Weekly comparison of BBMI model vs Vegas — spreads and over/under totals
59.2%
Overall O/U
1,958 picks (edge ≥ 2)
63.1%
Edge ≥ 6 pts
766 picks
63.7%
Edge ≥ 8 pts
430 picks
About the Edge Filter & Performance Stats
Headline stats (Overall, ≥6 pts, ≥8 pts) count only picks where edge ≥ 2 pts. The Vegas line is captured at a specific point in time — lines routinely move 1–2 points between open and tip-off, and can vary by a point or more across different books. A difference smaller than 2 pts is within normal market noise and does not represent a meaningful BBMI disagreement with Vegas. Those games are shown in the table but marked ~ and excluded from the record. The ≥6 and ≥8 pt tier labels were identified by analyzing this season's data rather than pre-specified before the season. Starting next season, edge tiers will be fixed in advance. All win percentages include 95% confidence intervals.
Minimum Edge (|BBMI Total − Vegas O/U|):
Filter by Team:
Tip: The model performs best when edge is highest. Try ≥ 6 points to see picks where BBMI most strongly disagrees with Vegas.
Summary Metrics
Sample Size
% Beats Vegas
Wagered
Won
ROI
1,958
59.2%
$195,800
$221,353
13.1%
—
95% CI: 57.0%–61.3%
—
—
—
Historical Results By Week
Team records indicate Win-Loss when BBMI picks that team to beat Vegas.
Weekly Summary
Sample Size
% Beats Vegas
Wagered
Won
ROI
3
0.0%
$300
$0
-100.0%
—
95% CI: 0.0%–56.2%
—
—
—
Weekly Performance Breakdown — All Weeks95% confidence intervals shown
Week
Picks
Win %
95% CI
ROI
4/4 – 4/10
3
0.0%
0.0% – 56.2%
-100.0%
3/28 – 4/3
6
33.3%
9.7% – 70.0%
-35.7%
3/21 – 3/27
18
50.0%
29.0% – 71.0%
-4.8%
3/14 – 3/20
25
52.0%
33.5% – 70.0%
-1.2%
3/7 – 3/13
152
56.6%
48.6% – 64.2%
+8.1%
2/28 – 3/6
214
54.7%
48.0% – 61.2%
+4.4%
2/21 – 2/27
192
59.9%
52.8% – 66.6%
+14.4%
2/14 – 2/20
197
64.0%
57.0% – 70.3%
+22.2%
2/7 – 2/13
191
60.7%
53.7% – 67.4%
+16.0%
1/31 – 2/6
194
54.1%
47.1% – 61.0%
+3.4%
1/24 – 1/30
190
66.3%
59.3% – 72.7%
+26.7%
1/17 – 1/23
215
60.9%
54.3% – 67.2%
+16.4%
1/10 – 1/16
214
56.1%
49.4% – 62.6%
+7.1%
1/3 – 1/9
147
63.3%
55.2% – 70.6%
+20.8%
Season Total
1,958
59.2%
57.0% – 61.3%
+13.1%
A wider 95% CI range indicates a smaller sample where week-level results are less conclusive. Season ROI is approximate.
🟡 Gold rows = Edge ≥ 5 pts — historically highest accuracy tier~ Faded rows = Edge < 2 pts — within normal line movement, excluded from stats
Understanding the Numbers — A Guide for New Visitors
1
Win % (% Beats Vegas)
The share of picks where BBMI correctly predicted which side of the spread would cover. The break-even point at standard −110 juice is ~52.4%. The 95% confidence interval beneath this number shows the plausible range for the true underlying rate — a wider interval means a smaller sample with less certainty. Only picks with edge ≥ 2 pts are included — differences smaller than 2 pts are within normal line movement and book-to-book variation, not a genuine model disagreement with the market.
2
ROI (Return on Investment)
Simulated return assuming a flat $100 wager per pick at −110 odds. Positive ROI means the model has generated paper profit over the tracked period. Past simulated performance does not guarantee future results, and real-world factors like line movement and juice vary.
3
Edge Column — including the ~ marker
Shown for every row — always a non-negative number representing how far apart BBMI and Vegas are on a given game. Rows with Edge ≥ 5 are highlighted in gold — this is the tier with historically the strongest accuracy. Rows marked ~ have edge < 2 pts — within normal line movement and book-to-book variation — and are shown for full transparency but excluded from stats. The Edge Filter above the table controls which games appear.
4
Weekly Breakdown Table
Shows week-by-week results so you can assess consistency rather than just the season headline. Week-to-week variance is normal — even a model with a genuine long-run edge will have losing weeks. Look for the cluster of weekly results around the season mean, and use the 95% CI column to judge how much weight to put on any single week.
5
Sample Size & Track Record
One season of data is a starting point, not a verdict. Forecasting research generally indicates that 3–4 seasons of prospective (pre-specified, out-of-sample) picks are needed to distinguish genuine skill from a good variance run with statistical confidence. BBMI's 2025–26 season results are promising — and will be tracked transparently each year to build an honest long-term record.
