FAQ

Sollert is a European football betting analytics platform. We build machine learning models that produce calibrated probability predictions for player prop markets — yellow cards, fouls, tackles, shots, goals, and assists. Every prediction is compared against live bookmaker odds to surface value bets with an edge score, and we provide half-Kelly staking recommendations to size bets relative to your bankroll.

We currently cover 5 leagues: Premier League, La Liga, Bundesliga, Serie A, and Ligue 1. Markets include yellow card probability, expected fouls, expected tackles, expected shots, expected goals, and expected assists.

Player prop markets — yellow cards, fouls, tackles, shots — are less scrutinised by bookmakers than match outcomes. Bookmakers spend most of their effort sharpening match result lines; player props, especially cards and fouls, are often priced with less rigour. This gives model-based tools more room to find genuine edges. It also means you're less likely to have your bookmaker account limited compared to value betting on match outcomes.

We train separate models for each market and position group (defenders, midfielders, forwards), then combine their outputs into a calibrated ensemble. All probabilities are post-processed to ensure calibration — when we say 65%, the historical frequency is close to 65%. We track accuracy metrics for every model version and retrain when calibration degrades.

Price gap is the difference between Sollert's model probability and bet365's implied probability for the same outcome, expressed in percentage points (pp). If our model says a player has a 38% chance to score and bet365's price implies 32%, the price gap is +6pp. Price gap is descriptive — it's just what shows up when you put two probabilities side by side.

Anything above 12 is solid — that's the band where conviction is at MODERATE or higher and the evidence base is decent. The 16–20 band is rare in production (around 1 in 500 predictions on a typical day) — those are the predictions with everything aligned. Below 8 is the “we'd pass” band — the score is honestly telling you we don't have a strong prediction here.

They measure different things. Probability is “how likely is the outcome” (e.g. 38% chance Salah scores). Sollert Score is “how solid is our forecast” — it asks whether the conviction tier, evidence base, and model decisiveness all line up. A 75% probability with weak evidence and counter-signals can have a low Sollert Score; a 22% probability with HIGH conviction and 50+ historical starts can have a high one. The score is meta-confidence in the prediction, not the probability of the outcome.

Three things can move the score between scoring runs: (1) Lineup confirmation — pre-lineup scores can change once the team-sheet is announced, because counter-signals (lineup uncertainty) clear and conviction may move up. (2) New evidence — a player who plays a fresh fixture moves into a higher sample-size bucket once their starts cross 5 / 15 / 30. (3) Model retrain — periodic retrains can shift the calibrated probability for a fixture, which in turn shifts the score's probability input. The score does not update mid-fixture; it freezes at scoring time and stays frozen through settlement.

If the player has very little history (fewer than 5 starts), the sample-size input drops to 0.0, capping the score around 12–14 even with HIGH conviction. This is intentional — a strong signal on a thin evidence base shouldn't get the same visual confidence as a strong signal backed by 50 historical fixtures.

Kelly staking is a mathematically optimal bet sizing formula. Given the price gap and the odds, it tells you what fraction of your bankroll to stake. Sollert uses half-Kelly (more conservative) to account for model uncertainty. You enter your bankroll once and the app scales all stake recommendations to it.

The Builder Lab lets you combine multiple predictions into a single multi-leg bet. Available on Edge and Sharp. It shows you the naïve combined probability (if legs were independent) and our adjusted probability (accounting for correlations between legs). This tells you whether the bet builder price offered by your bookmaker is fair value.

No. No predictions are guaranteed. Betting always carries the risk of losing money. Sollert surfaces price gaps — situations where our probability differs from bet365's implied probability — but individual bets will lose. Only ever bet what you can afford to lose.

Scout (free): Today's feed, match fixtures, basic cheat sheets.

Edge (£19/mo): Full price-gap analytics, best odds, half-Kelly staking, player profiles and trends, hit rate tracking, the Insights tab, and CSV data export. Includes a 7-day free trial.

You can upgrade or cancel anytime from your Account page. Cancellations take effect at the end of your current billing period — you won't be charged again.

Scout is free forever. Edge includes a 7-day free trial — cancel anytime, no questions asked. After 7 days it converts to £19/mo. Sharp does not offer a trial.

Payments are processed by Stripe. We accept all major credit and debit cards. We do not store card details — Stripe handles all payment data.

GamCare offers free, confidential support 24/7. Call 0808 8020 133 or visit gamcare.org.uk. You can also visit our Responsible Gambling page for more resources.

Every UK-licensed bookmaker is required by law to offer deposit limits, loss limits, and session time limits. You can usually find these in your account settings under 'Responsible Gambling' or 'Safe Gambling'. Set them before you start betting.

Yes. GamStop is a free national self-exclusion scheme. Registering excludes you from all UK-licensed gambling sites for a minimum of 6 months. Visit gamstop.co.uk.