File size: 41,802 Bytes
1790bde
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bb6d9a6
 
 
 
 
 
 
 
 
 
 
 
1790bde
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bb6d9a6
 
 
 
 
 
 
 
 
 
 
1790bde
bb6d9a6
1790bde
 
 
bb6d9a6
 
1790bde
 
 
 
 
 
 
bc8f4e7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1790bde
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
"""FastAPI backend for the World Cup 2026 prediction platform.

Serves everything from the SQLite store populated by services/model/pipeline.py,
plus on-the-fly match predictions from the saved Dixon-Coles model."""
from __future__ import annotations
import json
import sys
import time
import httpx
from pathlib import Path

# make services/model importable
MODEL_DIR = Path(__file__).resolve().parents[1] / "model"
sys.path.insert(0, str(MODEL_DIR))

from fastapi import FastAPI, HTTPException, Query  # noqa: E402
from fastapi.middleware.cors import CORSMiddleware  # noqa: E402
from pydantic import BaseModel  # noqa: E402

import config  # noqa: E402
from db import connect  # noqa: E402
from wcmodel import WorldCupModel  # noqa: E402
import bracket as B  # noqa: E402
import networkx as nx  # noqa: E402

app = FastAPI(title="World Cup 2026 Predictor", version="1.0")
app.add_middleware(
    CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])

MODEL = WorldCupModel.load()
PARAMS = MODEL.model.get_params()
_cache: dict = {}  # generic TTL cache: key -> {"ts": float, "data": any}


def attack_defence(team: str):
    a = PARAMS.get(f"attack_{team}")
    d = PARAMS.get(f"defence_{team}")
    return (round(float(a), 3) if a is not None else None,
            round(float(d), 3) if d is not None else None)


def latest_run_id(conn) -> int:
    r = conn.execute("SELECT id FROM model_runs ORDER BY id DESC LIMIT 1").fetchone()
    if not r:
        raise HTTPException(503, "model not built yet — run pipeline.py")
    return r["id"]


def _rows(q, *a):
    with connect() as conn:
        return [dict(r) for r in conn.execute(q, a).fetchall()]


# ------------------------------------------------------------------ meta
@app.get("/api/meta")
def meta():
    with connect() as conn:
        r = conn.execute(
            "SELECT * FROM model_runs ORDER BY id DESC LIMIT 1").fetchone()
        if not r:
            raise HTTPException(503, "model not built yet")
        d = dict(r)
        d["notes"] = json.loads(d["notes"]) if d.get("notes") else {}
        return d


# ----------------------------------------------------------------- teams
@app.get("/api/teams")
def teams():
    with connect() as conn:
        rid = latest_run_id(conn)
        rows = conn.execute(
            """SELECT t.name,t.group_label,t.confederation,t.elo,t.fifa_rank,
                      s.p_group_win,s.p_advance,s.p_r16,s.p_qf,s.p_sf,s.p_final,s.p_win
               FROM teams t LEFT JOIN simulations s
                 ON s.team=t.name AND s.run_id=?
               ORDER BY s.p_win DESC""", (rid,)).fetchall()
    out = []
    for r in rows:
        d = dict(r)
        d["attack"], d["defence"] = attack_defence(d["name"])
        out.append(d)
    return out


@app.get("/api/teams/{name}")
def team_detail(name: str):
    with connect() as conn:
        rid = latest_run_id(conn)
        t = conn.execute(
            """SELECT t.*,s.p_group_win,s.p_advance,s.p_r16,s.p_qf,s.p_sf,
                      s.p_final,s.p_win
               FROM teams t LEFT JOIN simulations s
                 ON s.team=t.name AND s.run_id=?
               WHERE t.name=?""", (rid, name)).fetchone()
        if not t:
            raise HTTPException(404, f"unknown team {name}")
        d = dict(t)
        d["attack"], d["defence"] = attack_defence(name)

        # recent form (last 12 played international matches)
        form = conn.execute(
            """SELECT date,home_team,away_team,home_score,away_score,tournament,neutral
               FROM matches
               WHERE home_score IS NOT NULL AND (home_team=? OR away_team=?)
               ORDER BY date DESC LIMIT 12""", (name, name)).fetchall()
        d["recent_form"] = [_form_row(dict(r), name) for r in form]

        # World Cup history
        wc = conn.execute(
            """SELECT COUNT(*) n,
                      SUM(CASE WHEN (home_team=? AND home_score>away_score)
                            OR (away_team=? AND away_score>home_score) THEN 1 ELSE 0 END) w
               FROM matches WHERE tournament='FIFA World Cup' AND home_score IS NOT NULL
                 AND (home_team=? OR away_team=?)""", (name, name, name, name)).fetchone()
        d["wc_history"] = {"matches": wc["n"] or 0, "wins": wc["w"] or 0}

        # this team's 3 group fixtures + predictions
        fixtures = conn.execute(
            """SELECT m.date,m.home_team,m.away_team,m.city,m.country,m.neutral,
                      p.p_home,p.p_draw,p.p_away,p.exp_home,p.exp_away,p.top_scores
               FROM matches m LEFT JOIN predictions p
                 ON p.home_team=m.home_team AND p.away_team=m.away_team AND p.model_run_id=?
               WHERE m.stage='group' AND (m.home_team=? OR m.away_team=?)
               ORDER BY m.date""", (rid, name, name)).fetchall()
        d["fixtures"] = [_fixture_row(dict(r)) for r in fixtures]
        d["squad"] = _load_squad(name)
        return d


def _load_squad(team: str):
    """Web-verified 26-man squad from data/squads.json (or [] if not yet added)."""
    f = config.DATA / "squads.json"
    if not f.exists():
        return []
    try:
        data = json.loads(f.read_text(encoding="utf-8"))
        return data.get("squads", {}).get(team, [])
    except Exception:
        return []


def _form_row(r: dict, team: str) -> dict:
    opp = r["away_team"] if r["home_team"] == team else r["home_team"]
    gf = r["home_score"] if r["home_team"] == team else r["away_score"]
    ga = r["away_score"] if r["home_team"] == team else r["home_score"]
    res = "W" if gf > ga else ("D" if gf == ga else "L")
    return {"date": r["date"], "opponent": opp, "gf": gf, "ga": ga,
            "result": res, "tournament": r["tournament"]}


# -------------------------------------------------------------- fixtures
def _fixture_row(r: dict) -> dict:
    r = dict(r)
    if r.get("top_scores"):
        r["top_scores"] = json.loads(r["top_scores"])
    return r


@app.get("/api/fixtures")
def fixtures():
    with connect() as conn:
        rid = latest_run_id(conn)
        rows = conn.execute(
            """SELECT m.date,m.home_team,m.away_team,m.city,m.country,m.neutral,
                      p.p_home,p.p_draw,p.p_away,p.exp_home,p.exp_away,p.top_scores,
                      p.over25,p.btts,
                      th.group_label
               FROM matches m
               LEFT JOIN predictions p
                 ON p.home_team=m.home_team AND p.away_team=m.away_team AND p.model_run_id=?
               LEFT JOIN teams th ON th.name=m.home_team
               WHERE m.stage='group' ORDER BY m.date,m.id""", (rid,)).fetchall()
    kmap = _kickoff_map()
    out = []
    for r in rows:
        d = _fixture_row(dict(r))
        d["kickoff"] = kmap.get(f"{d['home_team']}__{d['away_team']}")
        out.append(d)
    return out


# ---------------------------------------------------------------- groups
@app.get("/api/groups")
def groups():
    with connect() as conn:
        rid = latest_run_id(conn)
        teams = conn.execute(
            """SELECT t.name,t.group_label,t.confederation,t.elo,
                      s.p_group_win,s.p_advance,s.p_win
               FROM teams t LEFT JOIN simulations s
                 ON s.team=t.name AND s.run_id=?""", (rid,)).fetchall()
    g: dict[str, list] = {}
    for r in teams:
        d = dict(r)
        d["attack"], d["defence"] = attack_defence(d["name"])
        g.setdefault(d["group_label"], []).append(d)
    out = []
    for label in sorted(g):
        members = sorted(g[label], key=lambda x: (x["p_advance"] or 0), reverse=True)
        out.append({"group": label, "teams": members})
    return out


# -------------------------------------------------------------- predict
class PredictReq(BaseModel):
    home: str
    away: str
    neutral: bool = True


@app.get("/api/known_teams")
def known_teams():
    return sorted(MODEL.teams)


@app.post("/api/predict")
def predict(req: PredictReq):
    if not MODEL.can_predict(req.home):
        raise HTTPException(400, f"unknown team: {req.home}")
    if not MODEL.can_predict(req.away):
        raise HTTPException(400, f"unknown team: {req.away}")
    if req.home == req.away:
        raise HTTPException(400, "pick two different teams")
    return MODEL.match_summary(req.home, req.away, req.neutral)


# ------------------------------------------------------------- simulate
@app.get("/api/simulate")
def simulate():
    with connect() as conn:
        rid = latest_run_id(conn)
        rows = conn.execute(
            "SELECT * FROM simulations WHERE run_id=? ORDER BY p_win DESC", (rid,)
        ).fetchall()
    return [dict(r) for r in rows]


# -------------------------------------------------------------- bracket
@app.get("/api/bracket")
def bracket():
    """Projected bracket using each group's most-likely winner / runner-up."""
    with connect() as conn:
        rid = latest_run_id(conn)
        rows = conn.execute(
            "SELECT team,group_label,p_group_win,p_advance,p_win FROM simulations "
            "WHERE run_id=?", (rid,)).fetchall()
    by_group: dict[str, list] = {}
    for r in rows:
        by_group.setdefault(r["group_label"], []).append(dict(r))
    winner, runner = {}, {}
    for gl, members in by_group.items():
        order = sorted(members, key=lambda x: x["p_advance"], reverse=True)
        winner[gl] = sorted(members, key=lambda x: x["p_group_win"], reverse=True)[0]["team"]
        runner[gl] = next(m["team"] for m in order if m["team"] != winner[gl])
    # best-8 thirds by p_advance among 3rd-likely teams (approx: highest p_advance
    # teams not already winner/runner)
    used = set(winner.values()) | set(runner.values())
    thirds = sorted((m for ms in by_group.values() for m in ms if m["team"] not in used),
                    key=lambda x: x["p_advance"], reverse=True)[:8]
    third_groups = [t["group_label"] for t in thirds]
    third_team = {t["group_label"]: t["team"] for t in thirds}

    G = nx.Graph()
    slots = list(B.THIRD_SLOTS)
    for mid in slots:
        for grp in third_groups:
            if grp in B.THIRD_SLOTS[mid]:
                G.add_edge(("slot", mid), ("grp", grp))
    match = nx.bipartite.maximum_matching(G, top_nodes={("grp", g) for g in third_groups}) \
        if third_groups else {}
    slot_third = {mid: match[("slot", mid)][1] for mid in slots if ("slot", mid) in match}

    def resolve(slot, mid):
        kind, val = slot
        if kind == "W":
            return winner.get(val)
        if kind == "RU":
            return runner.get(val)
        return third_team.get(slot_third.get(mid))

    r32 = []
    for mid, (sa, sb) in sorted(B.R32.items()):
        r32.append({"match": mid, "stage": "R32",
                    "home": resolve(sa, mid), "away": resolve(sb, mid)})
    return {"round_of_32": r32, "feed": B.FEED,
            "winner": winner, "runner_up": runner, "thirds": third_team}


# ----------------------------------------------------------- calibration
@app.get("/api/calibration")
def calibration():
    with connect() as conn:
        rid = latest_run_id(conn)
        run = dict(conn.execute(
            "SELECT * FROM model_runs WHERE id=?", (rid,)).fetchone())
        run["notes"] = json.loads(run["notes"]) if run.get("notes") else {}
        bins = [dict(r) for r in conn.execute(
            "SELECT bin_lo,bin_hi,predicted,observed,count FROM calibration "
            "WHERE run_id=? ORDER BY bin_lo", (rid,)).fetchall()]
    return {"run": run, "reliability": bins}


def _load_site_password() -> str:
    import os
    p = os.environ.get("SITE_PASSWORD")
    if p:
        return p.strip()
    f = Path(__file__).resolve().parent / "site_password.txt"
    if f.exists():
        return f.read_text(encoding="utf-8").strip()
    return ""


class AuthReq(BaseModel):
    password: str


@app.get("/api/auth/required")
def auth_required():
    return {"required": bool(_load_site_password())}


@app.post("/api/auth")
def auth(req: AuthReq):
    pw = _load_site_password()
    if not pw:
        return {"ok": True, "required": False}
    return {"ok": req.password == pw, "required": True}


@app.get("/api/news")
def news():
    """Upcoming-team news, read from data/news.json (refreshed by the daily
    briefing cron / WebSearch). Static file so it always serves reliably."""
    f = config.DATA / "news.json"
    if not f.exists():
        return {"updated": None, "items": []}
    try:
        return json.loads(f.read_text(encoding="utf-8"))
    except Exception:
        return {"updated": None, "items": []}


@app.get("/api/health")
def health():
    return {"ok": True, "teams_in_model": len(MODEL.teams)}


@app.post("/api/reload")
def reload_model():
    """重训后热加载新模型 + 清缓存(无需重启后端)。"""
    global MODEL, PARAMS
    MODEL = WorldCupModel.load()
    PARAMS = MODEL.model.get_params()
    _cache.clear()
    _winner_cache["data"] = None
    _odds_cache["data"] = None
    return {"ok": True, "teams_in_model": len(MODEL.teams)}


# --------------------------------------------------------- live scores
_live_cache: dict = {"ts": 0, "data": []}
LIVE_TTL = 60  # seconds

@app.get("/api/live")
def live(date: str | None = None):
    """Live / today scores via TheSportsDB (free, no key). Cached 60s."""
    import datetime
    d = date or datetime.date.today().isoformat()
    now = time.time()
    if _live_cache["ts"] > now - LIVE_TTL and _live_cache.get("date") == d:
        return _live_cache["data"]
    try:
        url = f"https://www.thesportsdb.com/api/v1/json/3/eventsday.php?d={d}&s=Soccer"
        resp = httpx.get(url, timeout=12)
        events = resp.json().get("events") or []
        wc = [e for e in events if "World Cup" in (e.get("strLeague") or "")]
        out = []
        for e in wc:
            out.append({
                "home_team": e.get("strHomeTeam", ""),
                "away_team": e.get("strAwayTeam", ""),
                "home_score": e.get("intHomeScore"),
                "away_score": e.get("intAwayScore"),
                "status": e.get("strStatus", "NS"),  # NS/FT/HT/1H/2H/ET...
                "time": e.get("strTime", ""),
                "venue": e.get("strVenue", ""),
                "date": d,
            })
        _live_cache.update(ts=now, data=out, date=d)
        return out
    except Exception:
        return _live_cache.get("data", [])


# --------------------------------------------- the-odds-api 实时赔率
def _load_odds_key() -> str:
    import os
    k = os.environ.get("ODDS_API_KEY")
    if k:
        return k.strip()
    f = Path(__file__).resolve().parent / "odds_key.txt"
    if f.exists():
        return f.read_text(encoding="utf-8").strip()
    return ""

ODDS_API_KEY = _load_odds_key()
# the-odds-api 队名 -> 本库队名
ODDS_NAME_MAP = {
    "Bosnia & Herzegovina": "Bosnia and Herzegovina",
    "USA": "United States",
}
_odds_cache: dict = {"ts": 0, "data": None}
ODDS_TTL = 3 * 3600  # 3h,省配额(免费500/月)


def _devig(oh, od, oa):
    """De-margin three decimal odds -> normalized true probabilities."""
    inv = [1 / oh, 1 / od, 1 / oa]
    s = sum(inv)
    return [x / s for x in inv]


@app.get("/api/odds")
def odds(refresh: bool = False):
    """实时胜平负赔率(the-odds-api 全球庄家) + 模型对比 + EV 价值。缓存3小时。"""
    now = time.time()
    if not refresh and _odds_cache["data"] is not None and _odds_cache["ts"] > now - ODDS_TTL:
        return _odds_cache["data"]
    if not ODDS_API_KEY:
        return {"available": False, "note": "未配置 the-odds-api key", "matches": [], "quota": None}

    try:
        url = (f"https://api.the-odds-api.com/v4/sports/soccer_fifa_world_cup/odds/"
               f"?apiKey={ODDS_API_KEY}&regions=eu&markets=h2h&oddsFormat=decimal")
        r = httpx.get(url, timeout=15)
        quota = r.headers.get("x-requests-remaining")
        if r.status_code != 200:
            return {"available": False, "note": f"赔率接口返回 {r.status_code}", "matches": [], "quota": quota}
        raw = r.json()
    except Exception as e:
        return {"available": False, "note": f"获取失败({type(e).__name__})", "matches": [], "quota": None}

    out = []
    for m in raw:
        home = ODDS_NAME_MAP.get(m["home_team"], m["home_team"])
        away = ODDS_NAME_MAP.get(m["away_team"], m["away_team"])
        # collect best & average odds across bookmakers
        bh, bd, ba, sh, sd, sa, n = 0, 0, 0, 0.0, 0.0, 0.0, 0
        for bk in m.get("bookmakers", []):
            mk = next((x for x in bk.get("markets", []) if x["key"] == "h2h"), None)
            if not mk:
                continue
            od = {o["name"]: o["price"] for o in mk["outcomes"]}
            oh, od_, oa = od.get(m["home_team"]), od.get("Draw"), od.get(m["away_team"])
            if not (oh and od_ and oa):
                continue
            bh, bd, ba = max(bh, oh), max(bd, od_), max(ba, oa)
            sh += oh; sd += od_; sa += oa; n += 1
        if n == 0:
            continue
        avg = (sh / n, sd / n, sa / n)
        imp = _devig(*avg)
        mp = None
        if MODEL.can_predict(home, away):
            g = MODEL.predict_grid(home, away, neutral=True)
            mp = [float(x) for x in g.home_draw_away]
        ev = None
        if mp:
            ev = [round(mp[0] * bh - 1, 4), round(mp[1] * bd - 1, 4), round(mp[2] * ba - 1, 4)]
        out.append({
            "home_team": home, "away_team": away, "commence_time": m.get("commence_time"),
            "best": {"home": round(bh, 2), "draw": round(bd, 2), "away": round(ba, 2)},
            "avg": {"home": round(avg[0], 2), "draw": round(avg[1], 2), "away": round(avg[2], 2)},
            "implied": {"home": round(imp[0], 4), "draw": round(imp[1], 4), "away": round(imp[2], 4)},
            "model": None if not mp else {"home": round(mp[0], 4), "draw": round(mp[1], 4), "away": round(mp[2], 4)},
            "ev": None if not ev else {"home": ev[0], "draw": ev[1], "away": ev[2]},
            "n_books": n,
        })
    data = {"available": True, "note": f"{len(out)} 场 · the-odds-api", "matches": out, "quota": quota}
    _odds_cache.update(ts=now, data=data)
    return data


# ------------------------------------------- 夺冠盘(futures) & 模型×市场融合
_winner_cache: dict = {"ts": 0, "data": None}
WINNER_TTL = 6 * 3600


def _fetch_winner_market():
    """De-vigged champion probabilities from the-odds-api winner market."""
    now = time.time()
    if _winner_cache["data"] is not None and _winner_cache["ts"] > now - WINNER_TTL:
        return _winner_cache["data"]
    out = {"probs": {}, "odds": {}, "quota": None, "ok": False}
    if not ODDS_API_KEY:
        return out
    try:
        url = (f"https://api.the-odds-api.com/v4/sports/soccer_fifa_world_cup_winner/"
               f"odds/?apiKey={ODDS_API_KEY}&regions=eu&markets=outrights&oddsFormat=decimal")
        r = httpx.get(url, timeout=15)
        out["quota"] = r.headers.get("x-requests-remaining")
        data = r.json()
        if data and data[0].get("bookmakers"):
            bk = data[0]["bookmakers"][0]
            outs = bk["markets"][0]["outcomes"]
            inv = {}
            for o in outs:
                name = ODDS_NAME_MAP.get(o["name"], o["name"])
                inv[name] = 1.0 / o["price"]
                out["odds"][name] = o["price"]
            s = sum(inv.values())
            out["probs"] = {k: v / s for k, v in inv.items()}
            out["ok"] = True
    except Exception:
        pass
    _winner_cache.update(ts=now, data=out)
    return out


@app.get("/api/champion")
def champion(weight: float = 0.5):
    """综合夺冠概率 = (1-w)·模型 + w·市场,按队归一化。w 默认 0.5。"""
    cache_key = f"champion_{weight}"
    now = time.time()
    if _cache.get(cache_key) and _cache[cache_key]["ts"] > now - 1800:
        return _cache[cache_key]["data"]
    with connect() as conn:
        rid = latest_run_id(conn)
        rows = conn.execute(
            "SELECT team,group_label,p_win FROM simulations WHERE run_id=?", (rid,)
        ).fetchall()
    model = {r["team"]: r["p_win"] for r in rows}
    grp = {r["team"]: r["group_label"] for r in rows}

    mk = _fetch_winner_market()
    w = 0.0 if not mk["ok"] else max(0.0, min(1.0, weight))

    raw = {t: (1 - w) * model[t] + w * mk["probs"].get(t, 0.0) for t in model}
    z = sum(raw.values()) or 1.0
    out = [{
        "team": t, "group": grp[t],
        "model": round(model[t], 4),
        "market": round(mk["probs"].get(t, 0.0), 4) if mk["ok"] else None,
        "market_odds": mk["odds"].get(t),
        "blended": round(raw[t] / z, 4),
    } for t in model]
    out.sort(key=lambda x: x["blended"], reverse=True)
    result = {"weight": w, "market_available": mk["ok"], "quota": mk["quota"], "teams": out}
    _cache[cache_key] = {"ts": now, "data": result}
    return result


# ------------------------------------------------- 竞彩 odds (体彩) attempt
_jc_cache: dict = {"ts": 0, "data": None}
JC_TTL = 1800  # 30 min

@app.get("/api/jc_odds")
def jc_odds():
    """Attempt to pull official China Sports Lottery (竞彩足球) win/draw/loss odds.

    sporttery.cn is protected by a 瑞数 dynamic anti-bot WAF that blocks
    server-side and headless requests, so this usually returns available=False
    in hosted/sandboxed environments. On a real China residential network it
    may succeed, in which case odds auto-fill the value page. Cached 30 min."""
    now = time.time()
    if _jc_cache["data"] is not None and _jc_cache["ts"] > now - JC_TTL:
        return _jc_cache["data"]

    result = {"available": False, "note": "官方接口受反爬保护,未能自动获取,请手动填写赔率", "matches": []}
    try:
        url = ("https://webapi.sporttery.cn/gateway/jc/football/getMatchListV1.qry"
               "?poolCode=had,crs&channel=c")
        r = httpx.get(url, timeout=8, headers={
            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
                          "AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0 Safari/537.36",
            "Referer": "https://www.sporttery.cn/",
            "Accept": "application/json, text/plain, */*",
        })
        if r.headers.get("content-type", "").startswith("application/json"):
            data = r.json()
            matches = []
            for mi in (data.get("value", {}).get("matchInfoList") or []):
                for sm in (mi.get("subMatchList") or []):
                    had = sm.get("had") or {}
                    matches.append({
                        "home_team": sm.get("homeTeamAllName") or sm.get("homeTeamAbbName"),
                        "away_team": sm.get("awayTeamAllName") or sm.get("awayTeamAbbName"),
                        "odds_home": _f(had.get("h")), "odds_draw": _f(had.get("d")),
                        "odds_away": _f(had.get("a")),
                    })
            if matches:
                result = {"available": True, "note": "已获取官方竞彩赔率", "matches": matches}
    except Exception as e:
        result["note"] = f"自动获取失败({type(e).__name__}),请手动填写赔率"

    _jc_cache.update(ts=now, data=result)
    return result


def _f(x):
    try:
        return float(x)
    except (TypeError, ValueError):
        return None


# ------------------------------------------------------ live standings
@app.get("/api/standings")
def standings():
    """Per-group live standings from played matches + advancement probabilities.
    Before/while group stage is incomplete, rows still rank by model p_advance."""
    with connect() as conn:
        rid = latest_run_id(conn)
        played = conn.execute(
            "SELECT home_team,away_team,home_score,away_score FROM matches "
            "WHERE stage='group' AND home_score IS NOT NULL").fetchall()
        teaminfo = {r["name"]: dict(r) for r in conn.execute(
            "SELECT name,group_label,confederation,elo FROM teams").fetchall()}
        sims = {r["team"]: dict(r) for r in conn.execute(
            "SELECT team,p_group_win,p_advance,p_r16,p_win FROM simulations "
            "WHERE run_id=?", (rid,)).fetchall()}

    # init rows
    rows: dict[str, dict] = {}
    for name, info in teaminfo.items():
        s = sims.get(name, {})
        rows[name] = {
            "team": name, "group": info["group_label"],
            "confederation": info["confederation"], "elo": info["elo"],
            "played": 0, "win": 0, "draw": 0, "loss": 0,
            "gf": 0, "ga": 0, "gd": 0, "pts": 0,
            "p_group_win": s.get("p_group_win"), "p_advance": s.get("p_advance"),
            "p_win": s.get("p_win"),
        }

    for r in played:
        h, a = r["home_team"], r["away_team"]
        if h not in rows or a not in rows:
            continue
        hs, as_ = r["home_score"], r["away_score"]
        for t in (h, a):
            rows[t]["played"] += 1
        rows[h]["gf"] += hs; rows[h]["ga"] += as_
        rows[a]["gf"] += as_; rows[a]["ga"] += hs
        if hs > as_:
            rows[h]["win"] += 1; rows[h]["pts"] += 3; rows[a]["loss"] += 1
        elif hs < as_:
            rows[a]["win"] += 1; rows[a]["pts"] += 3; rows[h]["loss"] += 1
        else:
            rows[h]["draw"] += 1; rows[a]["draw"] += 1
            rows[h]["pts"] += 1; rows[a]["pts"] += 1
    for t in rows:
        rows[t]["gd"] = rows[t]["gf"] - rows[t]["ga"]

    any_played = any(r["played"] for r in rows.values())

    # group into 12 tables, rank
    groups: dict[str, list] = {}
    for r in rows.values():
        groups.setdefault(r["group"], []).append(r)

    def sort_key(x):
        if any_played:
            return (x["pts"], x["gd"], x["gf"], x["p_advance"] or 0)
        return (x["p_advance"] or 0,)  # pre-tournament: project by model

    # rank thirds across groups for best-8 highlight
    third_rows = []
    out = []
    for gl in sorted(groups):
        members = sorted(groups[gl], key=sort_key, reverse=True)
        for i, m in enumerate(members):
            m["rank"] = i + 1
            m["qualify"] = "auto" if i < 2 else ("third" if i == 2 else "out")
        third_rows.append(members[2])  # 3rd-placed team of this group
        out.append({"group": gl, "teams": members})

    # best-8 thirds
    third_sorted = sorted(
        third_rows,
        key=lambda x: ((x["pts"], x["gd"], x["gf"]) if any_played else (x["p_advance"] or 0,)),
        reverse=True)
    best8 = {t["team"] for t in third_sorted[:8]}
    for grp in out:
        for m in grp["teams"]:
            if m["qualify"] == "third":
                m["third_in8"] = m["team"] in best8

    return {"groups": out, "any_played": any_played,
            "thirds_race": [
                {"team": t["team"], "group": t["group"], "pts": t["pts"],
                 "gd": t["gd"], "gf": t["gf"], "p_advance": t["p_advance"],
                 "in8": t["team"] in best8}
                for t in third_sorted]}


# ------------------------------------------------- full schedule (group+KO)
STAGE_ORDER = {"group": 0, "R32": 1, "R16": 2, "QF": 3, "SF": 4, "3rd": 5, "F": 6}
STAGE_LABEL = {"group": "小组赛", "R32": "32强", "R16": "16强",
               "QF": "四分之一决赛", "SF": "半决赛", "3rd": "三四名", "F": "决赛"}


def _compute_group_standings():
    """From played group-stage matches, compute standings per group."""
    with connect() as conn:
        played = conn.execute(
            "SELECT home_team,away_team,home_score,away_score FROM matches "
            "WHERE stage='group' AND home_score IS NOT NULL"
        ).fetchall()
        team_group = {r["name"]: r["group_label"]
                      for r in conn.execute("SELECT name,group_label FROM teams").fetchall()}
    tables: dict[str, dict[str, dict]] = {}
    for r in played:
        g = team_group.get(r["home_team"])
        if not g:
            continue
        for t in (r["home_team"], r["away_team"]):
            tables.setdefault(g, {}).setdefault(t, {"pts": 0, "gf": 0, "ga": 0, "gd": 0})
        hs, as_ = r["home_score"], r["away_score"]
        tables[g][r["home_team"]]["gf"] += hs
        tables[g][r["home_team"]]["ga"] += as_
        tables[g][r["away_team"]]["gf"] += as_
        tables[g][r["away_team"]]["ga"] += hs
        if hs > as_:
            tables[g][r["home_team"]]["pts"] += 3
        elif hs < as_:
            tables[g][r["away_team"]]["pts"] += 3
        else:
            tables[g][r["home_team"]]["pts"] += 1
            tables[g][r["away_team"]]["pts"] += 1
    for g in tables:
        for t in tables[g]:
            tables[g][t]["gd"] = tables[g][t]["gf"] - tables[g][t]["ga"]
    return tables


def _resolve_bracket_from_standings(standings):
    """Given group standings, resolve R32 teams using actual results."""
    winners, runners = {}, {}
    all_thirds = []
    for g, tbl in standings.items():
        order = sorted(tbl, key=lambda t: (tbl[t]["pts"], tbl[t]["gd"], tbl[t]["gf"]),
                       reverse=True)
        if len(order) >= 2:
            winners[g] = order[0]
            runners[g] = order[1]
        if len(order) >= 3:
            all_thirds.append((g, order[2], tbl[order[2]]))
    all_thirds.sort(key=lambda x: (x[2]["pts"], x[2]["gd"], x[2]["gf"]), reverse=True)
    best8 = all_thirds[:8]
    third_team = {g: t for g, t, _ in best8}
    third_groups = [g for g, _, _ in best8]

    G = nx.Graph()
    slots = list(B.THIRD_SLOTS)
    for mid in slots:
        for grp in third_groups:
            if grp in B.THIRD_SLOTS[mid]:
                G.add_edge(("slot", mid), ("grp", grp))
    matching = nx.bipartite.maximum_matching(G, top_nodes={("grp", g) for g in third_groups}) \
        if third_groups else {}
    slot_third = {mid: matching[("slot", mid)][1] for mid in slots if ("slot", mid) in matching}
    third_for_slot = {mid: third_team.get(slot_third.get(mid)) for mid in slots}

    return winners, runners, third_for_slot


def _kickoff_map():
    """Map '{home}__{away}' -> ISO kickoff time (UTC) from the-odds-api data."""
    data = _odds_cache.get("data")
    if not data or not data.get("matches"):
        try:
            data = odds()
        except Exception:
            data = None
    m = {}
    if data and data.get("matches"):
        for x in data["matches"]:
            ct = x.get("commence_time")
            if ct:
                m[f"{x['home_team']}__{x['away_team']}"] = ct
    return m


@app.get("/api/schedule")
def full_schedule():
    """All fixtures: group-stage (with predictions) + knockout (with on-the-fly
    predictions once teams are determined from real results or simulation."""
    now = time.time()
    if _cache.get("schedule") and _cache["schedule"]["ts"] > now - 1800:
        return _cache["schedule"]["data"]
    with connect() as conn:
        rid = latest_run_id(conn)
        # group-stage fixtures
        gs = conn.execute(
            """SELECT m.date,m.home_team,m.away_team,m.city,m.country,m.neutral,m.stage,
                      m.home_score,m.away_score,
                      p.p_home,p.p_draw,p.p_away,p.exp_home,p.exp_away,p.top_scores,
                      p.over25,p.btts,
                      th.group_label
               FROM matches m
               LEFT JOIN predictions p
                 ON p.home_team=m.home_team AND p.away_team=m.away_team AND p.model_run_id=?
               LEFT JOIN teams th ON th.name=m.home_team
               WHERE m.stage='group' ORDER BY m.date,m.id""", (rid,)).fetchall()
        # Round-of-32 fixtures with real dates/scores. The 16 chronologically
        # first knockout games are the R32; later rounds are excluded so a
        # mis-resolved third-place slot can't borrow a later round's date.
        ko_rows = conn.execute(
            """SELECT date,home_team,away_team,home_score,away_score
               FROM matches WHERE stage='knockout'
               ORDER BY date, id LIMIT 16""").fetchall()

    # lookup: "{home}__{away}" -> real R32 date/score, so resolved R32 slots can
    # be placed on the calendar (else the dashboard's "即将开赛" section has no
    # dated knockout match and falls back to the last group matchday).
    ko_actual = {f"{r['home_team']}__{r['away_team']}": dict(r) for r in ko_rows}
    kmap = _kickoff_map()
    out = []
    for r in gs:
        d = _fixture_row({**dict(r), "stage_label": "小组赛"})
        d["kickoff"] = kmap.get(f"{d['home_team']}__{d['away_team']}")
        out.append(d)

    # Attempt to fill knockout slots from actual standings
    standings = _compute_group_standings()
    group_complete = all(
        len([t for t in tbl if tbl[t]["pts"] + tbl[t]["gf"] + tbl[t]["ga"] > 0]) >= 4
        and sum(1 for t in tbl for _ in range(1)) >= 4  # at least 4 teams played
        for tbl in standings.values()
    ) if len(standings) == 12 else False

    # Knockout teams are only shown once decided by real group results.
    # Before that, all knockout slots stay 待定 (no projected matchups).
    if group_complete:
        winners, runners, third_for_slot = _resolve_bracket_from_standings(standings)
    else:
        winners, runners, third_for_slot = {}, {}, {}

    # Build KO fixtures
    ko_teams: dict[int, tuple[str | None, str | None]] = {}
    for mid, (sa, sb) in sorted(B.R32.items()):
        ta = _resolve_slot(sa, mid, winners, runners, third_for_slot)
        tb = _resolve_slot(sb, mid, winners, runners, third_for_slot)
        ko_teams[mid] = (ta, tb)

    # Propagate through bracket
    ko_winners: dict[int, str | None] = {}
    for mid in (89, 90, 91, 92, 93, 94, 95, 96,
                97, 98, 99, 100, 101, 102, 104):
        a, b = B.FEED[mid]
        ko_teams[mid] = (ko_winners.get(a), ko_winners.get(b))

    # For each KO match, add on-the-fly prediction if both teams known
    for mid in sorted(ko_teams):
        ta, tb = ko_teams[mid]
        stage = B.STAGE_OF.get(mid, "KO")
        pred_data = {}
        if ta and tb and MODEL.can_predict(ta, tb):
            s = MODEL.match_summary(ta, tb, neutral=True)
            pred_data = {
                "p_home": s["p_home"], "p_draw": s["p_draw"], "p_away": s["p_away"],
                "exp_home": s["exp_home"], "exp_away": s["exp_away"],
                "top_scores": s["top_scores"], "over25": s["over25"], "btts": s["btts"],
            }
        # Once both teams are decided, attach the real fixture date/kickoff/score.
        kdate = khs = kas = kickoff = None
        if ta and tb:
            act = ko_actual.get(f"{ta}__{tb}")
            if act:
                kdate, khs, kas = act["date"], act["home_score"], act["away_score"]
            else:
                act = ko_actual.get(f"{tb}__{ta}")  # stored other way round
                if act:
                    kdate, khs, kas = act["date"], act["away_score"], act["home_score"]
            kickoff = kmap.get(f"{ta}__{tb}") or kmap.get(f"{tb}__{ta}")
        out.append({
            "date": kdate, "home_team": ta, "away_team": tb,
            "city": None, "country": None, "neutral": 1,
            "stage": stage, "stage_label": STAGE_LABEL.get(stage, stage),
            "match_id": mid,
            "home_score": khs, "away_score": kas,
            "group_label": None, "kickoff": kickoff,
            **pred_data,
        })

    _cache["schedule"] = {"ts": time.time(), "data": out}
    return out


@app.get("/api/ko_bracket")
def ko_bracket_view():
    """Round-of-32 bracket with real results + score predictions, and the R16
    pairings formed by the R32 winners. Built deterministically from the real
    knockout fixtures in the DB (authoritative pairings/scores) laid over the
    official bracket template — so third-place slots and penalty-shootout
    advancers come from actual data, not the flaky projection matching."""
    now = time.time()
    if _cache.get("ko_bracket") and _cache["ko_bracket"]["ts"] > now - 900:
        return _cache["ko_bracket"]["data"]

    with connect() as conn:
        ko = conn.execute(
            "SELECT date,home_team,away_team,home_score,away_score FROM matches "
            "WHERE stage='knockout' ORDER BY date, id").fetchall()
    r32_rows = [dict(r) for r in ko[:16]]        # 16 chronologically-first = R32
    later_rows = [dict(r) for r in ko[16:]]      # R16+ scheduled so far

    # teams appearing in a later round have, by definition, advanced past R32 —
    # this resolves penalty-shootout winners that a level score can't.
    advanced = {t for r in later_rows for t in (r["home_team"], r["away_team"])}
    by_team = {}                                 # team -> its R32 fixture row
    for r in r32_rows:
        by_team.setdefault(r["home_team"], r)
        by_team.setdefault(r["away_team"], r)

    winners, runners, _ = _resolve_bracket_from_standings(_compute_group_standings())
    kmap = _kickoff_map()

    def _slot_team(slot):
        kind, val = slot
        if kind == "W":
            return winners.get(val)
        if kind == "RU":
            return runners.get(val)
        return None                              # "3": resolved from the DB row

    def _actual_winner(row):
        hs, as_ = row["home_score"], row["away_score"]
        if hs is None or as_ is None:
            return None
        if hs != as_:
            return row["home_team"] if hs > as_ else row["away_team"]
        # level score -> decided on penalties; use who turns up in the next round
        h_adv, a_adv = row["home_team"] in advanced, row["away_team"] in advanced
        if h_adv != a_adv:
            return row["home_team"] if h_adv else row["away_team"]
        return None

    r32, winner_by_mid = [], {}
    for mid in sorted(B.R32):
        sA, sB = B.R32[mid]
        ta, tb = _slot_team(sA), _slot_team(sB)
        row = by_team.get(ta) or by_team.get(tb)
        date = kickoff = hs = as_ = winner = pred = None
        played = False
        if row:
            # fill an unknown (third-place) slot from the real opponent
            other = row["away_team"] if row["home_team"] in (ta, tb) else row["home_team"]
            if ta is None:
                ta = other
            if tb is None:
                tb = other
            date = row["date"]
            kickoff = (kmap.get(f"{row['home_team']}__{row['away_team']}")
                       or kmap.get(f"{row['away_team']}__{row['home_team']}"))
            if row["home_score"] is not None:
                played = True
                hs, as_ = ((row["home_score"], row["away_score"])
                           if row["home_team"] == ta
                           else (row["away_score"], row["home_score"]))
            winner = _actual_winner(row)
        if ta and tb and MODEL.can_predict(ta, tb):
            s = MODEL.match_summary(ta, tb, neutral=True)
            top = (s.get("top_scores") or [None])[0]
            pred = {"p_home": s["p_home"], "p_draw": s["p_draw"], "p_away": s["p_away"],
                    "score": [top[0], top[1]] if top else None}
            if winner is None and not played:            # project the not-yet-played winner
                winner = ta if s["p_home"] >= s["p_away"] else tb
        winner_by_mid[mid] = winner
        r32.append({
            "match_id": mid, "home": ta, "away": tb,
            "home_score": hs, "away_score": as_,
            "date": date, "kickoff": kickoff,
            "played": played, "winner": winner, "pred": pred,
        })

    r16 = [{"match_id": mid, "feeders": list(B.FEED[mid]),
            "home": winner_by_mid.get(B.FEED[mid][0]),
            "away": winner_by_mid.get(B.FEED[mid][1])}
           for mid in (89, 90, 91, 92, 93, 94, 95, 96)]

    data = {"r32": r32, "r16": r16}
    _cache["ko_bracket"] = {"ts": time.time(), "data": data}
    return data


def _get_projected_bracket():
    """Helper: get bracket from simulation projections."""
    with connect() as conn:
        rid = latest_run_id(conn)
        rows = conn.execute(
            "SELECT team,group_label,p_group_win,p_advance FROM simulations "
            "WHERE run_id=?", (rid,)).fetchall()
    by_group: dict[str, list] = {}
    for r in rows:
        by_group.setdefault(r["group_label"], []).append(dict(r))
    winner, runner = {}, {}
    for gl, members in by_group.items():
        order = sorted(members, key=lambda x: x["p_advance"], reverse=True)
        winner[gl] = sorted(members, key=lambda x: x["p_group_win"], reverse=True)[0]["team"]
        runner[gl] = next(m["team"] for m in order if m["team"] != winner[gl])
    used = set(winner.values()) | set(runner.values())
    thirds = sorted((m for ms in by_group.values() for m in ms if m["team"] not in used),
                    key=lambda x: x["p_advance"], reverse=True)[:8]
    third_groups = [t["group_label"] for t in thirds]
    third_team = {t["group_label"]: t["team"] for t in thirds}

    G = nx.Graph()
    slots = list(B.THIRD_SLOTS)
    for mid in slots:
        for grp in third_groups:
            if grp in B.THIRD_SLOTS[mid]:
                G.add_edge(("slot", mid), ("grp", grp))
    matching = nx.bipartite.maximum_matching(G, top_nodes={("grp", g) for g in third_groups}) \
        if third_groups else {}
    slot_third = {mid: matching[("slot", mid)][1] for mid in slots if ("slot", mid) in matching}
    third_for_slot = {mid: third_team.get(slot_third.get(mid)) for mid in slots}

    return {"winner": winner, "runner_up": runner, "thirds": third_team,
            "third_for_slot": third_for_slot}


def _resolve_slot(slot, mid, winners, runners, third_for_slot):
    kind, val = slot
    if kind == "W":
        return winners.get(val)
    if kind == "RU":
        return runners.get(val)
    if kind == "3":
        return third_for_slot.get(mid)
    return None