta != 0:
-        dcol = WIN_FILL if delta > 0 else LOSS_FILL
-        return [(f"({rating_str} ", RATING_FILL), (f"{'+' if delta > 0 else ''}{int(delta)}", dcol), (")", RATING_FILL)]
     return [(f"({rating_str})", RATING_FILL)]
 
 
diff --git a/BOT/transform.py b/BOT/transform.py
index e3cf8c4..7003fc1 100644
--- a/BOT/transform.py
+++ b/BOT/transform.py
@@ -168,6 +168,7 @@ def build_scoreboard_model(game: dict[str, Any], lang_column: str = "<English>")
         "duration": game.get("duration"),
         "map": str(game.get("mission_name") or ""),
         "mission_mode": str(game.get("mission_mode") or ""),
+        "tournament_id": tss.get("tournament_id"),
         "tournament_name": str(tss.get("tournament_name") or ""),
         "bracket": str(tss.get("bracket") or ""),
         "is_draw": is_draw,
diff --git a/BOT/wl.py b/BOT/wl.py
index be3700a..ff9e388 100644
--- a/BOT/wl.py
+++ b/BOT/wl.py
@@ -1,20 +1,99 @@
-"""TSS win/loss standings — STUB.
+"""TSS win/loss standings, scoped per tournament.
 
-SREBOT tracks squadron W/L globally. TSS needs the same idea but **scoped per
-tournament** (a team's record only means something inside its bracket), so the real
-implementation is deferred. Until then ``get_tss_standings`` returns ``None`` for
-every team and the scoreboard renders a neutral placeholder in the W/L slot.
+SREBOT tracks squadron W/L globally. TSS records only mean something inside a
+bracket, so standings are scoped to ``(tournament_id, team_id)`` and derived live
+from the rows already persisted per match — no separate counter table:
 
-TODO(tss-wl): persist per-(tournament_id, team_id) wins/losses and resolve here.
+* ``match_summary``      — one row per session, carries ``tournament_id``
+* ``player_games_hist``  — one row per used vehicle per player, carries
+  ``team_id``, ``session_id`` and ``victor_bool`` ('Win'/'Loss')
+
+A draw is already written as ``victor_bool = 'Loss'`` for *both* teams (the winning
+slot is empty), so counting over ``victor_bool`` treats a draw as a loss for both
+sides with no special handling.
+
+``insert_match``/``insert_player_games`` run before the scoreboard renders (see
+``tss_ws._handle_game``), so the record returned here includes the match on the
+board being drawn.
 """
 from __future__ import annotations
 
+import logging
+import os
+import sqlite3
 from typing import Any, Optional
 
+log = logging.getLogger("tssbot.wl")
 
-def get_tss_standings(team_id: Optional[str], tournament_id: Optional[Any] = None) -> Optional[dict[str, int]]:
-    """Return ``{"wins": int, "losses": int}`` for a team, or ``None`` if untracked.
+# One outcome per session (vehicle rows are duplicated across a player's lineup, so
+# collapse to a single victor_bool per session before counting — mirrors the
+# DISTINCT-first rule in storage.py's docstring), then tally wins vs losses.
+_STANDINGS_SQL = """
+SELECT
+    SUM(CASE WHEN victor = 'Win'  THEN 1 ELSE 0 END) AS wins,
+    SUM(CASE WHEN victor = 'Loss' THEN 1 ELSE 0 END) AS losses
+FROM (
+    SELECT pgh.session_id, MAX(pgh.victor_bool) AS victor
+    FROM player_games_hist pgh
+    JOIN match_summary ms ON ms.session_id = pgh.session_id
+    WHERE pgh.team_id = ? AND ms.tournament_id = ?
+    GROUP BY pgh.session_id
+)
+"""
 
-    Stub: always ``None`` until per-tournament W/L tracking lands.
+
+def get_tss_standings(
+    team_id: Optional[Any], tournament_id: Optional[Any] = None
+) -> Optional[dict[str, int]]:
+    """Return ``{"wins": int, "losses": int}`` for a team within one tournament.
+
+    Returns ``None`` when either id is missing, the DB is unavailable, the query
+    errors, or the team has no recorded matches in that tournament — in every such
+    case the scoreboard leaves the W/L slot blank (matching offline/sample renders).
     """
-    return None
+    if team_id is None or tournament_id is None:
+        return None
+    try:
+        tid = int(team_id)
+        tournament = int(tournament_id)
+    except (TypeError, ValueError):
+        return None
+
+    # Imported here so the module stays importable without STORAGE_VOL_PATH set
+    # (e.g. for unit tests that point _db_path at a temp database).
+    db_path = _db_path()
+    if db_path is None or not os.path.exists(db_path):
+        # No file → nothing to read (and don't let sqlite3.connect create a stray
+        # empty DB). Keeps offline/sample renders showing a blank slot.
+        return None
+
+    # NOT a ``mode=ro`` connection: the production DB is WAL (see storage._PRAGMAS),
+    # and a pure read-only open can fail when a -wal exists but the -shm wal-index
+    # does not (it can't be created read-only). A normal connection set to
+    # ``query_only`` reads WAL safely in every state while never writing. ``timeout``
+    # waits out a momentary write lock from the concurrent ingest path.
+    try:
+        conn = sqlite3.connect(db_path, timeout=5.0)
+        conn.execute("PRAGMA query_only=ON")
+    except sqlite3.Error:
+        return None
+    try:
+        row = conn.execute(_STANDINGS_SQL, (tid, tournament)).fetchone()
+    except sqlite3.Error as exc:
+        log.warning("[TSS-WL] standings query failed (team=%s tourn=%s): %s", tid, tournament, exc)
+        return None
+    finally:
+        conn.close()
+
+    if not row or row[0] is None:
+        return None
+    return {"wins": int(row[0] or 0), "losses": int(row[1] or 0)}
+
+
+def _db_path() -> Optional[str]:
+    """Resolve the battles DB path, or ``None`` if storage isn't configured."""
+    try:
+        from .storage import TSS_BATTLES_DB_PATH
+    except Exception:
+        return None
+    return str(TSS_BATTLES_DB_PATH)
diff --git a/tests/test_transform_model.py b/tests/test_transform_model.py
new file mode 100644
index 0000000..b299bd8
--- /dev/null
+++ b/tests/test_transform_model.py
@@ -0,0 +1,31 @@
+import pathlib
+import sys
+
+sys.path.insert(0, str(pathlib.Path(__file__).resolve().parents[1]))
+sys.path.insert(0, str(pathlib.Path(__file__).resolve().parents[2] / "SHARED"))
+
+from BOT.transform import build_scoreboard_model
+
+
+def _game():
+    return {
+        "_id": "abc",
+        "winner": "1",
+        "mission_name": "Kursk",
+        "tss": {
+            "tournament_id": 24965,
+            "tournament_name": "2x2 RBm Tanks",
+            "1": {"team_id": "111", "team_name": "A", "players": [{"uid": "1", "pvp_ratio": 1500}]},
+            "2": {"team_id": "222", "team_name": "B", "players": [{"uid": "2", "pvp_ratio": 1400}]},
+        },
+        "players": {
+            "1": {"name": "alice", "team": "1", "units": [{"unit": "t34", "used": True}]},
+            "2": {"name": "bob", "team": "2", "units": [{"unit": "pz", "used": True}]},
+        },
+    }
+
+
+def test_model_carries_tournament_id():
+    model = build_scoreboard_model(_game())
+    assert model is not None
+    assert model["tournament_id"] == 24965
diff --git a/tests/test_wl.py b/tests/test_wl.py
new file mode 100644
index 0000000..c795970
--- /dev/null
+++ b/tests/test_wl.py
@@ -0,0 +1,155 @@
+import pathlib
+import sqlite3
+import sys
+
+sys.path.insert(0, str(pathlib.Path(__file__).resolve().parents[1]))
+sys.path.insert(0, str(pathlib.Path(__file__).resolve().parents[2] / "SHARED"))
+
+from BOT import wl
+
+
+def _make_db(path: str, matches, players):
+    """Seed the two real tables with the columns get_tss_standings reads.
+
+    matches: list of (session_id, tournament_id)
+    players: list of (session_id, team_id, vehicle_internal, victor_bool)
+    """
+    conn = sqlite3.connect(path)
+    conn.execute(
+        "CREATE TABLE match_summary (session_id TEXT PRIMARY KEY, tournament_id INTEGER)"
+    )
+    conn.execute(
+        "CREATE TABLE player_games_hist ("
+        " UID TEXT, session_id TEXT, team_id INTEGER,"
+        " vehicle_internal TEXT, victor_bool TEXT)"
+    )
+    conn.executemany("INSERT INTO match_summary VALUES (?, ?)", matches)
+    conn.executemany(
+        "INSERT INTO player_games_hist (UID, session_id, team_id, vehicle_internal, victor_bool)"
+        " VALUES ('u', ?, ?, ?, ?)",
+        players,
+    )
+    conn.commit()
+    conn.close()
+
+
+def _point_at(monkeypatch, path):
+    monkeypatch.setattr(wl, "_db_path", lambda: str(path))
+
+
+def test_basic_wins_and_losses(tmp_path, monkeypatch):
+    db = tmp_path / "battles.db"
+    _make_db(
+        str(db),
+        matches=[("s1", 100), ("s2", 100), ("s3", 100)],
+        players=[
+            ("s1", 7, "t34", "Win"),
+            ("s2", 7, "t34", "Win"),
+            ("s3", 7, "t34", "Loss"),
+        ],
+    )
+    _point_at(monkeypatch, db)
+    assert wl.get_tss_standings(7, 100) == {"wins": 2, "losses": 1}
+
+
+def test_per_tournament_isolation(tmp_path, monkeypatch):
+    db = tmp_path / "battles.db"
+    # Same team_id appears in two tournaments; records must not bleed across.
+    _make_db(
+        str(db),
+        matches=[("a", 100), ("b", 100), ("c", 200)],
+        players=[
+            ("a", 7, "t34", "Win"),
+            ("b", 7, "t34", "Loss"),
+            ("c", 7, "t34", "Win"),
+        ],
+    )
+    _point_at(monkeypatch, db)
+    assert wl.get_tss_standings(7, 100) == {"wins": 1, "losses": 1}
+    assert wl.get_tss_standings(7, 200) == {"wins": 1, "losses": 0}
+
+
+def test_draw_is_loss_for_both_teams(tmp_path, monkeypatch):
+    db = tmp_path / "battles.db"
+    # A draw is persisted as victor_bool='Loss' for both participating teams.
+    _make_db(
+        str(db),
+        matches=[("s1", 100)],
+        players=[
+            ("s1", 7, "t34", "Loss"),
+            ("s1", 9, "pz", "Loss"),
+        ],
+    )
+    _point_at(monkeypatch, db)
+    assert wl.get_tss_standings(7, 100) == {"wins": 0, "losses": 1}
+    assert wl.get_tss_standings(9, 100) == {"wins": 0, "losses": 1}
+
+
+def test_vehicle_duplicate_rows_count_once(tmp_path, monkeypatch):
+    db = tmp_path / "battles.db"
+    # One player on the team flew two vehicles -> two rows, same session/outcome.
+    _make_db(
+        str(db),
+        matches=[("s1", 100)],
+        players=[
+            ("s1", 7, "t34", "Win"),
+            ("s1", 7, "is2", "Win"),
+        ],
+    )
+    _point_at(monkeypatch, db)
+    assert wl.get_tss_standings(7, 100) == {"wins": 1, "losses": 0}
+
+
+def test_string_team_id_is_coerced(tmp_path, monkeypatch):
+    db = tmp_path / "battles.db"
+    _make_db(str(db), matches=[("s1", 100)], players=[("s1", 7, "t34", "Win")])
+    _point_at(monkeypatch, db)
+    # The render model carries team_id as a string.
+    assert wl.get_tss_standings("7", 100) == {"wins": 1, "losses": 0}
+
+
+def test_missing_ids_return_none(tmp_path, monkeypatch):
+    db = tmp_path / "battles.db"
+    _make_db(str(db), matches=[("s1", 100)], players=[("s1", 7, "t34", "Win")])
+    _point_at(monkeypatch, db)
+    assert wl.get_tss_standings(None, 100) is None
+    assert wl.get_tss_standings(7, None) is None
+    assert wl.get_tss_standings("not-a-number", 100) is None
+
+
+def test_unknown_team_returns_none(tmp_path, monkeypatch):
+    db = tmp_path / "battles.db"
+    _make_db(str(db), matches=[("s1", 100)], players=[("s1", 7, "t34", "Win")])
+    _point_at(monkeypatch, db)
+    assert wl.get_tss_standings(999, 100) is None
+
+
+def test_missing_db_returns_none(tmp_path, monkeypatch):
+    _point_at(monkeypatch, tmp_path / "does_not_exist.db")
+    assert wl.get_tss_standings(7, 100) is None
+
+
+def test_reads_wal_mode_db_with_open_writer(tmp_path, monkeypatch):
+    # The production DB is WAL. Reproduce a live-ingest state: WAL journal, no
+    # autocheckpoint (so a non-empty -wal/-shm lingers), and a writer connection
+    # still open while we read — get_tss_standings must still return the record.
+    db = tmp_path / "battles.db"
+    _make_db(
+        str(db),
+        matches=[("s1", 100), ("s2", 100)],
+        players=[("s1", 7, "t34", "Win"), ("s2", 7, "t34", "Loss")],
+    )
+    writer = sqlite3.connect(str(db))
+    writer.execute("PRAGMA journal_mode=WAL")
+    writer.execute("PRAGMA wal_autocheckpoint=0")
+    writer.execute(
+        "INSERT INTO player_games_hist (UID, session_id, team_id, vehicle_internal, victor_bool)"
+        " VALUES ('u', 's3', 7, 't34', 'Win')"
+    )
+    writer.execute("INSERT INTO match_summary VALUES ('s3', 100)")
+    writer.commit()  # committed to the WAL, not checkpointed; writer stays open
+    try:
+        _point_at(monkeypatch, db)
+        assert wl.get_tss_standings(7, 100) == {"wins": 2, "losses": 1}
+    finally:
+        writer.close()