Another day, another winning session. March 3rd saw us go 3-2 on NCAAB ATS (+2.4 units) and 1-1 on NBA Totals. Not a night that’ll make the highlight reel, but that’s exactly the point — we’re grinding, stacking small edges, and letting the math compound over time.
Where We Stand
Here’s the season snapshot since we launched V9.4 on February 19th:
- NCAAB ATS: 16-10 (61.5%) — +15.0 units — +19.2% ROI
- NBA ATS: 4-3 (57.1%) — +2.2 units — +9.0% ROI
- NBA Totals: 2-1 (67%)
Positive units across every model. That’s not luck — that’s a system built on data doing what it’s supposed to do — the Donnie Dimes system.
The Part Most People Skip: Nightly Optimization
Here’s what separates The Lab from someone flipping coins on Twitter. Every single night after the games settle, our system runs a full optimization cycle — it’s why AI is the only edge left in modern betting. What does that actually mean?
- Rolling recalibration: The model parameters — things like home court advantage weight, strength of schedule multiplier, and the core prediction coefficient (k) — the same efficiency rating math that powers every pick — aren’t set in stone. They’re re-evaluated against the latest results every night.
- Backtesting against live data: We don’t just test against historical snapshots. We run honest, no-lookahead backtests using rolling statistics — the same data the model would have had available on game day. No cheating.
- Multi-sport expansion: NCAAB is the flagship, but the NBA model is live and we’re building out NHL and MLB for their upcoming seasons. Each sport gets its own architecture tuned to its market dynamics.
Tonight’s optimization moved our k-value from 0.30 → 0.24, with the updated backtest showing 63.4% win rate and +20.9% ROI. The model is getting sharper every day.
Why Boring Wins Are the Best Wins
A 3-2 night doesn’t get people excited. Nobody’s posting 3-2 on social media with fire emojis. But here’s the thing — sustainable profitability is built on nights exactly like this. You take your edge, you manage your sizing, and you let variance play out over hundreds of games.
The models don’t chase. They don’t tilt after losses. They don’t double down because “this one feels good.” They evaluate, assign probability, size accordingly, and move on. That’s the whole philosophy.
What’s Next
March Madness conference tournaments are right around the corner, which means more volume, more data, and more opportunities for the model to find edges. The NBA is in its stretch run with playoff positioning on the line — tighter games, more motivated teams, and totals markets that are increasingly exploitable.
We’re also deep into building out our MLB model ahead of Opening Day on March 27th. Early backtests are showing 52% win rate with +16.1% ROI on filtered plays — and we haven’t even started optimizing yet.
The grind doesn’t stop. Every night we get smarter. Every game is more data. And the bankroll keeps moving in the right direction.