April 2026 Recap: +295 Units, 32.4% ROI

+294.64 units. 3,634 bets. 32.4% ROI.

That's April 2026. Best month of the year so far. Clears February's 31.5%, well past March's 25.7%. A $10 flat bettor following every play cleared $2,940 in 30 days. Four months in, four green months. Running total: +1,073 units.

I shipped two upgrades on April 15. New LOL model, plus a data-driven player ban list. Both went straight into the second half of the month and lifted the whole portfolio. Two new sportsbooks went live on April 6. And a few hundred bets from the last week of April are still grading, so the final number will tick up from here.

60.5% win rate across 3,600+ graded bets. CS did the heavy lifting on volume. LOL caught fire after the model swap. VAL cooled. DOTA quietly recovered. COD was the quietest sport, still positive. The portfolio does what the portfolio does.

CS2 keeps printing

1,059–685 at 60.7% win rate. Over 1,700 graded bets in a single month, +149 units, +34.3% ROI. The CS rebuild from February 13th has now had two and a half full months to prove itself, and it keeps getting better. February was 19% ROI overall (because the old model dragged the first two weeks down), March was 21%, April is 34%. Trend line is up.

More than half of April's profit again came from CS. No other sport puts up this kind of volume, and the model is running well above the breakeven line on every batch.

New LOL model: the rebuild paid off

Teased this in last month's recap. Here's the data.

I shipped the new LOL model on April 15. Splitting April in half tells the story:

• April 1–14 (old model): 235–156, 60.2% win rate, +30.6% parlay ROI
• April 15–30 (new model): 279–166, 62.7% win rate, +53.7% parlay ROI

Two and a half points of win rate, +23 points of parlay ROI. On a sport that was already the most dialed model in the lineup. League was the one I least expected to find more juice on, and here we are.

Data-driven player ban list

Same day as the LOL upgrade, April 15, I shipped a model-wide player ban list. Discord announcement was simple:

“Players are only banned if there's statistically significant evidence they're underperforming expectation. The list is fully data-driven and updates over time. Backtested to improve opening accuracy, closing accuracy, and overall P&L. Live across the model now. No emotions. Just data.”

Roughly +0.3% accuracy lift across the board, applied to every book. That sounds small. On 4,000+ bets a month, it compounds into real dollars. And the bigger win is structural: every time someone in Discord says “ban this guy, his lines are cooked,” the answer is now “the algorithm already checked, and he's either on the list or he isn't.”

And yes, Satanic and Donk are on the list. You're welcome.

VALORANT cooled

302–204, 59.6% win rate, +25.3% ROI, +32 units. Last month I wrote “VAL hasn't gotten the memo” about books sharpening up new markets. April is the memo arriving.

66% in March, 59.6% in April. Still profitable, still well above breakeven, but the easiest edge of the year is now priced in. That's how every new market goes. A 60% floor is roughly where I'd expect a mature VAL model to land long-term anyway.

DOTA bounced back, COD stayed quiet

DOTA: 213–148, 59.1% win rate, +21.5% ROI, +19 units. Last month I called the March DOTA dip variance and said two months from now it'd probably flip. One month later, win rate is up almost two points, ROI is up 14. Not a flip yet but the regression I expected is showing up on the timeline.

COD: 109–79, 58.0% win rate, +13.6% ROI, +6 units. CDL was on a light schedule in April. Under 200 bets the whole month, the smallest sample of any sport. Don't read too much into the headline number on a sample that thin. The split underneath is the actually interesting part: UNDER bets went 89–58 (60.5%, +11 units), OVER went 20–21 (48.8%, -5 units). The model is finding clean edge on the UNDER side and basically nothing on the OVER side. Same pattern shows up in CS and DOTA at smaller magnitudes, but on COD the sample is small enough that it dominates everything else.

Five-sport portfolio doing five-sport portfolio things. VAL cooled, DOTA recovered. CS and LOL printed. COD was quiet. Spread the bets across five sports and the math averages out at +32% for the month.

Bovada and Stake are live

Two more straight-betting sportsbooks went live on the model on April 6.

Stake is the volume play. 1,776–1,348, 56.9% WR, +136.91 units, +4.4% ROI in April. Filter to the higher-EV plays only and ROI runs in the +9–11% range. One book, one month, traffic that's actually allowed to fire bets.

Bovada is lower volume, higher hit rate: 114–77, 59.7% WR, +22.40 units, +11.7% ROI. Filter to higher-EV plays and the win rate jumps into the 60s.

What's next

Everyone keeps asking me about the Polymarket model. Don't worry, I'm in the lab cooking. More when it's ready.

If you want to try it, there's a free 3-day trial. Five esports, twelve sportsbooks, scanning prop markets 24/7. Takes 15 minutes a day once you get the hang of it. See you in the May recap.