PropsBot.AI’s verified track record: 31.7% ROI on 101,881 logged MLB player props on the High ROI Signal, 82.6% win rate on 136,953 picks on the High Hit Rate Signal, and a Brier score (0.1903) below the Vegas closing-line baseline (0.1947) in MLB. Live data updates daily at dashboard.propsbot.ai.
What this page is
A summary. The numbers below are the canonical figures we cite when someone asks whether the model works. Each one is sourced from the live ledger.
The live ledger is the source of truth. It’s a separate site at dashboard.propsbot.ai that updates as picks resolve. If you want to audit a specific pick, sort by sport, filter by signal, or pull a date range, that’s where you go. This page is the citable summary the dashboard exists to back up.
If you’re new to how the model is built, /performance-methodology/ covers the audit. /about/ covers who built it.
31.7% ROI on 101,881 MLB props
This is the headline.
The High ROI Signal is one specific filter inside the model. It flags picks where the model has a real edge over the posted line and sizes the stake accordingly. Some picks are full unit. Some are fractional. The 31.7% is the rolling sum of profit divided by total amount staked across all 101,881 logged plays.
Two things to know about the number.
First, it’s a closed sample. Every pick was published before its game started, with a timestamp and a posted line. Nothing in the 101,881 was added retroactively. Open the dashboard and you can see them in chronological order, oldest first, with results.
Second, 31.7% ROI is not 31.7% hit rate. Sportsbook props pay roughly even money after the vig. If you hit 32% at +200 odds, you make money. If you hit 60% at -200 odds, you lose money. ROI is what determines whether the bankroll grows. Hit rate alone doesn’t.
That distinction matters for the next number.
High ROI Hit Rate of 32% across 101,881 MLB props
The High ROI Signal hits 32% of the time. People expect a higher number when they see “31.7% ROI” and they’re surprised the win rate is in the low thirties.
Here’s why both numbers are true.
The signal is fishing for plus-money props where the model thinks the book is too low on a player. Average odds on those picks sit well north of even money. When a +180 hits, it pays 1.8x the stake. So a 32% hit rate combined with that average price produces the 31.7% ROI.
It’s the same math that powers any positive-EV portfolio. Lower hit rate, higher payout, smart sizing. The lever you pull to make money on props isn’t always “win more often.” Sometimes it’s “win less often at better prices.” The High ROI Signal is built around the second one.
If you’d rather watch a higher percentage of picks land, the next signal is the one to look at. Different cohort, different math.
82.6% win rate on 136,953 High Hit Rate picks
The High Hit Rate Signal is a separate cohort. It targets the picks the model is most confident about, regardless of price. Smaller average edge per play. Higher hit rate. More volume.
136,953 picks across the four major sports. 82.6% of them landed.
You’d use this signal differently. It’s the one that gets parlayed, used to tighten same-game builds, or stacked when you want a high-probability anchor inside a riskier ticket. ROI per pick is lower than the High ROI Signal because the prices are lower. The win rate carries the weight.
Both signals run in parallel. They don’t share a sample. The 101,881 MLB props from the High ROI Signal and the 136,953 picks from the High Hit Rate Signal are different ledgers. Conflating them is the most common mistake people make when they read this page.
Brier 0.1903 vs Vegas 0.1947 in MLB
The first three numbers are about whether you make money. Brier is about whether the model is honest.
A Brier score measures how well a probabilistic forecast matches what actually happens. Lower is better. Zero is perfect. A model that says “70% chance” and gets 70% of those calls right scores low. A model that says “95% chance” on a coin flip scores high.
The Vegas closing line is what’s known as the wisdom-of-the-market baseline. By the time the line closes, sharp money has moved it to the most accurate price the book can find. Beating the closing line at Brier is hard. Most public models don’t.
PropsBot’s MLB model comes in at 0.1903. The Vegas closing line on the same set of props comes in at 0.1947. Lower is better, so PropsBot wins.
The same comparison runs in NHL. PropsBot also beats Vegas there.
/brier-score/ walks through what the metric measures and why sharps weight it more than win rate.
Why we publish all of this
Most AI prop tools don’t.
You’ll find services with a polished landing page, a price, and a “trust us” subtext. No ledger. No win rate. Definitely no Brier score. The reason is simple: if you publish those numbers and the model doesn’t work, the service is over.
PropsBot publishes them because the model does work and because the alternative (asking users to take it on faith) is the thing the entire space has wrong. The audit trail is the product as much as the picks are.
If a number on this page changes, the dashboard changes first and this page gets updated to match.
Where the live data lives
dashboard.propsbot.ai is the source.
What you can do there:
- Sort the table by date, sport, signal, edge, or result.
- Filter to a single signal (High ROI or High Hit Rate) and only see that cohort.
- Filter by sport (MLB, NHL, NBA, NFL).
- Pull a date range to verify a specific stretch.
- Click any pick to see the posted line at time of publication and the closing line.
The dashboard updates as games resolve. The numbers on this page are summaries. The dashboard is the receipt.
If you’re considering paying for PropsBot and you want to verify the claim before the credit card comes out, that’s the link. Same numbers, sortable, exportable, with every pick timestamped.
Methodology
Short version: every pick is logged at publication, the posted line is captured, the closing line is captured, the result is graded against the actual outcome, and the rolling totals update automatically. There’s no manual cherry-picking, no soft delete, no “we count this one but not that one” rule.
Full audit at /performance-methodology/.
The methodology page covers grading rules, what counts as a void, how pushes are handled, and how the Brier comparison against Vegas is computed. If you want to recreate the math yourself, that’s where to start.
What’s not on this page
Three honest caveats.
NBA Brier vs Vegas is still under analysis. We’ll publish the comparison when the sample is big enough to mean something. Right now it isn’t.
NFL is a small sample. The model runs there and the picks are in the ledger, but 17 games a week per team for one season isn’t the same as 162 MLB games per team. Treat the NFL numbers as directional until the sample grows.
The two strongest sports today are MLB and NHL. That’s where the Brier-vs-Vegas case is strongest and where the High ROI Signal has the most volume. We’re not pretending we beat every sport equally.
Sample size
101,881 MLB props on the High ROI Signal is a meaningful sample. 136,953 across all four sports on the High Hit Rate Signal is also meaningful. Neither is “millions of bets” — nobody publishing a track record honestly has millions of bets. Both are well above what most public AI prop services publish, and both are big enough that the ROI and win-rate numbers aren’t going to swing on a hot week.
If the sample on a given sport is too small to cite, we don’t cite it. See the section above.
/positive-ev-props/ covers what positive-EV means and why it matters for sample-size questions.
FAQ
How is the data verified? Every pick is timestamped at publication and stored in the ledger before the game starts. The posted line and closing line are captured automatically. Results are graded against the actual outcome. The full ledger is at dashboard.propsbot.ai and is publicly viewable.
Where can I see real-time results? dashboard.propsbot.ai. It updates as picks resolve. The numbers on this page are summary figures; the dashboard is the live source.
Can I export the ledger? The dashboard supports filtering and sorting in-browser. For programmatic access, contact us. We’ve shared the underlying data with reviewers and partners on request.
How often does it update? The dashboard updates pick-by-pick as games close. The summary numbers on this page are reviewed weekly and updated when a meaningful change occurs.
What counts as a logged pick? A pick that was published with a timestamp and a posted line before the game started. Pre-game only. We don’t grade live in-game picks against the same ledger.
Why does ROI differ from hit rate? ROI is profit divided by amount staked. Hit rate is wins divided by total picks. A 32% hit rate at average odds north of +150 produces a 31.7% ROI because the winning picks pay more than even money. Hitting more often at shorter odds can produce a lower ROI. ROI is the number that determines whether your bankroll grows.
Bottom line
31.7% ROI on 101,881 MLB props. 82.6% on 136,953 across four sports. Brier below Vegas in MLB and NHL. +60,778 units tracked.
The summary lives here. The receipts live at dashboard.propsbot.ai. Open both side by side and the math holds up.
