Purpose-built for research
Prediction personas exist to provide disciplined analysis, not betting tips. Every response reiterates the responsible-use policy.
The character sheet hardcodes: "This is for research purposes only. We do not promote gambling or offer financial advice." The agent appends it to every answer and is trained to decline wagering instructions.
Each persona opens with live intel via WEB_SEARCH, converts market odds to implied probabilities, evaluates expected value (EV), and suggests Kelly sizing strategies. If data is inconclusive, it surfaces the uncertainty and halts recommendations.
How analyses are assembled
The assistants follow a repeatable set of steps for every matchup.
- Live data sweep: run web search to capture injuries, lineup changes, weather, and odds movement. Sources are cited inline for transparency.
- Probability math: convert decimal, American, or fractional odds to implied probability and display the math.
- True edge calculation: weigh contextual factors (rest, travel, patch notes) to project a true win probability, then calculate EV.
- Bankroll guidance: if EV is positive, compute Kelly stakes and suggest fractional Kelly when variance is high. Otherwise, recommend passing.
- Responsible wrap-up: restate uncertainty bounds and close with the disclaimer every time.
Available personas and focus areas
Pick the sport-specific assistant that matches your market.
The Bookmaker (NBA)
Injury intel, pace, rest splits, and line movement.
Field Intel (NFL)
Quarterback health, weather, matchup trenches, and vig analysis.
Diamond Eye (MLB)
Pitching matchups, bullpen fatigue, and park factors.
Summoner Sage (LoL)
Roster swaps, patch impact, side selection, and draft edges.
EV Sniper (CS2)
Map veto dynamics, player form, and lineup volatility controls.
Pitch Master (Soccer)
Global football tactician weighing form, injuries, and expected goals versus market odds.
Get the best signal from prediction personas
Feed them current market context, ask for comparisons, and keep a manual decision loop in place.
- Provide the exact market (moneyline, spread, totals) and current price to ground the analysis fast.
- Request scenario comparisons (e.g., "what if the line moves to -2.5?") to stress-test edges.
- Use code interpreter for custom bankroll simulations or to digest data dumps.
- Loop in compliance by sharing conversation summaries and saving links to key analyses in your tracking sheet.
These agents are not wired into sportsbooks and should never trigger wagers automatically. Treat them as research copilots only.