Odds: Bet365 pre-match prices. Live data: ESPN public API. Model: Poisson in-play estimator with xG time-decay and game-state adjustments. Flat stake: £20 per acca.

Betting Portfolio & P&L

All tracked accas with results. Flat £20 stake per bet.

How We Found This Bet

A complete walkthrough of the research process, from scanning fixtures to building a value accumulator on World Cup 2026 playoff night (March 26, 2026).

Step 1: Fixture Scanning

What we did

Searched for all football matches today. Identified 8 UEFA World Cup 2026 playoff semi-finals plus 2 inter-confederation playoffs as the marquee fixtures.

Why playoffs are good for accas

• One-leg knockout format creates pressure — favourites don't always perform
• High-stakes games produce predictable patterns (defensive starts, late goals)
• Bookmakers price based on group stage form, but playoff dynamics differ
• Multiple concurrent matches = more legs to combine

Data sources used

• ESPN API — free, public, real-time scores + goalscorer data
• Covers.com — expert picks and consensus odds
• Wincomparator — odds comparison across bookmakers
• Racing Post — form analysis and tipster consensus

Step 2: Odds Collection

Match odds gathered (decimal)

Primary odds sourced from Bet365, cross-referenced on Wincomparator. Bet365 was chosen as the primary book due to wide market availability (match results, BTTS, scorer markets, totals) and competitive pricing on international football.

Step 3: Sentiment & Form Analysis

Method

For each match, we aggregated sentiment from multiple sources:

• Expert tips — Covers, Racing Post, CBS Sports, Goal.com predictions
• Historical form — last 5/10 match results, home/away splits
• Head-to-head — playoff pedigree, knockout tournament history
• Team news — injuries, suspensions, squad announcements
• Key stats — goals per game, clean sheet rates, BTTS rates

Key findings

Step 4: Value Identification

The formula

Edge = True Probability - Implied Probability

Where Implied Probability = 1 / Decimal Odds

A positive edge means the bet is value — the bookmaker is offering better odds than the true chance.

Our edge map

5 of 7 legs have a positive edge. The two slight negatives (Wales, Poland) are still close to fair value and are needed as "banker legs" to give the acca structural reliability.

Step 5: Acca Construction

Design principles

• Mix bet types — match results + BTTS + scorer markets reduce correlation
• 2-3 banker legs — short-odds selections (Italy, Poland, Denmark) provide structural stability
• 2-3 value legs — longer-odds selections (BTTS, scorer markets) where we have the biggest edges
• Target total odds — user wanted ~150/1 for a £20 stake = ~£3,000 return

Odds multiplication

2.50 × 1.78 × 1.74 × 1.82 × 1.98 × 1.98 × 2.70 = ~149.16

£20 × 149.16 = £2,983

Risk assessment

• LOW RISK Italy Win & U2.5 — Italy dominant, N. Ireland park the bus
• LOW RISK BTTS Ukraine/Sweden — 73% true probability, best value leg
• MEDIUM Wales Win, Poland Win, Czechia Win — home favourites, should deliver
• HIGH RISK Scorer markets — individual scorers are inherently volatile

Step 6: Live Tracking Setup

Tech stack

• Odds source: Bet365 — primary bookmaker for odds research, market scanning, and bet placement
• Live data: ESPN public API (site.api.espn.com) — free, no auth, real-time scores + goalscorers
• Probability engine: Poisson-based in-play model running client-side in JavaScript
• Hosting: Nginx container on Docker port 8070
• SSL: Caddy reverse proxy with automatic Let's Encrypt certificates
• Domain: bet.ictedgefund.com via A record → server IP

Probability model details

• Team win: Pre-match expected goals (xG) per team, Poisson-distributed over remaining minutes. Game-state adjusted (losing team gets +25% attack rate, winning team -15%)
• Under X goals: Combined xG rate projected over remaining time, summing Poisson PMF for allowed remaining goals
• BTTS: Conditional probability — if one team has scored, only need P(other team scores ≥ 1 in remaining time)
• Scorer: Pre-match anytime scorer probability decayed exponentially by elapsed time. P(scores) = 1 - (1-p90)^(remaining/90)

Methodology: Poisson Model Explained

What is the Poisson distribution?

Goals in football follow a Poisson process — they're rare, independent events occurring at a roughly constant rate. If a team is expected to score 1.5 goals per 90 minutes (their xG), we can calculate the probability of them scoring exactly 0, 1, 2, 3... goals.

P(k goals) = (lambda^k * e^-lambda) / k!

In-play adaptation

As the match progresses, we scale the xG rate by the fraction of time remaining:

remaining_xG = pre_match_xG * (remaining_minutes / 90)

We then enumerate all possible remaining scorelines (up to 5 additional goals per team) and sum the joint probabilities to get win/draw/loss chances.

Game state adjustments

Teams trailing push forward (higher xG), teams leading sit deeper (lower xG). We apply +25% / -15% multipliers based on current score difference. This is a simplified version of the Dixon-Coles (1997) correction.

Limitations

• Pre-match xG estimates are manually set, not from a statistical model
• Red cards, injuries, and tactical shifts aren't captured
• Scorer probabilities assume uniform scoring rate across the match
• No correlation adjustment between legs in the same match (legs 2 & 7 share Wales/Bosnia)

Glossary