Track record

Every prediction timestamped before kick-off and settled automatically. Hit rate, calibration and return are measured against the closing market price.

Window Jun 7 – Oct 20

Hit rate: 49.8%; 558 of 1120 picks
Return · Closing price: -12.8%; flat stake, closing odds, 585 bets
Settled: 1,120; 1X2 picks with a final score
Longest losing run: 5

Monthly hit rate

JunJulAugSepOct

Small months swing wildly; read them with the sample size in view.

Latest results

57 matches rated today

Create a free account and see every prediction for 7 days. No card needed.

Start 7 days free Already a member? Sign in

Calibration, leagues and marketsThe full report — calibration curves, per-league and per-market records, return at opening and closing market prices, weekly review.

1,120 decided matches · model version dc-1.0, dc-1.2-cal. 0 rows had no stored outcome flag and were re-derived from the score. 1X2 Brier is multi-class (0 best, 2 worst; 0.667 is a uniform guess). Over/Under and BTTS Brier are binary (0 to 1; 0.25 is a coin flip). The two scales are not comparable.

Monthly hit rate

orange bars = fewer than 30 matches

Month	n	W–L	Hit rate	Brier
Oct 2026	258	150–108	58.1%	0.565
Sep 2026	352	154–198	43.8%	0.628
Aug 2026	398	199–199	50.0%	0.602
Jul 2026	87	42–45	48.3%	0.605
Jun 2026	25	13–12	52.0%	0.607

Small months swing wildly; read them with the sample size in view.

Calibration

calibrated confidence vs observed hit rate

Confidence	n	Predicted	Observed	Gap
30–40%	224	35.5%	37.9%	+2
40–50%	434	47.2%	49.1%	+2
50–60%	98	51.0%	51.0%	−0
60–70%	98	68.8%	67.3%	−1
70–80%	119	76.2%	77.3%	+1
90–100%	2	100.0%	100.0%	+0

Retrospective (today's curve). A well-calibrated model sits on the diagonal: picks published at 60% should land about 60% of the time. Gaps beyond 10 points are marked. Bins with fewer than 10 picks are listed but not plotted. Calibration curve: covered, 1,098 matches, fitted Oct 19.

Walk-forward (earlier months only)

Confidence	n	Predicted	Observed	Gap
30–40%	128	34.3%	45.3%	+11
40–50%	113	44.2%	59.3%	+15
50–60%	146	54.1%	39.0%	−15
60–70%	69	66.2%	78.3%	+12
70–80%	41	74.4%	78.0%	+4
80–90%	3	81.8%	66.7%	−15
90–100%	2	100.0%	50.0%	−50

Each month is scored with a curve fitted on the months before it. 510 early matches form the warm-up and are not scored; first scored month: Sep 2026. Pick Brier raw 0.2332 → calibrated 0.2404 (binary scale).

By league

League	n	1X2	Over/Under 2.5	Both teams to score	Brier
1. Divisjon	136	55.1%75–61	71.3%97–39	69.1%94–42	0.578
Brasileirão	115	47.8%55–60	47.0%54–61	50.4%58–57	0.606
Championship	109	37.6%41–68	46.8%51–58	52.3%57–52	0.692
Ekstraklasa	95	52.6%50–45	55.8%53–42	60.0%57–38	0.596
Eliteserien	93	52.7%49–44	67.7%63–30	61.3%57–36	0.559
LaLiga	78	51.3%40–38	67.9%53–25	53.8%42–36	0.567
Eredivisie	73	46.6%34–39	75.3%55–18	74.0%54–19	0.629
Liga Portugal	67	52.2%35–32	52.2%35–32	55.2%37–30	0.547
Superligaen	66	47.0%31–35	40.9%27–39	51.5%34–32	0.652
Süper Lig	63	46.0%29–34	58.7%37–26	46.0%29–34	0.592
Premier League	61	47.5%29–32	52.5%32–29	52.5%32–29	0.617
Serie A	61	59.0%36–25	52.5%32–29	54.1%33–28	0.551
Ligue 1	57	52.6%30–27	61.4%35–22	50.9%29–28	0.613
Bundesliga	46	52.2%24–22	69.6%32–14	58.7%27–19	0.618

Only leagues with settled matches are listed. Hit rate and W–L per market; Over/Under 2.5 and both-teams-to-score are settled from the score. Green from 60%, red under 45% for 1X2 and under 50% for goal markets (10+ matches).

Strong markets

league × market · ≥15 matches · ≥70% · 180 days

Market	League	Zone	W–L	Hit rate
Over 2.5	EliteserienNOR	≥75%	14/15	93%
	2. Divisjon Avd. 2NOR	≥75%	25/29	86%
	3. Divisjon Avd. 2NOR	≥75%	37/43	86%
	Highland / LowlandSCO	≥75%	22/27	81%
	Major League SoccerUSA	≥75%	12/15	80%
BTTS Yes	Super LeagueSUI	≥70%	13/16	81%
	Major League SoccerUSA	≥70%	20/27	74%
	3. LigaGER	≥70%	16/22	73%
	Tweede DivisieNED	≥70%	23/32	72%
Under 2.5	Liga Profesional ClausuraARG	≥75%	21/27	78%
1X2	EliteserienNOR	≥70%	16/17	94%
	Liga PortugalPOR	≥70%	15/17	88%
	National DivisionLUX	≥70%	13/15	87%
	1. DivisjonNOR	≥70%	12/15	80%
	2. Divisjon Avd. 2NOR	≥70%	12/15	80%

Uncovered leagues included; women's, youth, friendly and cup matches are not showcased. A league drops off at the Monday refresh when it falls below the threshold.

By market

Market	n	W–L	Hit rate	Brier
1X2	1120	558–562	49.8%	0.602
Over/Under 2.5	1120	656–464	58.6%	0.240
Both teams to score	1120	640–480	57.1%	0.243

Over/under 2.5 and both-teams-to-score are scored on the raw model probability and settled from the final score. Their Brier is binary (0.25 = coin flip).

Flat-stake return

Closing priceAug 14 – Oct 20

Bets: 585 · 285–300
Odds coverage: 52% · 585 of 1120 decided matches have a price for this phase; 535 without a snapshot are outside the ROI denominator.
Staked: 585.00
Returned: 510.32
Profit: -74.68 (-12.8%)
Bookmaker favourite: 52.3% · Brier 0.584
Providers: 1xBet_Spain, API-Football/Bet365, Bet365_Brazil, Bet365_Spain, Bet365_UK CPM, Bet365_UK affiliate, Bet365_default, Betfair UK Sportsbook_UK, Betfred_United Kingdom, Bwin_France, Marcaapuestas_Spain, Paddy Power_UK, Sky Bet_UK, Spreadex_UK, SuperBet_Poland, Unibet_France, unknown

Opening priceAug 14 – Oct 20

Bets: 638 · 317–321
Odds coverage: 57% · 638 of 1120 decided matches have a price for this phase; 482 without a snapshot are outside the ROI denominator.
Staked: 638.00
Returned: 568.31
Profit: -69.69 (-10.9%)
Bookmaker favourite: 52.7% · Brier 0.578
Providers: 1xBet_Spain, Bet365_Brazil, Bet365_Spain, Bet365_UK CPM, Bet365_UK affiliate, Bet365_default, Betfair UK Sportsbook_UK, Betfred_United Kingdom, Bwin_France, Marcaapuestas_Spain, Paddy Power_UK, Sky Bet_UK, Spreadex_UK, SuperBet_Poland, Unibet_France, unknown

One unit on the model's 1X2 pick at the last captured odds before kick-off (closing, else opening). Only matches with an odds snapshot count, so this window is shorter than the full record. The bookmaker line shows how often the market favourite won over the same matches, with its margin-free Brier. The closing price is a benchmark of what the market knew at kick-off; the opening price is closer to what was available when the prediction was published. Neither is a realised return.

Signal report

market × league × signal · which signal actually hits?

When the model says 70% for a selection, does it hit 70%? When it disagrees with the market by 10 points, is it right? For each market these tables bucket the model probability of the chosen side (level) and the gap between model and margin-free bookmaker probability (edge, percentage points) and show the hit rate. Green from 65%, red under 50%; buckets with fewer than 10 matches are faded.

1X2 · model level

League	<50%	50–60%	60–70%	≥70%
All leagues	41.7%258–360	48.8%126–132	67.8%101–48	76.8%73–22
1. Divisjon	49.2%30–31	48.6%17–18	64.0%16–9	80.0%12–3
Brasileirão	37.9%25–41	53.1%17–15	81.8%9–2	66.7%4–2
Championship	36.5%27–47	45.5%10–12	40.0%4–6	0.0%0–3
Ekstraklasa	47.2%34–38	64.7%11–6	80.0%4–1	100.0%1–0
Eliteserien	40.0%18–27	47.4%9–10	50.0%6–6	94.1%16–1
LaLiga	37.2%16–27	41.2%7–10	92.3%12–1	100.0%5–0
Eredivisie	37.0%10–17	44.0%11–14	57.1%8–6	71.4%5–2
Liga Portugal	32.3%10–21	46.2%6–7	66.7%4–2	88.2%15–2
Superligaen	46.7%14–16	50.0%11–11	44.4%4–5	40.0%2–3
Süper Lig	26.5%9–25	50.0%6–6	85.7%12–2	66.7%2–1
Premier League	41.5%17–24	50.0%4–4	62.5%5–3	75.0%3–1
Serie A	52.8%19–17	41.7%5–7	88.9%8–1	100.0%4–0
Ligue 1	51.4%19–18	57.1%8–6	60.0%3–2	0.0%0–1
Bundesliga	47.6%10–11	40.0%4–6	75.0%6–2	57.1%4–3

1X2 · edge vs market

League	≤−5	−5…0	0…+5	+5…+10	>+10
All leagues	60.3%105–69	53.7%73–63	49.3%70–72	42.2%38–52	33.0%33–67
Championship	36.8%7–12	66.7%12–6	38.1%8–13	21.4%3–11	16.7%4–20
LaLiga	68.0%17–8	56.3%9–7	61.1%11–7	18.2%2–9	12.5%1–7
Brasileirão	61.9%13–8	50.0%9–9	53.3%8–7	54.5%6–5	33.3%3–6
Eredivisie	57.1%16–12	33.3%2–4	53.8%7–6	40.0%2–3	25.0%3–9
Premier League	64.7%11–6	40.0%6–9	50.0%7–7	57.1%4–3	12.5%1–7
Liga Portugal	52.9%9–8	44.4%4–5	63.6%7–4	80.0%8–2	36.4%4–7
Ligue 1	66.7%10–5	52.9%9–8	45.5%5–6	50.0%2–2	33.3%3–6
Bundesliga	40.0%2–3	75.0%9–3	42.1%8–11	55.6%5–4	0.0%0–1
Serie A	71.4%5–2	42.9%3–4	60.0%3–2	20.0%1–4	60.0%3–2
Süper Lig	85.7%6–1	60.0%3–2	0.0%0–1	50.0%2–2	71.4%5–2
1. Divisjon	77.8%7–2	75.0%3–1	66.7%2–1	40.0%2–3	100.0%3–0
Eliteserien	33.3%1–2	33.3%1–2	50.0%3–3	0.0%0–2	100.0%2–0
Ekstraklasa	100.0%1–0	50.0%2–2	25.0%1–3	0.0%0–1	100.0%1–0
Superligaen	–	50.0%1–1	0.0%0–1	50.0%1–1	–

Over/Under 2.5 · model level

League	<50%	50–60%	60–70%	≥70%
All leagues	–	53.0%282–250	61.9%232–143	66.7%142–71
1. Divisjon	–	73.3%11–4	76.5%26–8	69.0%60–27
Brasileirão	–	49.4%40–41	40.6%13–19	50.0%1–1
Championship	–	48.6%35–37	42.9%12–16	44.4%4–5
Ekstraklasa	–	42.6%23–31	78.1%25–7	55.6%5–4
Eliteserien	–	59.4%19–13	69.4%25–11	76.0%19–6
LaLiga	–	65.9%27–14	70.8%17–7	69.2%9–4
Eredivisie	–	61.1%11–7	82.4%28–6	76.2%16–5
Liga Portugal	–	54.8%17–14	51.6%16–15	40.0%2–3
Superligaen	–	32.4%11–23	52.4%11–10	45.5%5–6
Süper Lig	–	57.9%22–16	58.8%10–7	62.5%5–3
Premier League	–	57.5%23–17	45.0%9–11	0.0%0–1
Serie A	–	54.5%18–15	48.0%12–13	66.7%2–1
Ligue 1	–	62.1%18–11	63.6%14–8	50.0%3–3
Bundesliga	–	50.0%7–7	73.7%14–5	84.6%11–2

Over/Under 2.5 · 1X2 disagreement (above threshold)

League	<5	5–10	≥10
All leagues	65.7%46–24	71.9%23–9	83.7%36–7
Eredivisie	73.3%11–4	75.0%3–1	94.4%17–1
Bundesliga	71.4%10–4	100.0%6–0	–
LaLiga	55.6%5–4	60.0%3–2	100.0%5–0
1. Divisjon	100.0%5–0	60.0%6–4	66.7%2–1
Ligue 1	60.0%3–2	–	50.0%2–2
Liga Portugal	50.0%1–1	60.0%3–2	100.0%2–0
Championship	33.3%1–2	100.0%1–0	75.0%3–1
Premier League	0.0%0–3	100.0%1–0	100.0%2–0
Eliteserien	80.0%4–1	–	100.0%1–0

Both teams to score · model level

League	<50%	50–60%	60–70%	≥70%
All leagues	–	53.1%340–300	60.6%223–145	68.8%77–35
1. Divisjon	–	68.8%11–5	71.7%38–15	67.2%45–22
Brasileirão	–	50.5%48–47	50.0%10–10	–
Championship	–	51.9%42–39	52.0%13–12	66.7%2–1
Ekstraklasa	–	58.5%38–27	60.7%17–11	100.0%2–0
Eliteserien	–	55.6%20–16	60.4%29–19	88.9%8–1
LaLiga	–	53.1%26–23	56.0%14–11	50.0%2–2
Eredivisie	–	75.0%15–5	70.2%33–14	100.0%6–0
Liga Portugal	–	53.1%26–23	68.8%11–5	0.0%0–2
Superligaen	–	57.9%22–16	38.1%8–13	57.1%4–3
Süper Lig	–	43.6%17–22	47.6%10–11	66.7%2–1
Premier League	–	56.3%27–21	38.5%5–8	–
Serie A	–	52.1%25–23	61.5%8–5	–
Ligue 1	–	42.9%15–20	68.4%13–6	33.3%1–2
Bundesliga	–	38.1%8–13	73.7%14–5	83.3%5–1

Both teams to score · edge vs market

League	≤−5	−5…0	0…+5	+5…+10	>+10
All leagues	60.0%30–20	58.8%80–56	62.7%131–78	46.5%73–84	40.0%30–45
Championship	100.0%3–0	58.8%10–7	63.0%17–10	33.3%12–24	38.5%5–8
LaLiga	100.0%1–0	62.5%5–3	56.3%18–14	48.0%12–13	45.5%5–6
Brasileirão	66.7%2–1	43.8%7–9	48.0%12–13	45.0%9–11	57.1%4–3
Eredivisie	50.0%3–3	70.6%12–5	86.2%25–4	80.0%8–2	100.0%2–0
Premier League	100.0%2–0	57.1%12–9	52.4%11–10	43.8%7–9	0.0%0–1
Liga Portugal	100.0%1–0	85.7%6–1	50.0%12–12	53.3%8–7	50.0%5–5
Ligue 1	66.7%4–2	64.3%9–5	69.2%9–4	33.3%4–8	27.3%3–8
Bundesliga	50.0%5–5	38.5%5–8	86.7%13–2	50.0%3–3	50.0%1–1
Serie A	66.7%2–1	33.3%2–4	57.1%4–3	83.3%5–1	28.6%2–5
Süper Lig	40.0%2–3	25.0%1–3	50.0%2–2	25.0%1–3	28.6%2–5
1. Divisjon	66.7%2–1	85.7%6–1	57.1%4–3	50.0%1–1	50.0%1–1
Eliteserien	100.0%2–0	75.0%3–1	75.0%3–1	0.0%0–1	0.0%0–1
Ekstraklasa	0.0%0–4	100.0%2–0	100.0%1–0	66.7%2–1	0.0%0–1

Both teams to score · 1X2 disagreement (above threshold)

League	<5	5–10	≥10
All leagues	62.8%54–32	64.0%32–18	78.3%47–13
Eredivisie	75.0%12–4	71.4%5–2	87.0%20–3
Bundesliga	77.8%14–4	71.4%5–2	–
LaLiga	50.0%5–5	50.0%3–3	66.7%4–2
1. Divisjon	50.0%3–3	75.0%9–3	66.7%2–1
Championship	42.9%3–4	100.0%1–0	88.9%8–1
Ligue 1	62.5%5–3	66.7%2–1	80.0%4–1
Brasileirão	75.0%3–1	100.0%2–0	60.0%3–2
Premier League	33.3%1–2	25.0%1–3	66.7%2–1
Eliteserien	66.7%4–2	50.0%1–1	100.0%2–0
Süper Lig	0.0%0–2	0.0%0–2	33.3%1–2

How to read it (as of 2026-09-12): in 1X2 the hit rate falls as the edge grows, so a model drifting away from the market is not being proven right; in the goal markets the level works: Over 2.5 and both-teams-to-score selections above 60% hit, no edge required. Edge tables only include matches with recorded odds (no bookmaker price is stored for Over/Under). Today's picks are built on this report. Disagreement tables: absolute gap (percentage points) between the model's 1X2 pick and the vig-free market, only for Over 2.5 ≥ 65% and BTTS Yes ≥ 60% picks. As of 13 Sept, goal markets hit markedly better at ≥10 points of disagreement (Over 13/13, BTTS 22/24), but most of the sample is Eredivisie; it joins the rule once non-Eredivisie data accumulates.

Showcase record

rule E · frozen picks · last 30 days

The showcase pick filters the raw engine output through one rule: goal market first (Over 2.5 model at 65%+ or both-teams-to-score at 60%+), otherwise 1X2 only when the model sits less than 5 points above the market, never 1X2 in weak leagues; everything else is 'no pick'. Picks freeze 3 hours before kick-off and are scored here as frozen.

Market	n	Hit rate
All picks	115	0.6173913043478261
1X2	42	0.5714285714285714
Over/Under 2.5	21	0.6666666666666666
Both teams to score	52	0.6346153846153846
No pick	85	0.425

League	n	Hit rate	No pick
Eredivisie	20	0.7	0
LaLiga	14	0.7857142857142857	9
Serie A	13	0.5384615384615384	5
Bundesliga	12	0.6666666666666666	1
Süper Lig	11	0.5454545454545454	6
Championship	10	0.4	2
1. Divisjon	8	0.625	0
Ligue 1	5	1	12
Brasileirão	4	0.75	23
Ekstraklasa	4	0.75	0
Premier League	4	0.25	14
Eliteserien	4	0.5	1
Liga Portugal	3	0.3333333333333333	10
Superligaen	3	0.3333333333333333	2

Rule from the 13 Sept 2026 backtest (231 matches with odds): combined rule 113/183 = 62%, coverage 79%; 1X2 alone 43%, bookmaker favourite 47%. Moves to the showcase after one week of live observation.

Weekly progress

computed every Monday · covered leagues · official version

Week	1X2			Over/Under 2.5			Both teams to score
	n	Hit rate	Log-loss	n	Hit rate	Log-loss	n	Hit rate	Log-loss
2026-W42	133	54.9%	0.968	133	54.1%	0.682	133	54.1%	0.692
2026-W41	97	61.9%	0.924	97	61.9%	0.664	97	57.7%	0.673
2026-W38	81	48.1%	1.028	81	61.7%	0.689	81	55.6%	0.688
2026-W37	67	35.8%	1.044	67	49.3%	0.714	67	49.3%	0.711
2026-W36	85	49.4%	1.052	85	64.7%	0.658	85	60.0%	0.647
2026-W35	68	52.9%	0.985	68	63.2%	0.653	68	61.8%	0.655
2026-W34	50	44.0%	1.105	50	62.0%	0.659	50	46.0%	0.692
2026-W33	27	37.0%	1.169	27	66.7%	0.614	27	74.1%	0.589
2026-W32	13	46.2%	1.001	13	38.5%	0.710	13	61.5%	0.712
2026-W31	3	33.3%	1.082	3	33.3%	0.670	3	33.3%	0.723
2026-W30	3	33.3%	0.973	3	66.7%	0.691	3	33.3%	0.683
Last 11 weeks	627	50.1%	1.011	627	59.0%	0.673	627	56.1%	0.677

Each row is the matches played and settled in that ISO week; the numbers never change afterwards. Lower log-loss is better (1X2 random = 1.099; binary markets = 0.693). ECE is the mean gap between predicted and observed probability over 10 bins. A single week is noisy; decisions are taken on the rolling window. On the 477 matches with a closing price: model 48.0%, bookmaker favourite 51.8%, ROI -13.5%. Display calibration (curve fitted one week earlier): confidence Brier raw 0.2351 → calibrated 0.2420.

Alerts from the latest review (2026-W42):

1X2 ECE 0.081 > 0.05
Display calibration made 1X2 confidence Brier worse for 3 consecutive weeks

Weekly lessons (shadow)

graded from settled matches · 2026-W42 · exploratory

Every settled match gets a grade card per market: did the model's selection come in, what the bookmaker's pre-match favourite did, who was right when they disagreed, and how the pre-match assessment status fared. The week's cards are summed into cells below. This changes nothing yet: picks, model and assessment rules stay as they are until at least four weeks of shadow record justify a versioned change.

6 cards from 2 matches · covered leagues · Assessment status known pre-match for 0 cards; 6 cards had no pre-match snapshot and are excluded from the status rows.

Cell	n	Hit	95%	Model avg
1X2	2	100%	34–100%	52%
1X2 · Model and bookmaker agreed	1	100%	21–100%	44%
1X2 · Model level <50%	1	100%	21–100%	44%
1X2 · Model level 50–60%	1	100%	21–100%	59%
Over/Under 2.5	2	50%	9–91%	60%
Over/Under 2.5 · Model level 50–60%	1	0%	0–79%	52%
Over/Under 2.5 · Model level 60–70%	1	100%	21–100%	67%
Both teams to score	2	0%	0–66%	58%
Both teams to score · Model level 50–60%	1	0%	0–79%	53%
Both teams to score · Model level 60–70%	1	0%	0–79%	64%

Lessons this week

No cell cleared the 20-match bar with a clear signal this week. That is normal; lessons need volume.

Cells with fewer than 20 cards are shown in grey and never produce a lesson. Status rows use only snapshots written before kick-off; nothing is graded retrospectively with today's rules. Lessons are exploratory and feed a versioned rule change only after a multi-week shadow record.

Computed Tue, Oct 20, 09:36. Recomputed hourly. How the numbers are defined: methodology. · Open data