# How I pick European football bets | EuroFootballPicks

> From a shortlist of about 200 matches in 26 European leagues, through a goals and xG model and team news, to a few picks posted before kick-off.

- Source: https://eurofootballpicks.com/how-i-pick.html
- Updated: 2026-11-19
- Author: LP75 (Libor Pavlíček), EuroFootballPicks
- Markdown version of the page for AI assistants and agents. No bookmaker ads or affiliate links. For adults 18+ only.

---

1. **A weekly shortlist.** I go through roughly two hundred matches from 26 European leagues: table position, points and goals of both teams, odds and the chances from my model.
2. **The model.** It works with goals and expected goals (xG) of both teams this season, home and away separately. It tells me where the favourite is genuinely strong and where the odds only look attractive.
3. **Team news.** Before posting I check injuries and suspensions in two independent sources. A missing key player can be enough to drop a pick.
4. **Only the strongest.** On match days I post a few picks I trust most, sometimes just one, sometimes none.
5. **Posted before kick-off, settled after.** Every pick goes out with a timestamp on Telegram and into the [record](https://eurofootballpicks.com/record.html), where it stays, whatever the result.

No pick is a guarantee. Football is a low-scoring game and favourites lose. That is why I show every result, including the bad days.

## How the model works

1. **Attack and defence strength.** For every match of the season I take each team's “performance”: 60% expected goals (xG) and 40% actual goals. It is compared with the league average, separately for home and away sides, giving an attack and a defence index (1.00 = league average).
2. **Last season as an anchor.** Early in the season there are few matches, so last season's strength is mixed in with the weight of ten matches. Promoted teams start as a weaker side (attack 0.85, defence 1.15). The weight of last season fades with every match played.
3. **Expected goals for the match.** League average × one team's attack × the opponent's defence, with a small home adjustment (×1.03).
4. **Probabilities.** From the expected goals, a Poisson distribution gives the chance of every score up to 10–10 (with the Dixon–Coles adjustment for low scores, which makes draws more accurate). Summing them gives home/draw/away, over 2.5 goals and both teams to score.
5. **What the model does not know:** injuries, line-ups, motivation, a new coach or the weather. I check those myself — absences from two independent sources — and treat the model as a guide, not an oracle.

The settings were chosen with a backtest over two seasons of 14 leagues (against results and closing odds). The test results are in the table below.

## How accurate is the model?

A backtest on 5,351 matches from 14 European leagues (August 2025 – November 2026). For every match the model only knew the data available before kick-off, exactly as it does live. The table shows: when the model gave an outcome (home win, draw, away win) a certain chance, how often did it actually happen?

| Model chance | Outcomes | Model expected | Actually happened | Bookmakers expected |
| --- | --- | --- | --- | --- |
| **0–10%** | 183 | 8% | **6%** | 9% |
| **10–20%** | 1,335 | 16% | **15%** | 17% |
| **20–30%** | 6,669 | 26% | **26%** | 26% |
| **30–40%** | 3,739 | 34% | **33%** | 34% |
| **40–50%** | 2,299 | 45% | **46%** | 45% |
| **50–60%** | 1,108 | 54% | **55%** | 54% |
| **60–70%** | 499 | 64% | **70%** | 64% |
| **70–80%** | 177 | 74% | **78%** | 74% |
| **80–100%** | 44 | 83% | **75%** | 80% |

The model is well calibrated: when it says 50–60%, the outcome happens about 55% of the time. With favourites at 60–80% it is on the cautious side — they win a few points more often than it expects.

**An accurate estimate is not the same as profit.** Bookmakers estimate just as well, and their odds include a margin. When we let the model bet retrospectively, it would have lost money — more in [why you should never pay for tips](https://eurofootballpicks.com/paid-tips.html). Data: [JSON (CC BY 4.0)](https://eurofootballpicks.com/data/model-calibration.json).
