Connect
Use SGPLab in Claude
Ask Claude about prop hit rates, correlations and lines, and get answers computed from the same data as this site.
· Set it up
Server URL: https://sgplab.io/api/mcp
Claude (web and desktop): open Settings, then Connectors, choose to add a custom connector, and paste the URL. No account or API key is needed.
Claude Code: claude mcp add --transport http sgplab https://sgplab.io/api/mcp
· Try asking
- Which NBA players have gone over their points line most often this season?
- What range does the betting market imply for tonight's top NBA scorers?
- How often do NBA players go over their points line while their team fails to cover?
- Where do NFL final margins land most often, and what is a half-point worth around 3 and 7?
- Which sportsbooks charge the lowest hold on NBA player points?
- In NFL games with 15+ mph wind, how have receiving yards overs done at DraftKings?
· What it can and can’t do
The connector covers the NFL, NBA, NHL, MLB, WNBA and college football. Historical results are graded against real sportsbook closing lines, and answers link back to the matching page here so you can check the full table.
It only reads research data. It can’t place bets, see a sportsbook account, or build a betslip. Results describe what happened and are not picks; see the disclaimer. SGPLab is intended for adults who can legally bet where they live.
Current prices at individual sportsbooks and the views that need a Pro subscription aren’t available through the connector. Upcoming lines are consensus across books.
· Tools
| Tool | What it does |
|---|---|
slate | Upcoming games for a league that sportsbooks have priced, with start time and how many markets are posted. |
projections | Projections for players on the upcoming slate, read off the betting market: for each player and stat, a distribution fitted to every priced rung of the alternate-line ladder (consensus across books). |
lines | Current consensus lines for the upcoming slate: the median price across sportsbooks (pick'em sites excluded), only where at least two books post it. |
hit_rates | How often each player went over (or under) their prop line, graded game by game against a real book's closing line, with the record, hit rate, and the flat 1-unit profit at the actual prices. |
combo_hit_rates | How often each player cleared EVERY one of several prop lines in the same game (or stayed under all of them), graded against the consensus closing line. |
player_card | One player's recent games with the result against each night's prop line, per stat, plus a record for every stat that book priced. |
player_correlation | How often a player hit a combo where EACH leg has its own direction, e.g. points over + rebounds under, or points over + his team fails to cover. |
team_correlation | How often each TEAM hit a combo of segment × market legs, each with its own direction. |
league_correlations | Realized correlation (phi) of common SGP combos across a whole league. |
prop_splits | How overs and unders settled across the whole league for one stat, split by the size of the line (stars vs role players), or by price for anytime props (HR, goals, anytime TD). |
ladder | What riding a whole alternate-line ladder would have paid: the book's main line plus each harder rung above it, staked 0.5 / 0.25 / 0.2 / 0.15 / 0.05 units by distance from the main line, graded against what happened. |
volatility | Which players swing most (or least) from game to game for a stat, after removing the swings expected from their volume. |
team_duos | How often a team's two top-lined players (by the book's line that night) both went over, or both under. |
key_numbers | Where final margins and totals land, and how often games lined at a number pushed. |
hold | The average hold (vig) each sportsbook charges by market, from closing two-way prices. |
game_flow | Rotation effects reconstructed from play-by-play. |
team_predictability | How hard a team's player props are to price, and which games tonight are most disrupted. |
minutes_overlap | How much two teammates share the floor relative to what their minutes alone predict (NBA/WNBA). 1.0 = independent, above 1 = subbed together, below 1 = staggered. |
prop_backtest | Backtest a prop bet (stat + over/under + book's closing main line) under filters such as the team's spread, the game total, home/away, the line or price, rest days, season, team/opponent, position, or the line's rank on the team (1 = the market's top option). |
describe_data | Lists the raw research tables available to query_data (box scores, play-by-play, schedules and scores, quarter/inning scores, rosters, injuries, consensus closing player-prop lines, alt-line ladders and game lines), or the columns of one table. |
query_data | Run one read-only DuckDB SELECT over the raw research tables (see describe_data). |
· If something looks wrong
- No games or props for a league: that league may be out of season, or books may not have posted props yet. Football props can appear days before kickoff, others on game day.
- A player or stat isn’t found: the tool replies with what it can match, and Claude usually retries. Try the full name, or ask which stats that league has.
- A “too many requests” error: the limit is 120 calls per 10 minutes. Wait a few minutes and try again.
- A first answer is slow: a large league file may be loading, which can take several seconds.
· Privacy and support
We log each tool call, but not your conversation or your IP address in that log. Details are in the privacy policy. Questions or problems: [email protected].