The Free Meta Sweep Was Lying to You. I Took It Down.

Up front: this one is me admitting a number on my own site was wrong. If you're on the free tier and you ran a meta sweep — the button that scored your deck against the top archetypes at once — the win rate it gave you was too low. Not noisy-around-the-truth low. Systematically low. If you cut a card because the sweep said your deck was weak, that call was made on bad information, and I'm sorry.

Single matchups (your deck vs one specific deck) were always fine, and they just got better. Details below.

What went wrong

Every simulated matchup is played out many times, because a single game tells you nothing — someone draws well and wins. The more games, the more the luck averages out. The sweep touches every archetype at once, so it costs a lot more to run than a single matchup does. To keep it affordable to give away, the free sweep ran very few games per matchup.

I assumed that made it noisier. Rougher number, right ballpark. That assumption was wrong, and here's why.

The rating you see isn't the average of how the games went. It's closer to the worst of the ways your deck can be played — if there's a line of play that beats you, I'd rather tell you your deck is fragile than flatter it. That's deliberate and I'm keeping it. But "the worst result across several attempts" over a handful of games mostly finds bad luck, not a bad matchup. Run it a few more times and the worst one you happened to see is usually just the unluckiest one, not a real weakness.

So the error doesn't scatter around the truth. It sits underneath it. And it gets worse the fewer games you run.

How much it was off

I measured it: the same 28 matchups, same decks, same seeds, run at the free sweep's depth and at proper depth. Average across all of them:

Eight points is enough to turn "this deck is fine" into "this deck is bad." But the average hides the real damage. Individual matchups were far worse, and they clustered on one kind of deck — the ones built around blockers, which are roughly half the decks people actually play:

Look at the first row. The free sweep told you that matchup was unwinnable — 1.5%, don't bother. It's a coin flip. That is not a rough estimate. That's a wrong answer with a confident face on it.

What I changed

Single matchups now run 10x more games, for everyone. Free and Patreon run the exact same simulation at the exact same depth. There is no "free version" of a matchup any more — if you run your deck against a specific deck, you get the same number a patron gets. That's the part I'm actually happy about: the free tier got more accurate, not less.

The meta sweep is now Patreon T2. I know how that reads — a thing that was free is now paid, and that's a fair thing to be annoyed about. But the honest version of the choice was: ship the sweep cheap and wrong, or run it deep for everyone and eat a server bill I can't cover, or put it behind the tier that pays for the servers. I picked the third. Shipping the first one is what got me into this post.

Old sweep results are still visible, and they're still the old numbers. I left them there rather than deleting your history, but if a sweep on your deck predates this change, treat the number as too low — especially against blocker decks. Re-run the matchups you actually care about instead; those are free and they're accurate.

The usual caveat

All of this is simulator data from a tool I built, not tournament results. It plays both decks out under the real rules — resources, timing, keywords, blocking — and makes a competent, consistent decision at each step. It won't find every line a great player would. Read it as "what happens when both decks are piloted reasonably," not as gospel.

Which is exactly why the sample size mattered so much, and why I'd rather show you fewer numbers I trust than more numbers I don't.

If you can't afford the Patreon T2, reach out in https://www.gundambay.com/contact.