Expert soccer predictions from TipMaster, since 1997: crowd groups, accuracy and model benchmark.
Scored across 7 tools
Four tools (find_matches, list_matches, get_consensus, get_match_prediction) all return match-level data, and three of them return prediction content, so an agent must read carefully to know which to pick (e.g. summary lines vs. full records vs. single-match lookup). Descriptions do draw distinctions (find_matches includes prediction summaries, list_matches omits prediction fields, get_consensus takes id/filters) which keeps it workable, but the boundaries are soft.
All names are snake_case verb_noun (find_matches, get_accuracy, get_consensus, get_expert_picks, get_match_prediction, get_model_comparison, list_matches). The get_/find_/list_ retrieval verbs are used consistently and predictably.
Seven tools is well within the ideal 3-15 range for a focused predictions domain. Each tool maps to a distinct slice of functionality (match discovery, prediction detail, consensus, expert picks, accuracy stats, model benchmark).
The surface covers listing matches, single-match predictions, consensus, expert picks, accuracy statistics, and a model comparison benchmark. Minor gaps exist, such as no competition/team enumeration or group leaderboard tool, but core prediction workflows are covered end to end.
Football (soccer) matches with TipMaster expert predictions, filtered by team, competition (e.g. "Bundesliga", "Premier League", "LaLiga"), mode, status and kickoff window: upcoming (locked), live and finished (played) matches of the last 10 days. Each line has teams, kickoff, status, result, the most likely score and the home/draw/away call of the Top 20%, Top 25% and All players groups.
Inputs: mode, team, limit, toUtc, +3 more.
How often TipMaster players got football (soccer) matches right, per group (Top 20%, Top 25%, All players): majorityCorrect is the share of matches where the group’s most common home/draw/away call was right, outcomeCorrect the share of individual predictions with the right outcome, exactCorrect the share with the exact final score (National only). National covers the last 8 weeks, each week’s groups ranked on the 8 weeks before it; Classic covers this season, ranked by career record since 1997, and career gives each group’s average career points per match since 1997. A group’s rates appear once it has 50 matches with a published split (reason too_few_matches until then). Pass competition (National league name) or game (Classic) to narrow it.
Inputs: game, mode, competition.
Full prediction records: the crowd’s home/draw/away shares, the top 3 exact scores and most likely score (National), the Top 20% / Top 25% / All players groups (bands) and the expert cohort. Pass an id for one match, or mode/status filters for a list. outcome is null while fewer than 20 players have predicted.
Inputs: id, mode, status.
The current TipMaster Classic round: 25 real football (soccer) matches per game (btm = Bundesliga TipMaster, tmi = TipMaster International), each with the expert cohort’s home/draw/away split and majority pick (expertPick), and the whole crowd’s split beside it. The expert cohort is the top quarter, by career points per match, of this round’s players with at least 100 career Classic matches, a record kept since 1997. Expert predictions are published once the round’s entries close (Friday 18:00 Europe/Berlin, releaseUtc); before that released is false, expertPick and expert are null and expertSample is 0. Shares need at least 20 players.
Inputs: game.
The TipMaster expert prediction for one football (soccer) match, found by team names ("Bayern vs Freiburg", one team name also works) or by id. Returns the most likely final score (mostLikelyScore, National only), the top 3 exact scores, and the home/draw/away split of three groups ranked by their own record: Top 20%, Top 25% and All players. Once played it carries the result. Shares are final once status is locked or played (a Classic match with status open still moves); the person asking has nothing to enter. Classic matches have no score (scoreRecorded false), only home/draw/away. A group with fewer than 20 players is withheld (reason). Other close matches are listed in alternatives.
Inputs: id, match.
Experimental football outcome model trained on stored Sportmonks results. Returns a chronological holdout benchmark and comparisons with National expert groups on the same matches. Includes sample sizes, date windows, accuracy, Brier score and log loss. Paired rates stay unavailable below 50 published matches. Does not promise high confidence or future accuracy.
Inputs: no named parameters.
Every published football (soccer) match (teams, kickoff, entry deadline, status, result) without the prediction fields.
Inputs: mode, status.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Play your TipMaster soccer prediction game: competitions, confirmed scores, results and standings.
Football match outcome probabilities recorded before kick-off, and an open record of results.
Predicted scores, win chances and checked sport stats for 29 competitions. Free and licensed tools.
Free football match probabilities (Dixon-Coles), season projections and a public track record.
Add one secure layer between your agents and this server.