Lyon
Crystal PalaceMost likelyLyon win · 44%
Best price2.21 Marathonbet
Market-anchored: the market's vig-free price (1 sharp book) + model lean (max ±3 pp)
These are calibrated probabilities. See how closely they've matched real outcomes
Prediction as of 15 Oct 2026
Lyon and Crystal Palace face off in a compelling UEFA Europa League matchup, with statistical models suggesting a tight contest. The probabilities reflect a finely balanced affair, with Lyon assigned a 44% chance of victory, while the draw and a Crystal Palace win are both rated at 28%. Given the proximity of these figures, this fixture sits on the knife-edge of a toss-up, with the model indicating only a slight lean toward the home side rather than a definitive outcome. The lean toward Lyon is supported by several key metrics. The model highlights Lyon’s attacking expected goals (xG) form at home as a significant factor, while simultaneously pointing to Crystal Palace's defensive xG form on the road as an element that helps differentiate the two sides. Furthermore, the model’s Pi-rating—a comprehensive measure of team strength—provides additional support for the home side's marginal advantage. It is worth noting that the model's stance on Lyon is broadly consistent with current market sentiment. Both clubs enter the fixture with relatively steady recent form. Lyon’s last five matches have yielded three wins, a draw, and a loss (W, D, W, W, L), while Crystal Palace has returned three wins, a draw, and a loss in their equivalent span (W, L, W, D, W). As these two sides have no history of previous meetings, there is little precedent to rely on, placing more emphasis on current tactical output. Regarding availability, no injuries have been reported for either squad. However, it is important to note that confirmed lineups have not yet been released. Consequently, the absence of reported injuries should not be interpreted as an indication that either side will field their first-choice XI, as squad rotation or the resting of key players remains a possibility before kick-off.
The AI writes the prose. It never produces or adjusts a probability — those come only from the model.
| Bookmaker | Lyon | Drawpays 1% more than the typical book | Crystal Palacepays 6% more than the typical book |
|---|---|---|---|
| Bet365 | 2.20 | 3.70 | 3.10 |
| Betano | 2.18 | 3.70 | 3.15 |
| Betfair | 2.20 | 3.60 | 3.00 |
| Marathonbet | 2.21best | 3.72best | 3.22best |
Bookmaker names shown for reference only. No affiliate links.
Two ways to size a hypothetical stake on the model's call — Lyon at 2.21 (Marathonbet).
Stays in your browser — never sent to us.
Flat 1% of bankroll — the same rule as our public AI bankroll simulation.
Quarter-Kelly at the best listed price, with a hard 5%-of-bankroll cap.
Hypothetical sizing in currency-neutral units, not betting advice. Predictions are probabilities, not guarantees. 18+.
↑ stronger · = even · ↓ weaker opponent
↑ stronger · = even · ↓ weaker opponent
No reported absences.
No reported absences.
No previous meetings on record.
Recent form — context only, not a tip.
No line movement recorded yet
No line movement recorded yet
No line movement recorded yet
How the model sees this match — the drivers and, after full-time, how the prediction scored.
This UEFA Europa League fixture between Lyon and Crystal Palace presents as a tight contest, with the model assigning Lyon a 44% probability of victory, while the draw and a Crystal Palace win each sit at 28%. The model’s slight lean toward Lyon is primarily driven by the side's attacking expected goals (xG) form at home, supported by a favorable Pi-rating and Crystal Palace’s defensive xG form away from home. Despite these factors, the overall distribution suggests an open match, leaving the outcome far from certain. The model’s assessment of Lyon remains broadly consistent with current market sentiment, and with no reported injuries for either side, the tactical landscape remains subject to potential squad rotation once official lineups are confirmed.
The AI writes the prose. It never produces or adjusts a probability — those come only from the model.
How we make these predictionsThe driver chips attribute the gradient-boosted component of an ensemble whose headline is additionally anchored to the market where a consensus exists — they explain that part of the number, not all of it.