A single EdgeAI scouting report is the result of five independent analysis passes stitched together by an AI model. Here's the whole pipeline, in the order it runs.
1. Pulling the raw signals
For any given player prop — say, "Aaron Judge over 1.5 total bases" — EdgeAI first collects five distinct data streams:
Sportsbook line from The Odds API. This is what we're trying to beat.
Historical game logs from the official league API (MLB Stats API, NHL API, etc.). We pull the last 30 games.
Opposing starter or matchup data — for MLB this is the opposing pitcher's ERA, WHIP, K/9, and pitching hand.
Situational factors — ballpark factor for MLB, back-to-backs for NBA/NHL, weather for outdoor games.
Platoon splits — how the batter performs against left vs right-handed pitching, with a 30-at-bat minimum sample threshold.
2. Windowing the stats
Raw averages lie because they ignore how a player is trending right now. EdgeAI computes three overlapping windows:
Baseline (30 games) — the long-term rate. What "normal" looks like for this player.
Edge window (15 games) — the primary signal. This is what gets compared against the sportsbook line.
Recent form (10 games) — the trend indicator. If recent form has shifted 20%+ from baseline, the AI is told to weight it heavier.
Why three windows? Because a player who's averaged 1.6 hits per game for the season but 0.6 in his last 5 games is a very different bet than one who's steady. Layering the windows exposes the trend.
3. Computing signal confidence
Confidence is where a lot of prop-betting tools cheat. EdgeAI computes it explicitly:
How often did the player beat the line over the edge window?
How consistent was the pattern (versus one outlier game inflating the average)?
How large is the sample against the current matchup type (this pitcher's hand, this venue)?
The number you see on the scouting report — 40%, 62%, 78% — is a real fraction backed by the data, not a marketing gauge.
Signal confidence is not a win probability. Even a 90% confidence pick can lose — variance in sports is huge. Confidence measures how strongly the recent data supports the lean, nothing more.
4. Handing the picture to the AI
All of the above is compressed into a structured prompt and sent to a large language model. The prompt includes:
The player, prop type, and current sportsbook line.
Every stat window with its sample size.
Opposing starter block: quality classification, ERA, WHIP, K/9, IP per start.
Platoon matchup: batter hand, pitcher hand, batting average vs each, sample size.
A rule-set the AI must follow: how to weight recency, how park effects interact with pitcher K/9, when to lean UNDER on hits vs OVER on runs.
The AI returns a structured JSON output: recommendation, confidence rating, one-line summary, list of supporting factors, list of risk factors, and a data quality tag.
5. Grading and post-processing
After the AI response comes back, EdgeAI does one more pass:
Cap extreme edge percentages (a ±500% edge on a 0.5-hit prop usually reflects a math artifact, not a real edge).
Mute the color intensity when confidence is under 60% — a huge edge with weak support should not read as a green light.
Flag the pick as Best Value if it's the strongest edge in its game, and High Confidence if statistical support is unusually strong.
Store the raw AI response and the input prompt for audit — every insight can be traced back.
What you see in the app
All of that ends up as a single-page scouting report:
A big OVER or UNDER recommendation.
The Model Edge and Signal Confidence numbers.
A projection-vs-line bar so you can eyeball the gap.
A recent-form sparkline showing which of the last 10 games beat the line.
The matchup panel with opposing pitcher, platoon splits, and ballpark factor.
The AI's plain-language summary, supporting factors, and risk factors.
Frequently asked
How often is the data updated?
Sportsbook lines refresh every 10 minutes as game time approaches. Player game logs refresh every 4 hours. Once a game is within an hour of first pitch, EdgeAI polls the sportsbook line every 10 minutes to catch late line movement.
What if the sportsbook line changes after I unlock an insight?
EdgeAI shows the line as it was when the report was generated. Always verify the current line at your sportsbook before placing the bet — that's the number that actually settles.
Does EdgeAI account for injuries?
Yes. Each report checks the player's injury status. If a player is flagged Day-to-Day, Out, or IL, that's shown in the report header and factored into confidence.
Does EdgeAI take my bets?
No. EdgeAI is a decision-support tool. You see the analysis and place your own wagers with your preferred sportsbook.