The BTP model is a machine. It does not have feelings, a Bet365 account, or opinions about Wigan. It produces probability estimates. This post is data analysis — it is not a tipster service, a recommendations column, or a suggestion that you do anything at all with any of this. The model has never watched a match in its life. It doesn’t even have a life. It has features.
The Model’s Weekend Stack — GW43/GW44: Two Patents + a Heinz
Three separate sets this weekend, across two divisions:
- A Championship Patent — three model-aligned home calls on Saturday’s 3pm card.
- A League One Patent — three picks spread across Saturday and Sunday.
- A combined Heinz — the top six EV selections across both leagues, regardless of which way the model points. A wider, braver net.
WHY THE TWO-PATENT-PLUS-HEINZ STRUCTURE
The Patents are the clean, model-aligned picks. The Heinz takes a wider net — including three selections that are NOT the model’s top call but carry strong market-vs-model EV.
After last week’s Cooked! post-mortem, where the one non-model-aligned “value” pick (Bolton away) was the ugliest loss of the midweek, we’re being explicit about the separation. Patents = model’s first choice. Heinz = the contrarian plays included.
The Bet Structures — Plain English
PATENT — 3 PICKS, 7 BETS
A Patent is 3 selections covering 7 total bets: three singles, three doubles and one treble. Any single selection landing returns something; two or three winning scales up fast.
Minimum to see a return: 1 selection wins. All three landing: full treble plus all three doubles plus all three singles pay.
HEINZ — 6 PICKS, 57 BETS
A Heinz is 6 selections covering 57 total bets: 15 doubles, 20 trebles, 15 four-folds, 6 five-folds, and 1 six-fold accumulator. Named after Heinz’s “57 varieties”.
Minimum to see any return: 2 selections win (one double pays). Singles are not included — unlike a Lucky 63.
HOW EV WORKS — REFRESHER
Bet365 publish decimal odds. Divide 1 by the odds and you get the market’s implied probability. Divide our model’s probability by that implied probability and you get the EV ratio. EV 1.00 = model and market agree. EV 1.50 = model thinks the outcome is 50% more likely than the market does.
EV does not guarantee a correct call on any single fixture. It measures divergence, not certainty. Over a large sample, positive-EV selections should return more than they cost. Over any single weekend, the variance is huge. We are not at “large sample” yet — we are tracking it in public.
Filters Applied (Same Across All Three Sets)
- EV ratio ≥ 1.20 — the model has to be at least 20% more confident than the market.
- Model probability ≥ 25% — no speculative outsiders; every pick has a meaningful chance of landing.
- Draw probability < 40% — avoids fixtures the model itself rates as tossups (no clear direction to back).
- For the Patents only: the pick must be the model’s top call (its highest-probability outcome). “Value” picks against the model’s top call are excluded.
- For the Heinz: picks are ranked by EV regardless of whether they match the model’s top call. The 3 picks that aren’t model top calls are flagged with a 💥 icon so you can see exactly which are the contrarian plays.
🏆 Championship Patent — Three Model-Aligned Home Calls
Saturday 18 April, all three kicking off at 3:00pm. All three are the model’s top call. All three are home picks with model probability above 45% and EV above 1.20 against Bet365’s implied prices.
CHAMPIONSHIP · SAT 3:00PM
Bristol City v Norwich
SELECTED: BRISTOL CITY HOME
✓ Model’s top call
EV RATIO
1.59
odds: 2.60
57.5%
36.1%
Norwich have slumped, Ashton Gate form is strong, and the model rates this as the weekend’s most confident home call. The market agrees Bristol are favourites but prices them at 2.60 — the model sees something closer to 1.75 in pure probability terms.
CHAMPIONSHIP · SAT 3:00PM
Watford v Sheffield Utd
SELECTED: WATFORD HOME
✓ Model’s top call
EV RATIO
1.34
odds: 2.75
45.8%
34.1%
Sheffield United have underperformed their name all season and sit 17th. Watford at home against a drifting visitor is exactly the quadrant the model likes.
CHAMPIONSHIP · SAT 3:00PM
Hull City v Birmingham
SELECTED: HULL HOME
✓ Model’s top call
EV RATIO
1.21
odds: 2.45
46.5%
38.5%
Hull sit 7th on 68 points, one point outside the play-offs, and this is effectively a six-pointer against their own season. Birmingham sit 15th with nothing to play for. The model likes Hull at home in exactly this scenario.
⚽ League One Patent — Three Model-Aligned Picks
Spread across Saturday and Sunday. All three are the model’s top call. One home pick, two away picks — the League One model reads these road trips as underpriced by the market.
LEAGUE ONE · SAT 3:00PM
Reading v Cardiff
SELECTED: READING HOME
✓ Model’s top call · 🔥 Biggest EV of the weekend
EV RATIO
2.47
odds: 4.75
47.2%
19.1%
This is the weekend’s standout EV. Cardiff are priced as clear favourites at 1.60, giving the market just 19.1% on a Reading home win. The model sees Reading at 47.2% — a 28-point gap. Either the model is drastically wrong, or the market is mispricing the gap between 2nd-place Cardiff and a 10th-placed Reading side with good home form.
LEAGUE ONE · SAT 3:00PM
Northampton v Doncaster
SELECTED: DONCASTER AWAY
✓ Model’s top call (narrowly)
EV RATIO
1.78
odds: 1.65
35.0%
55.1%
Model has Doncaster 35.0%, home 34.2%, draw 30.8% — nearly 3-way-even, with Doncaster the top call by the thinnest margin. But the market prices Doncaster as clear favourites at 1.65. The model says this is closer than the price suggests — hence the big EV.
LEAGUE ONE · SAT 11:30AM
Barnsley v Bradford
SELECTED: BRADFORD AWAY
✓ Model’s top call
EV RATIO
1.35
odds: 2.15
56.2%
42.3%
Bradford sit 4th on 71 points, chasing the play-offs. Barnsley are 12th and drifting. The model’s 56.2% confidence on Bradford away is the single largest model probability of the weekend for either team. Oakwell can be awkward for visitors but the form gap is real.
🍳 The Combined Heinz — Top 6 EV Across Both Leagues
Pure top-six by EV ratio across both divisions, ignoring whether the pick is the model’s top call. Three of these selections overlap with the Patents above. The other three are Championship value picks where the model disagrees with the market but isn’t the model’s top call — we flag these with 💥.
THE 6 HEINZ SELECTIONS
| FIXTURE | PICK | ODDS | EV | MODEL | TYPE |
|---|---|---|---|---|---|
| Reading v Cardiff | Reading Home | 4.75 | 2.47 | 47.2% | ✓ Top Call |
| Wrexham v Stoke | Stoke Away | 4.75 | 1.98 | 39.2% | 💥 Value |
| Sheff Wed v Charlton | Sheff Wed Home | 4.75 | 1.80 | 35.5% | 💥 Value |
| Northampton v Doncaster | Doncaster Away | 1.65 | 1.78 | 35.0% | ✓ Top Call |
| Bristol City v Norwich | Bristol City Home | 2.60 | 1.59 | 57.5% | ✓ Top Call |
| Millwall v QPR | QPR Away | 5.00 | 1.52 | 28.4% | 💥 Value |
Coloured rows: ✓ = also in the relevant Patent (model’s top call). Plain rows: 💥 = value pick, not the model’s top call but the EV is strong enough to cross the 1.20 threshold.
HEINZ MATHS — 57 BETS, 6 SELECTIONS
At unit stakes per bet:
- Total stake = 57 units
- Model-probability for all 6 landing (the six-fold): 0.472 × 0.392 × 0.355 × 0.350 × 0.575 × 0.284 = 1.19%
- Six-fold payout if all 6 land: 4.75 × 4.75 × 4.75 × 1.65 × 2.60 × 5.00 = ~2,299 per unit
- Minimum 2 wins to see any return (one double pays). Unlike a Lucky 63, a Heinz has no singles
The typical Heinz return curve is 3-of-6 landing for break-even-ish, 4+ for clear profit. Variance is significant.
The Full EV Landscape — All 24 Priced Fixtures
All 24 fixtures across both matchdays sorted by best EV outcome. 💎 = Patent selection (model’s top call, EV ≥ 1.20). 💥 = Heinz value pick (not model’s top call, EV ≥ 1.20). Grey rows = didn’t qualify.
What’s Not Here (And Why)
NOTABLE EXCLUSIONS
- Mansfield v Luton — EV-rich (draw 1.49, home 1.39) but draw probability 40.4%, just above the 40% filter. The model isn’t confident which direction this goes — we don’t guess.
- Leyton Orient v Rotherham — model has Orient at 55.2% home but EV only 0.98 (market agrees). Nothing to fade when the market prices it the same as we do.
- Blackburn v Coventry — the weekend’s highest-profile fixture and potentially Coventry’s promotion clincher. EV 1.14 on Blackburn home, below our 1.20 threshold. No pick.
- Portsmouth v Leicester — the relegation decider. Home EV 1.18, just under the cut. Narratively irresistible, mathematically just short.
- Bolton v Huddersfield — EV 1.19 on Bolton home, one penny under the cut. Near-miss.
Discipline over drama. That is the whole point of having thresholds.
THE HONEST FRAMING
Three sets, one weekend, two leagues, one honest promise: we publish the follow-up whether these land or get cooked. Last week’s Cooked! post-mortem is the template. The filter did its job in that one — the picks didn’t.
This week we are being explicit about the Patent / Heinz separation. The Patents are the cleanest of the clean — model’s top calls only, EV and probability thresholds passed. The Heinz includes three brave calls (Stoke away at Wrexham, Sheff Wed home against Charlton, QPR away at Millwall) that are not the model’s top call but carry strong EV. You can see the picks; you can see the labels; you can make your own judgments.
Over a large sample, positive-EV picks should return more than they cost. Over a single weekend, variance is enormous. We are tracking this in public because hiding the bad ones is fraud and claiming skill from four hits in a row is self-delusion. Back next week with the results — win, lose, or cooked.
Championship predictions use our Platt-calibrated Logistic Regression (goals_logreg_v1_cal). League One uses the original Random Forest (leagueone_goals_v1). Bet365 decimal odds fetched 16 April 2026 via API-Football; implied probabilities normalised to remove overround. Odds and EV ratios correct at time of writing — prices may have moved since publication. Why different models? · Whole-season model scorecard
