February 2026 Recap: +342 Units, 31.5% ROI

+342 units. 4,339 bets. 31.5% ROI.

That's the best month the model has ever had. A $10 flat bettor following every play cleared $3,400+ in 28 days. January was a great month. February doubled it.

But what makes this month different isn't just the number. I shipped three major model updates in February. New DOTA, new COD, and a full CS2 rebuild. All three hit immediately. Every single model finished the month above a 60% win rate. That doesn't just happen.

60.5% win rate across 4,339 bets. No single game carried. No single day carried. Just thousands of +EV spots, day after day, compounding. The boring answer, but the real one.

The CS2 problem

Let me start with CS2 because it's been the elephant in the room for a while. If you were in Discord the past couple months, you saw the messages:

“ngl i'm a larry glazer but the cs model is fucking ass”

“Will there be a new cs model any time soon?”

“Gonna keep shit talking it till it hits”

Look, CS2 is a volatile game. Match fixing, upsets, the kind of randomness you don't get in League or DOTA. But the complaints weren't wrong either. The edge had dulled since the model first launched in August. Books adjusted, win rates came down from where they were in the fall. It was still profitable, but not at the level I wanted it. So I rebuilt it.

The new CS2 models went live February 13th. Here's the full month including the old model's first two weeks:

19% ROI for the full month, which is still insane. 19% return on every dollar wagered is an infinite money printer in any other context. But the new model is on another level. Since Feb 13:

46% ROI. 62% win rate. On CS2. With that kind of volume. The thing people were telling each other not to touch is now one of the strongest models in the lineup. To the people who kept shit-talking it: you made me fix it faster. Appreciate that.

New DOTA 2 model

DOTA was already one of the best performers in the arsenal. The old model didn't need an architecture change. It needed better data. I'd been working on the data quality pipeline since October, and it finally shipped in February. The result: up to 50% more bettable lines at 60%+ probability, and backtests showing 7% higher accuracy than the old model across the expanded coverage.

54% ROI. More volume, higher accuracy. When you already have the best model and you make it better, this is what happens.

COD is back

I had a COD model last year but took it down. New season, new game, old model didn't translate. So I rebuilt it from scratch starting in January, ran it through a few weeks of internal testing, and pushed it live February 19th. More volume and a higher hit rate than the old version ever managed.

63% win rate in ten days. Early, obviously. But it's the same pattern as every new model launch. Fresh markets, fat edges, lines that haven't been sharpened yet. If you're not betting COD props right now, you're leaving money on the table.

VALORANT and LOL update

VALORANT is in its second full month. January was 62% ROI, which was always going to come down. Here's where it landed:

39% ROI. Books are adjusting but they haven't caught up yet. League also hit above 60% for the month. When every model is performing at that level, it's not a hot streak. It's edge.

Thunderpick is coming

BIG ANNOUNCEMENT: Thunderpick straight betting is launching on the model in March. Real sportsbook, real odds, real player props, not DFS. Available everywhere, no taxes, and they're offering up to a $600 deposit match for new signups.

We're handing out early access to members now. Sign up through our Thunderpick link with code LCSLARRY, deposit at least $25, and you're in. The DFS prop model has proven itself over tens of thousands of bets. Adding straight betting opens up a whole new set of +EV plays that weren't possible before.

Meanwhile, Twitter thinks I'm washed because the free picks haven't been running hot lately. Variance on a small sample of cherry-picked plays will do that. The actual model, all 4,339 bets of it, just had its best month ever. Make of that what you will.

If you want to try it, there's a free 3-day trial. Five esports, eleven sportsbooks, scanning prop markets 24/7. Takes about 15 minutes a day. March is shaping up to be even bigger with Thunderpick and COD both ramping up. See you next month.