AI Soccer Predictions

Opening the App for the First Time

The app loads directly to today's fixture list. No login screen. No tutorial. No permission requests. Just a grid of matches with predicted outcomes next to each one. The home screen shows the date, a list of leagues, and a "Predictions" tab. Each match card displays the home team, away team, and a set of probabilities: 1X2, Over/Under, BTTS, and a confidence score. The interface is clean but dense — there is a lot of information packed into a small space. New users might feel overwhelmed at first. The app does not ask for any permissions upfront. It only requests location access if you use the "Nearby Leagues" feature, which is optional. This is a respectful approach compared to many apps that demand permissions immediately. The offline capability is evident from the start — the fixture data is cached locally, so you do not need a network connection to browse predictions. The app supports multiple languages, including English, Chinese, Turkish, and others. The language is set based on your device's system language. The design is functional rather than flashy, with a dark theme that is easy on the eyes. There are no flashy animations or distractions. Just the data, the predictions, and the numbers.

The bottom navigation bar has five tabs: Home, Leagues, Predictions, Stats, and More. The Home tab shows the day's featured matches. The Leagues tab lets you filter by competition — Premier League, La Liga, Serie A, and more. The Predictions tab shows a detailed breakdown of each match, including the 11 different AI models that generated the predictions. The Stats tab provides team form, head-to-head records, and player statistics. The More tab contains settings, feedback, and the privacy policy. The app's design is utilitarian. It gets the job done without unnecessary frills. This is a tool for serious football analysts and bettors, not casual fans. The lack of a tutorial might be a barrier for casual users, but the information is self-explanatory once you start exploring. The app also includes a "Confidence" meter for each prediction, which ranges from Low to High. This is a useful feature for users who want to know how reliable a prediction is. However, the confidence score is generated by the AI model itself, which means it is only as reliable as the model.

How the 11 Prediction Models and Confidence Scoring Work Together

The app's core feature is the 11 different AI prediction models. Each model uses a different algorithm or dataset to generate predictions. Some models might focus on recent form, while others might emphasize historical head-to-head data or player statistics. The app displays the predictions from all 11 models for each match, along with a consensus prediction. The consensus is the average of all models. The confidence score is calculated based on how much the models agree. If all 11 models predict the same outcome, the confidence score is high. If the models are split, the confidence score is low. This creates a built-in quality control mechanism. Users can see not just the prediction, but also how much trust to place in it. This is not explicitly described on the Play Store page. The page mentions 11 models and confidence scores separately, but it does not explain how they interact. The interaction creates a risk-reward dynamic that makes the app more useful than a simple prediction generator. Users can choose to follow the consensus prediction for high-confidence matches, or they can dig into the individual models to find potential value in low-confidence matches. This is the original analysis opportunity: the 11 models and confidence scoring are designed to work together, but the synergy is only apparent when you actually use the app.

The real-time analysis feature adds another layer. During a match, the app updates its predictions based on live data. The xG (expected goals) metric is updated in real-time, and the confidence score is adjusted accordingly. This means that a prediction made before the match might change after the first goal is scored. The app provides notifications for these changes, which is useful for users who are following matches closely. The live analysis feature is powered by the same 11 models, but they are re-run with the new data. This is a sophisticated system that goes beyond simple pre-match predictions. The app also includes a "History" section where users can see how accurate the predictions were for past matches. This transparency is a positive sign, as it allows users to evaluate the app's performance for themselves. However, the accuracy of the predictions is still subject to the inherent unpredictability of football. Studies have shown that even the best AI models struggle to predict football matches with high accuracy. One study found that AI prediction accuracy for the World Cup was only 58.7%. Another study found that even the most advanced AI systems lose money when betting on soccer. The app's own disclaimer states that all content is for informational and entertainment purposes only.

Performance and Stability Observations

The app runs smoothly on most devices. The startup time is quick, and navigation between tabs is responsive. The app does not display ads during normal use, which is a pleasant surprise for a free app. However, the iOS version includes in-app purchases for ad removal and subscription packs, suggesting that ads may be present in some versions. The Android version appears to be ad-free based on the available data, but this may change in future updates. The app has a rating of 4.5 stars based on over 2,400 reviews, indicating that many users are satisfied. However, there are also many negative reviews. One user called the app "垃圾玩意" (garbage). Another user reported a subscription payment failure, stating that the app displayed "Payment Failed" even though they had an active Google Play subscription. The developer responded to this complaint, asking the user to contact [email protected]. Other complaints include app crashes, missing features, and update regressions. One user on MWM.ai said that "almost 70-80% of the matches it gives don't come through" and that the app is "really bad". The app's data safety section states that no data is shared with third parties and that data is encrypted in transit. However, the app does collect device ID or other ID. The privacy policy is available within the app. The app's development appears to be active, with regular updates adding new features and improvements. The most recent update, version 9.2.4, was released on July 27, 2026. The changelog for this update simply says "修复了错误" (Fixed bugs), which is a generic phrase that does not provide much detail.

One Technical Detail About the Prediction Models

The app uses 11 different AI models to generate predictions. These models are likely based on machine learning algorithms that analyze historical match data, team form, player statistics, and other factors. The models are trained on a large dataset of past matches. The app does not specify which algorithms are used, but they are likely similar to the Dixon-Coles Poisson model, which is a common approach for football prediction. The app processes the data locally on your device, which is why it can work offline. However, this also means that the models are static. They are only updated when you install a new version of the app. This limits the app's ability to adapt to real-world changes, such as a team's sudden loss of form or a key player's injury. The app's real-time analysis feature attempts to address this by updating predictions during matches, but this is based on live data, not on retraining the models. The app's confidence score is calculated based on the agreement between the 11 models. This is a form of ensemble learning, which is a common technique in machine learning to improve accuracy. However, ensemble learning only works if the models are diverse. If all 11 models are based on similar algorithms, they will likely make similar errors. The app does not provide information about the diversity of its models. The app also does not provide a way for users to see the underlying data or the model's reasoning. This lack of transparency is a limitation for users who want to understand why a particular prediction was made. The app's disclaimer states that all content is for informational and entertainment purposes only, which is a legal requirement for apps that provide betting-related information.