PropsBot has tracked 594,000+ AI Signal predictions across NFL, NBA, MLB, and NHL — every prediction logged with the player, market, line, direction, Confidence Score, Edge Score, and final result. Our High ROI Signal has produced a 16.8% ROI across 8,571 graded NFL picks (51.7% win rate, +1,443 units) 25.9% across 81,410 graded NBA picks (47.1% win rate, +21,051 units), and 15.2% across 9,913 graded NHL picks (44.9% win rate, +1,508 units), as of September 29, 2026, and our model’s probability calibration beat Vegas in an MLB calibration study dated July 12, 2026.
How We Measure Performance
We track every single prediction PropsBot issues — no cherry-picking. Each pick is logged with the player and market (e.g., Travis Konecny over 1.5 shots on goal), the sportsbook line at the moment of prediction, the model’s projected probability, the Confidence Score and Edge Score, and the final outcome.
You can browse the full ledger live at dashboard.propsbot.ai/Performance — filter by sport, signal, market, or date range. We don’t claim “guaranteed locks.” We don’t recycle picks. We show our math and our misses.
AI Signals — The Filters That Drive PropsBot’s Headline Numbers
💰 High ROI Signal — “Where the math says you’ll profit long-term”
The High ROI signal flags props where the model’s projected probability is meaningfully higher than the sportsbook’s implied probability after vig is removed. These are positive-expected-value (+EV) bets.
| Sport | ROI | Win Rate | Graded Picks | Units Profit |
|---|---|---|---|---|
| NFL | 16.8% | 51.7% | 8,571 | +1,443u |
| NBA | 25.9% | 47.1% | 81,410 | +21,051u |
| NHL | 15.2% | 44.9% | 9,913 | +1,508u |
| MLB | Current MLB signal results: live ledger at dashboard.propsbot.ai | |||
🎯 High Hit Rate Signal — “The picks our model agrees on most strongly”
The High Hit Rate signal flags props where multiple internal models converge on the same direction with high confidence. These are picks where the model’s conviction is strongest.
| Sport | Win Rate | Graded Picks | ROI |
|---|---|---|---|
| NFL | 74.4% | 22,210 | −5.0% |
| NBA | 76.6% | 200,885 | -1.4% |
| NHL | 84.4% | 35,027 | −2.7% |
| MLB | current MLB signal results are on the live ledger at dashboard.propsbot.ai | ||
Calibration — How We Know Our Probabilities Are Honest
A picks service that says “70% chance to hit” is meaningless if the actual hit rate at that confidence is 50%. To prove our probabilities are honest, we measure calibration — how well predicted probabilities match actual outcomes.
The standard metric is the Brier score — used in weather forecasting, prediction markets, and quantitative finance. Lower is better.
| Sport | PropsBot Brier | Vegas Brier | Edge |
|---|---|---|---|
| MLB | 0.1903 | 0.1947 | +0.0044 ✓ |
| NHL | 0.1846 | 0.1865 | +0.0019 ✓ |
Translation: In the July 12, 2026 MLB calibration study, when PropsBot’s model said a prop had a 70% chance to hit, the actual hit rate was closer to 70% than the sportsbook’s own implied probability was.
This is the core trust claim PropsBot makes that no other AI player prop service does — verifiable, third-party-checkable, and updated continuously. See live calibration plot →
How PropsBot Generates Predictions
PropsBot runs an ensemble of independent machine-learning models that analyze multiple data streams simultaneously for every player prop on the board:
Player Performance Trends
- Rolling averages over the last 5, 10, and 20 games
- Recent form vs season baseline
- Performance by venue (home/away/neutral)
- Performance vs specific defenses or pitcher quality
- Usage rate and minutes/innings/snap-count trends
Opponent Matchup Analysis
- Opponent defensive efficiency by stat category
- Pace of play and possession context
- Pitcher strength (MLB) — handedness splits, recent form, BvP history
- Goalie strength (NHL)
- Coverage scheme (NFL)
Sportsbook Implied Probability
- Live sportsbook line ingestion across DraftKings, FanDuel, BetMGM, Caesars, ESPN BET
- Vig removal to compute true implied probability
- Cross-book comparison for line-shopping edge
Live Line Movement
- Movement direction and magnitude
- Sharp vs public side detection
- Steam moves
- Late line corrections
These four streams feed our Confidence Score and Edge Score.
The Confidence Score Explained
The Confidence Score (0–100) measures how strongly multiple models agree on a prediction. Higher = stronger conviction.
Hit rate by Confidence Score band (across all sports, all signals):
Most subscribers filter the daily slate by Confidence Score 75+ before reviewing matchups in detail.
The Edge Score Explained
The Edge Score = (PropsBot Probability − Sportsbook Implied Probability) × 100
A +15 Edge Score means our model estimates a 15-percentage-point advantage over what the sportsbook prices in. Edge is a measure of mispricing, not certainty — pair high Edge with high Confidence for the strongest plays.
How We Validate Our Results
- Every prediction is logged with player, market, line, direction, Confidence, Edge, and result.
- No cherry-picking — all picks meeting the signal threshold are included in headline metrics.
- Live verification — you can audit the full ledger at dashboard.propsbot.ai/Performance right now.
- Daily updates — sample sizes grow continuously as more games settle.
Important Disclaimers
PropsBot is a research and analytics tool, not a picks service. Past performance does not guarantee future results. Sports outcomes involve inherent randomness, and no model can predict every result accurately.
Always bet within your means. If you or someone you know has a gambling problem, call 1-800-GAMBLER or visit ncpgambling.org.
Frequently Asked Questions
How does PropsBot’s High ROI Signal compare to typical sports betting ROI?
A breakeven sports bettor wins ~52.4% of -110 bets. A “good” sharp bettor profits at 5–10% ROI long-term. PropsBot’s High ROI Signal has produced 16.8% ROI across 8,571 graded NFL picks, 25.9% across 81,410 graded NBA picks, and 15.2% across 9,913 graded NHL picks (as of September 29, 2026) — above the sharp-bettor threshold, with sample sizes large enough to be statistically meaningful.
Why does PropsBot’s calibration matter?
Most picks services don’t quantify how accurate their probabilities are — they just count wins and losses. Calibration tells you whether the probabilities themselves are honest. PropsBot’s MLB Brier score of 0.1903 beat Vegas’s own implied probabilities (0.1947), and its NHL Brier score of 0.1846 beat Vegas’s 0.1865, in a calibration study dated July 12, 2026, meaning our 70%-confidence picks hit close to 70% of the time in that sample.
Where can I see the live performance data?
At dashboard.propsbot.ai/Performance. Every prediction is tracked. You can filter by sport, signal, market, and date range.
Are these numbers updated live?
Yes. The dashboard refreshes as new games settle. The numbers on this page reflect the live state of the ledger.
Why does ROI vary by sport?
Different sports have different market efficiencies. MLB props (especially pitcher and batter props) tend to have softer lines than NBA props because there are more games, fewer betting eyes per market, and stronger structural inputs (pitcher matchups, weather). PropsBot’s signal performs strongest in MLB historically.
