One site doesn't recycle what you pull. We checked whether that helps you.
Beezie's machines genuinely drain and restock. What that does to your odds is stranger than we expected.
Most gacha machines put what you pull straight back in. Sell a card back on Phygitals and it's available to the next opener within about twenty seconds — the pool barely moves. Beezie doesn't do that. Their team told us they hold swapped items aside and let a machine run down until the good cards are gone, then restock it.
That's an unusual claim, so we checked it against 25,488 machine snapshots we'd already recorded. It's true, and you can see it in the stock counts: one machine drained from 813 cards to 145, jumped back to 797, drained to 435, jumped to 807. Another fell to 32 before a restock took it back over 2,000. A third restocked 17 times in 475 hours. That's a real sawtooth, not marketing.
Which raises the obvious question: if grails leave the pool and don't come back, is a freshly restocked machine worth more than a drained one? Here is our honest answer. Comparing fresh against drained within each cycle, the gap is −2.78 percentage points, with a 95% confidence interval from −10.22 to +2.59. That interval comfortably contains zero, so we cannot show you a timing edge — but it is also wide, and it rests on just 16 drain cycles. Too few to rule one out either. Anyone quoting a tighter number on this is counting five-minute polls as if they were independent observations. They aren't; a machine barely moves over hours.
The direction is the interesting part, because it isn't the one the story predicts. Three of four machines lean the other way — the drained machine reads slightly richer. There's a mechanical reason: Beezie's published grail odds are simply how many grails remain divided by how many cards remain, and draining shrinks that denominator faster than pulls remove the rare cards. One machine sat at 24 cards with 3 grails still in it — a 12.5% grail rate that no fresh machine ever offers.
Before anyone builds a strategy on that: don't. Deep drain happened in 18 of 103 cycles, usually briefly, and usually ended by a restock rather than by you getting there first. The median deeply-drained machine sat at 1.034 against 1.048 fresh. Out of 25,488 snapshots, exactly 3 showed a machine priced better than 1.20. It is a real quirk of the arithmetic, not an opportunity.
One more thing, because we'd rather say it than have you find it. While checking this, we discovered our own Beezie return figures were too generous. On a buyback the machine's treasury pays out a gross amount and the player receives 94% of it after the swap fee — and we had been counting the gross. Corrected, the site-wide recovery we show for Beezie drops from 103.8% to 97.8%, and every machine reads negative again. The old number wasn't Beezie being generous. It was us measuring the wrong side of the fee.