CME SETUP BOARD · every situation the engine can see

ReceiptThe expansion gate is armed and running in shadow at precision 0.4623 and lift 1.53× on the production rule, which is the honest operating point rather than the flattering one. The 0.5669 headline it was first quoted at read three features that can only be known after the fact, and one of them was the rule's own entry minute.

What is real: the gate separates expansion days from the rest, and it does it before the trade rather than after. What is thin: at 0.287 gated trades a session the machine needs about 349 sessions to reach a hundred trades, so it will not have an opinion worth arguing with inside a year. You will have seen a hundred of these long before it does.

ReceiptA powered zero rather than a shrug: −0.0078 R at t −0.206 across 1,369 clusters, with the test able to see an effect of 0.1226 R against a 0.17853 R target. Both directions were tried and both are empty, so there is no sign hiding inside it waiting to be found.

What died: the automatic version, because the machine has to guess which way the break goes and its guess is worth nothing. What is real, and this is the whole card: handed the direction in advance, the identical entry, stop, target and toll earn +0.4612 R a trade at a 51.28% win rate. So the geometry pays, and the only piece missing from it is the direction. The machine has none; you might.

ReceiptTwo instrument families digest their gaps at genuinely different rates, measured on our own corpora: both equity indices give back about 67% of the gap by the close, while both large-gap commodities give back about 42%. Gold closes its gap outright on 17.05% of gap days and crude on 17.91%, so five gaps in six never come home.

What died: the mechanical fade at a chosen fraction, and its continuation mirror with it, at −0.0726 R across 841 trades and t −2.058. That geometry is spent on both corpora and must not be re-run. What is real: the digestion rates themselves, and the finding that the low-volatility tercile digests fifteen to seventeen points more than the high one, which points the opposite way to the folklore everybody repeats.

ReceiptCompression really does carry forward, and the size of the carry survived the hardest control this desk owns: M2K lifts +18.65pp pooled against a +10.64pp bar, and MNQ's corrected carry is +5.27pp for compression and +8.44pp for expansion.

What died: arming on yesterday. Below about three weeks the carry cannot be told apart from a drifting volatility level, so a rule that fires because yesterday was quiet is mostly firing because the month is quiet. What is real: the multi-week regime, which belongs in the size dial and in your expectations for the session. And the honest warning underneath all of it is this: nobody is obliged to break a coil, so there is no compelled counterparty here the way there is at a settlement or an expiry.

ReceiptRefused at design this morning, and refused for want of events rather than for want of an effect. The event fires 98 times on 90 clustered days across seven years and three months against a floor of 761 clusters, which is short by 8.5×. The regime that actually makes it a climax, a burst inside an already-mature leg, holds five clustered days in the entire corpus.

What is real: the situation, and your ability to see it forming. What is honest: it was never refuted, it was found unmeasurable at the size that makes it a climax, and those are different sentences. Five days in seven years is nothing to a statistical test and quite a lot to a pair of eyes that only need to be right on the day. The machine can flag it forming; the judgment is yours.

ReceiptRefused at design with zero readings spent and zero dollars spent. The pooled window carries 220 clusters at a minimum detectable effect of 0.2784 R, short of target by 1.56×, and on the calendar-matched contrast that carries the verdict it is short by 1.69×. The named releases are worse: inflation short by 2.98×, payrolls by 3.06×, and the rate decision by 3.70×.

What is real: the volatility, which is why the temptation exists in the first place. What the map says: we cannot gather enough of these events to learn anything from them inside any corpus we will ever hold, so a rule built here would be a rule fitted to a handful of afternoons. Stand aside mechanically. If you take one by hand, take it knowing that this board has no receipt to give you and no geometry it can defend on your behalf.