Published July 12, 2026. Figures are a dated research snapshot; the public ledger remains the live source of truth.
Quick Answer
PropsBot’s High ROI Signal recorded 31.7% ROI across 101,881 logged MLB player props. Its separate High Hit Rate Signal recorded an 82.6% win rate across 136,953 picks spanning four sports. For probabilistic accuracy, the MLB model’s Brier score was 0.1903 versus 0.1947 for the Vegas closing-line baseline; lower is better.
These are not blended figures. Each number measures a different cohort or question. The public PropsBot results dashboard contains the timestamped ledger, while the verified track-record summary and performance methodology explain the calculations and grading rules.
Research Snapshot
| Measurement | Reported result | What it answers |
|---|---|---|
| High ROI Signal, MLB | 31.7% ROI on 101,881 props | Did the plus-money, edge-filtered portfolio return more than it staked? |
| High ROI Signal hit rate, MLB | 32% on the same 101,881 props | How often did that higher-payout cohort win? |
| High Hit Rate Signal | 82.6% on 136,953 picks across four sports | How often did the separate high-probability cohort win? |
| MLB Brier score | 0.1903 | How closely did PropsBot’s probabilities match actual outcomes? |
| Vegas MLB baseline | 0.1947 | How did the model compare with closing-market probabilities on the same measurement? |
Why ROI and Hit Rate Must Be Separated
A model can win less often and still produce a positive return when winning prices are long enough. It can also post a high win rate while producing little or negative return if the prices are too short. That is why this report does not use a single headline percentage as a universal measure of model quality.
The High ROI Signal is designed around mispriced opportunities, including plus-money props. The reported 32% hit rate and 31.7% ROI therefore describe the same cohort without contradicting one another. A winning prop at +180 returns more than a winning prop at -180, so price determines how hit rate turns into return.
The High Hit Rate Signal answers a different question. It isolates higher-probability selections and therefore produces a higher win rate, but it should not be compared directly with the High ROI portfolio as though the two signals contain the same picks or prices.
Why Brier Score Matters
Brier score measures probability calibration. A forecast is rewarded when a stated probability matches how often that outcome occurs and penalized when confident predictions are wrong. A lower score is better, with zero representing perfect probabilistic forecasting.
For MLB, PropsBot reported a Brier score of 0.1903 compared with 0.1947 for the Vegas closing-line baseline. The comparison matters because closing prices aggregate information from the market near game time. It asks whether PropsBot’s probabilities were better calibrated, not merely whether a recent group of picks won.
Readers who want the underlying definition and worked examples can use the PropsBot Brier score guide. The methodology page explains how the baseline is derived and how picks are aligned for comparison.
What Was Logged
The PropsBot methodology requires a pick to be recorded before the game starts with a timestamp and posted line. Results are graded after the event, and the closing line is captured for market comparison. The public dashboard is intended to make individual records filterable rather than asking readers to trust a summary claim.
This report is a dated layer above that ledger. It makes the current headline findings easier to cite, but it does not replace the underlying records. When a summary number and the live dashboard differ because additional games have resolved, the dashboard should be treated as the current source.
What This Report Does Not Claim
- It does not claim that every PropsBot sport has identical performance.
- It does not combine the High ROI and High Hit Rate cohorts.
- It does not treat win rate as a substitute for price or return.
- It does not present backfilled selections as pregame predictions.
- It does not guarantee future profit. Model performance and market conditions change.
The current track-record disclosure identifies MLB and NHL as the strongest sports for the Brier-versus-Vegas case. It also notes that NBA analysis is still developing and that the NFL sample is smaller. Those limitations are part of the result, not fine print to remove from it.
How to Audit the Findings
- Open dashboard.propsbot.ai.
- Filter the ledger by sport and signal rather than mixing cohorts.
- Choose a date range and inspect the posted line, result, and available closing-line record.
- Compare the filtered totals with the definitions on the PropsBot track-record page.
- Review grading, void, push, probability, and baseline rules in the methodology.
How to Use This Research
For bettors, the report is a due-diligence starting point rather than a substitute for checking today’s line. Use it to understand which signal a claim refers to, then evaluate the current prop at the price that is actually available. Historical evidence can support trust in a process, but it cannot make a moved or stale line playable.
For analysts, reviewers, and journalists, the report provides definitions that keep the cohorts separate and a direct route to the underlying ledger. Cite the sample, sport, signal, metric, and access date together. A sentence that reports “82.6% win rate” without naming the High Hit Rate cohort would remove the context needed to interpret the figure responsibly.
Interpreting AI Sports Betting Performance
AI sports betting products often market a win rate without defining the sample, odds, or grading rules. That makes comparison difficult. A more useful performance record includes at least the pick timestamp, market and line, odds, result, sample size, return calculation, probability calibration, and known limitations.
Closing-line comparison is also important. A single bet can lose after being placed at an excellent number, while another can win despite being placed at a poor number. Results matter over time, but price quality helps identify whether the decision process was sound before the outcome was known.
PropsBot’s public approach is therefore built around three layers: portfolio return, signal-specific hit rate, and calibration against a market baseline. None is sufficient alone. Together they provide a more complete view of whether the model finds value, how frequently different cohorts win, and whether its stated probabilities are honest.
Citation Notes
Suggested attribution: “According to PropsBot’s July 2026 MLB model performance report, its High ROI Signal recorded 31.7% ROI across 101,881 logged MLB player props, while its MLB Brier score was 0.1903 versus a 0.1947 Vegas closing-line baseline.”
Please link either this report or the public ledger when citing the figures. Because the ledger updates as games resolve, include an access date when referencing live totals.
Frequently Asked Questions
Is the 82.6% win rate part of the 31.7% ROI sample?
No. The High Hit Rate and High ROI signals are separate cohorts. Combining them would make both figures misleading.
Does 31.7% ROI mean PropsBot wins 31.7% of its bets?
No. ROI is profit divided by amount staked. The same MLB High ROI cohort has a reported 32% hit rate because many winning selections were priced above even money.
Why compare with the closing line?
The closing market incorporates late information and market action. It provides a demanding baseline for evaluating the calibration and price quality of a model.
Where are the individual picks?
The individual, timestamped records are available at dashboard.propsbot.ai. This page is a citable research summary.
Are future returns guaranteed?
No. Historical performance does not guarantee future results. Users should verify current prices, use appropriate bankroll controls, and treat a pass as a valid decision.
