Thursday August 27 · 4:00 to 8:00 PM ET · eight names — 3v4 for the opener, then even 4v4 · nobody sits. Night 79 of a record that starts 2025-02-09 — and only the third Thursday in all of it. Joe called in at lunchtime: no double shift after all, so he is in at five and the lobby fills to eight — which makes every series after the opener an even 4v4, the best-shaped night this lobby has had in weeks. Four graded series, plus three optional modes Kyle and Zach cooked up: take them or leave them, the sheet stands on its own. Teams first, rules and the math below. Host: rotating.
Power levels — current form straight off the True Strength board, recent games weighed heaviest (half-life 90 days). 9,000 is the lobby average, and double someone's power level means the model makes you a 2-to-1 favourite over them:
The question marks — priced on bands, explained in “How the night is priced” below:
Out this week: Don (out) · Justin (out) · Gomeh (out) · Melo (out) · Jared (out) · Alan (out) · Zion (out) · Sam (out) · Mo (out) · Bryan (out)
Seven for the opener, eight from five. The gold cards are the graded series, and they are the night — the sheet stands on its own if nobody feels like experimenting. The purple cards are OPTIONAL: try them, skip them, or bin one after a round. They are not priced, not graded and go nowhere near anyone's rating, and the rules on them are first drafts meant to be argued with. Anyone with time today should make them better:
seven in the lobby — the opener runs 3v4 with bots filling to 6v6
JOE LANDS — his double shift fell through and he called it in. Eight is EVEN, so every series from here is a straight 4v4: same number of humans a side, same two bots a side, and the extra-body guesswork drops out of the odds entirely. Best-shaped night this lobby has had in weeks
The sheet picked six heroes, two per role, and both teams field the same six. Hero advantage cancels exactly, so what is left is who actually played better — the purest read of the night, and it counts for nothing. On purpose.
Take the sheet’s last split and repurpose it: one side hides, the other hunts, then swap so everybody gets a turn at both jobs.
The one mode that wants an odd split. One human is the BOSS, alone, on something with beef, bots as minions. Everybody else is the RAID. Seven versus one, and the one is supposed to be scary.
The lobby dies when it dies. The four graded series get harvested and graded against this sheet the same as every week. The optional modes do not — if the Mirror or Boss Fight shows up in anyone's numbers, that's a bug, and it'll get fixed.
2+ hours notice = exempt. No notice? Clock starts at 4:00, stacks every 15 minutes: your #1 hero banned for your first series → top 2 → top 3 → one more every 15.
Exempt tonight: Joe — he gave notice at lunchtime and lands at 5:00. For the other seven the clock is live at 4:00.
Last week the pre-night favourite went 3–7 and the model earned that: Melo debuted 9-1-1 and finished with the most kills in the lobby, which no prior built from eleven previous debuts was ever going to see coming. Tonight is the first night all year with no first-timers, so there is no debut prior in play at all — and one graduation: Jonathan cleared the twenty-recent-game bar and prices as a rated player for the first time. That leaves Chris as the only question mark in the room (Chris 6,675 (anywhere from 4,750 to 9,375) · Joe 8,950 (anywhere from 6,325 to 12,650)), swept across his band on every call. The band under each series' odds is where the call lands across all of that, including the extra-body term for 3v4 (+32 pts, CI -64..+121) — every series tonight is uneven, so every band carries it. A wide band is the sheet telling the truth.
One system, finally actually one system. The balancer on The Tools was rebuilt in August to run the sheet's model instead of Elo — but this file was still fitting its own version of current form, with a hard cliff at ninety days instead of the shipped model's smooth decay. Tonight they disagreed badly enough to matter: this sheet would have priced Ice at +82 while the site said +27, and published a run sheet crowning Ice the top of the lobby while the website crowned Tomas. Fixed at the source — the run sheet and the balancer now read the same numbers out of the same model, so the tool you screenshot and the sheet you argue with cannot contradict each other any more. Tonight's board: Tomas 13,225 · Kyle 12,075 · Michael 10,875 · Ice 10,525 · Zach 9,650 · Joe 8,950≈ · Jonathan 6,950 · Chris 6,675≈.
The fairness receipt — every number here is computed by the balancer, not typed in. All 4 graded series land inside 46–54 (worst call 3.6 points off a coin flip, mean 1.8), no two players are teammates more than 3 of 4 series, and the extra-bot seat is carried by Chris, Michael, Tomas in the single uneven series, and by nobody after that. How much unfairness the rotation is allowed to buy was swept for this exact pool, not inherited — the allowance landed at 4 points. The tighter setting was available and rejected on purpose: it produced a night where Joe played all three of his series with the same teammate, which is the very thing that got a sheet binned on 8/14. Four points of odds is inside the model's own tie margin; a welded pair is not. With only 8 names the whole search space is small enough to enumerate, so these 4 series are the provably best night available under those rules — not a good guess.
Every player who ever debuted here, what their first 20 games said, and what happened next:
That table is the argument for patience, and tonight it has two live examples. Tomas opened at a power level of 2,675 — still the lowest opening on record — and across his career he has become 3.4× the player who debuted, which is how somebody who started at the very bottom of that column walks in tonight ranked first in the whole lobby. (Career and tonight's current form are different windows, so the numbers in this table won't match the board up top — that column is a lifetime read, the board is recent form.) Jonathan spent five nights priced off too little data and tonight he is a rated player: no band, no asterisk.
And Chris is the clearest picture of the thing working, because you can watch it happen night by night. He opened at 1,625 off eight games on 8/14 — the harshest opening this record has ever handed anyone — then 6,450 a week later, and 6,675 tonight. That is 4.1× stronger in three nights, and the honesty band around him has more than halved on the way. He is the last person here still carrying a band at all, which is not a judgement — it is just the width of what four nights can prove. A rating needs about twenty recent games before it means much, so if your number looks wrong, the answer is games, not argument. Nobody is judged on night one. Or night four.
11 customs, twelve names, nobody sat. The pre-night favourite went 3–7 — the worst per-game night this sheet has had — while side A won 5 against 4.8 expected. Both of those are true at the same time: the night's aggregate was almost exactly right while the individual calls were a coin flip, which is what a night hinging on one unpriceable debut looks like. Melo went 9-1-1 with a 42-bomb; the prior had him where it has every first-timer. Kill leaders: Melo 320/69 · Kyle 305/66 · Ice 288/73 · Zach 263/84.
What the night could and could not settle: Chris re-fit to +10 (95% range -395..+415) — real movement, and he carries that read into tonight. Gomeh and the debuts sat on opposite sides nearly all night, so "Gomeh is worse" and "the debuts are better" explain the same results, and the grade refused to pick between them. That refusal is the feature, not a gap. The repetition number was ugly (Chris+Zach together 10 of 11 games) but that one is not the sheet's to answer for — the lobby improvised its own closers after Tomas left at 6:30. Full grade preserved in the data; dossiers in the Hero Lab.
Odds = a Bradley-Terry team strength model fit on our full record of decisive customs, validated at 58% accuracy on 172 held-out even-sided games it had never seen, with confidence shrunk by 0.5 to what our stated odds have actually cashed at. Ratings are current form, last 90 days. Every band on this sheet is computed by sweeping the unknowns (debut range, every thin-fit band, the extra-body CI on uneven blocks) — a wide band is the sheet telling the truth. Role shares are colour only; every veteran flexes. Full receipts at marvelgooners.com.