Every Pre-configured fixture prediction in the app is built from a statistical model trained on real match data — goals, shots, corners, cards, form, head-to-head results, and how each team performs relative to their league. The model processes all of this and produces a full set of match probabilities and market lines before a ball has been kicked. This enables users to have complete access to a suite of prediction data without the need for a constant internet connection (which is especially important for anyone who has tried using the internet at a packed football stadium).
However, this approach has a clear limitation… It uses all the available data, not necessarily the data isolated to the starting 11 of each team. Team news changes the picture as a team may be missing key players through injury, have a new striker signed from overseas, or a manager may rotating ahead of an important cup tie. These details are important, as while general fluctuations in performance deviate around the mean, trimming the noise can aid with accuracy.
The Custom Prediction service addresses this directly. Rather than relying on the model's read of each team as a whole, users are able to submit their own expected starting XI for both sides before the match. When submitted, the model then re-runs with those specific players enhanced in the weighting. This produces a prediction that reflects the matchday reality the user anticipates. The data fields output are the same as a user gets in the pre-configured version, generated by the same engine, but is also informed by the line-up.
The result is a prediction that rewards preparation. Users who follow team news closely, track injury updates, and understand how a manager tends to set up for particular opponents can turn that knowledge into a more precise output than a purely model-driven prediction can provide.
CUSTOMISED PREDICTION GENERATION
Custom Generation Features
Select any upcoming fixture and submit your own starting XIs to get a prediction based on customised line-ups. These can be ones used for the actual games, or your own anticipated ones ahead of time to try and secure better odds.
11 slots per team — assign position (GK locked, outfield choose Defender/Midfielder/Forward) then select player.
Player dropdown is sectioned: position-matched squad players first, then all other team players below. This means you can play anyone in any position and the model accounts for it.
"Search for player…" searches the entire database, not just the team's squad. This means users can find players from other clubs or new signings. Users can also bild hypothetical squads to see how different potential signings might influence outcomes.
"Player not found / Unknown…" lets you type any name freely for players with no data at all. The database won’t be able to account or all players, so users are able to add Unknown players to the line-ups and assign an Influence rating (Neutral / Low / High / Very High) so the model can weight their impact despite having no stats.
Tactical pitch review renders both XIs on a pitch before submitting with icons positioned by role in positional lanes.
Interactive Player Icons: Tap a chip to change that player; drag onto another chip to swap their positions; drag to an empty lane to reassign the position.
Requests are saved locally on submit and a copy of the line-up added to each returned prediction so a user can recall the team selection used.
Randomise XI auto-fills a 4-4-2 from squad data as a quick starting point.
Returned predictions utilise the same notebook and bet text generation features used by the pre-configured predictions, allowing the user to mix custom and pre-configured fixtured in bet requests.