We do not manufacture an edge. We live and breathe with probability advantage.

A regime is an assumption about which history is relevant. This protocol exists to keep that assumption explicit (it is written here), testable (it has to pass an admission test, re-run nightly) and reversible (a health monitor says when it is fading). It changes which evidence we weight — it never changes the cap, the risk budget, or the requirement that the forward odds confirm the trade.

The Regime Protocol

Some groups have traded under a genuinely different set of odds since a specific event — energy since the war, memory and the semiconductor capex chain since the summer of 2025. Testing those names on ten years of pooled history understates what the current regime is doing; testing them on the regime alone trusts a few dozen events. The protocol blends the two, adaptively, and shows its work.

Rule 1 · Theme

Defined by exposure, not by sector label

A theme is a named group that shares an economic driver. Any leg of a ratio or calendar spread counts. Membership is a list in one shared module — every study reads the same definition.

Rule 2 · Regime start

A dated catalyst, written down

Each regime begins on a specific date tied to a specific event. The date is stated on this page and can be challenged. A regime without a catalyst is a story, not a regime.

Rule 3 · The estimator

An adaptive blend, not a cut and not a weight

blended = λ · (in the regime) + (1 − λ) · (all history), with λ = n ÷ (n + 10). Few regime analogs → history carries it. As the regime adds events it takes over on its own. A fixed 2:1 event weight was tested and rejected: it moved a 93% regime read from 61% to 62%.

Rule 4 · Like-for-like base

The base rate gets the same blend

An idea is judged against the regime's own drift, not the old one. If a name has gone up 79% of the time in the regime, a 94% setup has +15 points of edge — not +37. This is the guard against a bull run flattering every long.

Rule 5 · Rank and size on the blend

With the thin flag and the cap intact

Themed ideas rank and Kelly-size on the blended numbers. Under 8 regime events the read is flagged thin. The single-name cap and the risk budget do not move: a bigger blended Kelly is a reason to reach the cap, never to exceed it.

Rule 6 · Show all three, always

History · regime · blended — no black box

Every themed idea displays the all-history read, the regime read with its sample size, and the blend. If the three disagree, that disagreement is the information.

blended = λ · regime + (1 − λ) · history · λ = n_regime ÷ (n_regime + 10) · applied to win %, median and the base rate · thin under 8 regime events · horizon 21 sessions

Adopted regimes — registry, admission test and health

Pooled across each theme on three canonical setups. Admission requires ≥ 60 pooled regime events, a pooled lift of ≥ 8 points over history, and ≥ 40% of names (with ≥ 8 regime events) beating their own history by ≥ 10 points. And — the drift check — the pooled regime win has to beat the theme's own regime base rate (any day, same names, same direction) by ≥ 5 points: a group that has simply trended since its start date lifts every same-direction trade, and that is drift, not a setup edge. Health compares the last 60 sessions of the regime with the regime-to-date: fading = the recent window has fallen back to within 5 points of history. The evidence tables are sorted by regime lift — the names at the top are where the regime is real; the ones at the bottom are where it is not.

Energy · war regime

since 2026-03-01regime intact23 of 69 members have test history

The war / supply shock: product cracks and refining margins re-priced, crude carried a persistent risk premium, and energy decoupled from the broad tape.

Canonical setup · pooled across the theme	All history	In the regime	Lift	Edge over regime drift	Breadth ≥+10	Names worse	Admission	Health
Completed reversal from the low (long) ↗	54% · +0.8% n6075	64% · +3.8% n229	+9.7 pts	+4.0 pts drift 60%	44% of 16	2	does not qualify fails: drift	regime intact last 60s: 71% n70 · prior 61%
Hammer (long) ↗	60% · +2.3% n7562	65% · +4.2% n433	+5.0 pts	+4.7 pts drift 60%	26% of 23	6	does not qualify fails: lift, breadth, drift	regime intact last 60s: 76% n111 · prior 61%
Fade the reversal from the high (short) ↗	42% · -1.6% n5911	44% · -0.8% n236	+2.1 pts	+4.1 pts drift 40%	33% of 12	6	does not qualify fails: lift, breadth, drift	regime intact last 60s: 49% n72 · prior 42%

Completed reversal from the low (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
HO=F ↗	38% · -1.9% n52	100% · +14.6% n8	+61.5 pts	66% · +5.5% λ 0.44	59%	+6.9
VLO ↗	61% · +2.9% n313	95% · +10.0% n19	+33.4 pts	83% · +7.5% λ 0.66	79%	+4.4
RB=F ↗	45% · -0.7% n56	78% · +12.5% n9	+33.2 pts	60% · +5.6% λ 0.47	65%	-4.3
MPC ↗	64% · +3.7% n323	94% · +12.8% n18	+30.9 pts	83% · +9.5% λ 0.64	77%	+6.5
PSX ↗	57% · +1.3% n301	83% · +6.6% n12	+26.2 pts	71% · +4.2% λ 0.55	70%	+1.5
OXY ↗	46% · -0.4% n265	67% · +2.1% n6 thin	+20.3 pts	54% · +0.6% λ 0.38	51%	+2.7
COP ↗	55% · +1.2% n303	75% · +6.7% n8	+19.6 pts	64% · +3.6% λ 0.44	53%	+10.7
CVX ↗	56% · +0.8% n312	75% · +5.1% n8	+18.6 pts	65% · +2.7% λ 0.44	57%	+7.4
SLB ↗	46% · -1.5% n275	54% · +8.7% n13	+7.3 pts	51% · +4.2% λ 0.57	56%	-4.9
OKE ↗	61% · +1.7% n326	67% · +3.5% n12	+5.7 pts	64% · +2.7% λ 0.55	61%	+2.9
TRGP ↗	65% · +3.4% n345	69% · +2.5% n13	+4.6 pts	67% · +2.9% λ 0.57	64%	+3.6
BKR ↗	46% · -1.0% n321	50% · -0.8% n12	+3.6 pts	48% · -0.9% λ 0.55	49%	-0.2
EOG ↗	53% · +0.7% n312	56% · +2.8% n9	+3.0 pts	54% · +1.7% λ 0.47	55%	-0.8
XLE ↗	53% · +0.4% n306	54% · +0.6% n11	+1.6 pts	54% · +0.5% λ 0.52	59%	-5.0
LNG ↗	52% · +0.5% n318	53% · +1.9% n17	+1.0 pts	52% · +1.3% λ 0.63	58%	-5.2
CL=F ↗	54% · +0.1% n54	50% · -0.9% n6 thin	-3.7 pts	52% · -0.3% λ 0.38	56%	-3.7
EPD ↗	57% · +0.9% n307	50% · -0.2% n8	-6.7 pts	54% · +0.4% λ 0.44	56%	-2.8
DVN ↗	50% · +0.1% n301	43% · -1.9% n7 thin	-7.3 pts	47% · -0.8% λ 0.41	52%	-5.1
XOM ↗	53% · +0.2% n307	43% · -5.2% n7 thin	-9.9 pts	49% · -2.0% λ 0.41	58%	-9.2
KMI ↗	54% · +0.6% n344	43% · -1.8% n7 thin	-11.5 pts	50% · -0.4% λ 0.41	52%	-2.0
FANG ↗	53% · +1.3% n229	33% · -1.2% n6 thin	-20.0 pts	46% · +0.3% λ 0.38	58%	-11.9
WMB ↗	58% · +1.4% n333	30% · -2.1% n10	-27.7 pts	44% · -0.3% λ 0.50	52%	-7.6
NG=F ↗	29% · -2.4% n72	0% · -12.0% n3 thin	-29.2 pts	22% · -4.7% λ 0.23	46%	-23.7

Hammer (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
MPC ↗	62% · +4.3% n386	92% · +12.3% n25	+30.3 pts	83% · +10.1% λ 0.71	79%	+4.5
VLO ↗	58% · +2.5% n401	87% · +11.1% n30	+28.3 pts	80% · +8.9% λ 0.75	82%	-1.9
TRGP ↗	63% · +3.9% n468	91% · +4.0% n11	+27.7 pts	78% · +4.0% λ 0.52	63%	+14.3
PSX ↗	64% · +3.6% n321	79% · +6.6% n24	+15.0 pts	75% · +5.7% λ 0.71	73%	+1.7
FANG ↗	65% · +4.5% n425	78% · +2.1% n9	+13.1 pts	71% · +3.4% λ 0.47	58%	+13.3
EPD ↗	62% · +1.9% n344	75% · +2.1% n8	+12.8 pts	68% · +2.0% λ 0.44	56%	+11.4
NG=F ↗	43% · -1.9% n136	52% · +0.7% n21	+9.8 pts	49% · -0.2% λ 0.68	45%	+4.5
CL=F ↗	58% · +2.2% n69	68% · +4.3% n31	+9.7 pts	65% · +3.8% λ 0.76	60%	+4.8
SLB ↗	55% · +0.8% n397	64% · +3.3% n25	+8.8 pts	62% · +2.5% λ 0.71	57%	+4.7
OKE ↗	62% · +1.8% n331	70% · +2.2% n20	+7.8 pts	67% · +2.1% λ 0.67	61%	+6.3
HO=F ↗	64% · +4.0% n66	69% · +7.2% n32	+5.2 pts	68% · +6.5% λ 0.76	66%	+2.1
XOM ↗	62% · +2.0% n306	65% · +3.6% n23	+3.1 pts	64% · +3.1% λ 0.70	59%	+5.2
RB=F ↗	62% · +2.8% n74	65% · +8.0% n37	+2.7 pts	64% · +6.9% λ 0.79	70%	-5.9
BKR ↗	53% · +0.6% n443	55% · +1.2% n20	+2.0 pts	54% · +1.0% λ 0.67	48%	+6.4
CVX ↗	62% · +2.0% n322	60% · +1.9% n15	-1.5 pts	61% · +2.0% λ 0.60	57%	+3.2
KMI ↗	57% · +1.6% n340	54% · +0.7% n13	-3.6 pts	55% · +1.1% λ 0.57	50%	+5.1
XLE ↗	61% · +2.4% n326	57% · +2.3% n14	-3.6 pts	59% · +2.3% λ 0.58	59%	-0.4
LNG ↗	60% · +2.6% n473	53% · +3.8% n15	-7.0 pts	56% · +3.3% λ 0.60	58%	-1.6
COP ↗	58% · +2.2% n349	50% · +3.1% n12	-8.2 pts	54% · +2.7% λ 0.55	53%	+0.7
WMB ↗	64% · +2.0% n382	50% · +0.4% n12	-13.6 pts	56% · +1.1% λ 0.55	51%	+5.4
EOG ↗	60% · +2.3% n385	44% · -0.3% n9	-15.6 pts	53% · +1.1% λ 0.47	55%	-2.2
OXY ↗	54% · +0.8% n387	35% · -2.8% n17	-19.2 pts	42% · -1.4% λ 0.63	51%	-8.9
DVN ↗	63% · +3.9% n431	20% · -3.8% n10	-42.6 pts	41% · +0.0% λ 0.50	52%	-11.1

Fade the reversal from the high (short) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
NG=F ↗	42% · -0.6% n76	80% · +8.2% n5 thin	+37.9 pts	55% · +2.4% λ 0.33	53%	+1.8
CVX ↗	39% · -1.9% n285	71% · +2.2% n7 thin	+32.8 pts	52% · -0.2% λ 0.41	43%	+9.5
XOM ↗	39% · -1.4% n291	71% · +1.0% n7 thin	+32.6 pts	52% · -0.4% λ 0.41	42%	+10.3
XLE ↗	42% · -1.2% n254	67% · +2.9% n9	+24.6 pts	54% · +0.7% λ 0.47	41%	+12.4
WMB ↗	42% · -1.4% n285	65% · +0.8% n17	+22.2 pts	56% · -0.0% λ 0.63	50%	+6.3
EPD ↗	39% · -0.9% n284	55% · +0.3% n20	+15.6 pts	50% · -0.1% λ 0.67	44%	+5.7
BKR ↗	46% · -1.2% n304	58% · +4.1% n24	+11.9 pts	55% · +2.5% λ 0.71	52%	+2.6
KMI ↗	44% · -1.0% n296	53% · +0.1% n15	+9.4 pts	50% · -0.4% λ 0.60	49%	+0.1
COP ↗	42% · -1.4% n332	50% · -5.1% n6 thin	+7.8 pts	45% · -2.8% λ 0.38	46%	-1.2
EOG ↗	43% · -1.7% n305	50% · -0.5% n6 thin	+7.4 pts	45% · -1.2% λ 0.38	45%	+0.0
CL=F ↗	48% · -0.3% n64	50% · -4.0% n4 thin	+1.6 pts	49% · -1.3% λ 0.29	45%	+3.9
RB=F ↗	49% · -0.7% n41	50% · +1.6% n2 thin	+1.2 pts	49% · -0.3% λ 0.17	38%	+11.2
DVN ↗	44% · -2.3% n328	43% · -1.6% n7 thin	-1.6 pts	44% · -2.0% λ 0.41	48%	-3.9
FANG ↗	43% · -2.2% n259	40% · -5.6% n5 thin	-2.9 pts	42% · -3.3% λ 0.33	42%	-0.2
OKE ↗	37% · -1.6% n286	33% · -0.8% n12	-3.8 pts	35% · -1.1% λ 0.55	38%	-3.3
OXY ↗	46% · -0.6% n302	40% · -9.2% n10	-5.7 pts	43% · -4.9% λ 0.50	49%	-5.8
SLB ↗	45% · -1.9% n293	39% · -3.0% n18	-5.8 pts	41% · -2.6% λ 0.64	43%	-2.3
PSX ↗	41% · -1.8% n318	33% · -8.0% n9	-7.9 pts	38% · -4.7% λ 0.47	32%	+5.8
TRGP ↗	31% · -3.7% n290	15% · -3.5% n13	-16.0 pts	22% · -3.6% λ 0.57	36%	-14.0
LNG ↗	40% · -2.1% n319	20% · -4.7% n5 thin	-19.8 pts	33% · -2.9% λ 0.33	42%	-8.5
MPC ↗	39% · -2.4% n322	18% · -15.8% n17	-21.5 pts	26% · -10.8% λ 0.63	23%	+2.3
HO=F ↗	46% · -1.3% n83	14% · -3.6% n7 thin	-31.5 pts	33% · -2.2% λ 0.41	40%	-7.3
VLO ↗	41% · -2.0% n294	0% · -14.7% n11	-40.8 pts	19% · -8.7% λ 0.52	25%	-6.0

Rates - bond complex + curves

since 2026-03-01FADING21 of 32 members have test history

The same war / inflation impulse that re-priced energy: yields rose with crude (10y 4.05 -> 5.00), the front end repriced for a hawkish Fed, and the curve FLATTENED. Adopted by Brad 2026-09-18 on judgment - the nightly admission test is shown beside it, unedited.

Canonical setup · pooled across the theme	All history	In the regime	Lift	Edge over regime drift	Breadth ≥+10	Names worse	Admission	Health
Completed reversal from the low (long) ↗	54% · +0.1% n4584	34% · -0.4% n141	-19.6 pts	-1.0 pts drift 35%	0% of 8	8	does not qualify fails: lift, breadth, drift	FADING last 60s: 29% n31 · prior 36%
Hammer (long) ↗	44% · -0.8% n183	25% · -1.4% n4	-18.7 pts	-4.8 pts drift 30%	0% of 0	0	does not qualify fails: events, lift, breadth, drift	thin recent sample last 60s: — n0 · prior 25%
Fade the reversal from the high (short) ↗	42% · -0.2% n3843	61% · +0.2% n200	+19.2 pts	-3.7 pts drift 65%	67% of 9	0	does not qualify fails: drift	regime intact last 60s: 65% n75 · prior 58%

Completed reversal from the low (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
GBM=F ↗	36% · -0.2% n25	67% · +0.2% n3 thin	+30.7 pts	43% · -0.1% λ 0.23	40%	+3.3
GBS=F ↗	30% · -0.1% n20	33% · -0.6% n3 thin	+3.3 pts	31% · -0.2% λ 0.23	39%	-7.8
EXHB.DE ↗	47% · -0.0% n389	50% · -0.2% n6 thin	+3.2 pts	48% · -0.1% λ 0.38	48%	-0.1
BND ↗	54% · +0.1% n263	57% · +0.3% n7 thin	+2.7 pts	56% · +0.2% λ 0.41	51%	+4.5
MUB ↗	69% · +0.5% n318	62% · +0.6% n8	-6.1 pts	66% · +0.5% λ 0.44	53%	+12.9
EXHC.DE ↗	51% · +0.0% n491	44% · -0.1% n9	-6.5 pts	48% · -0.0% λ 0.47	49%	-0.9
MBB ↗	60% · +0.1% n311	50% · -0.1% n10	-9.5 pts	55% · +0.0% λ 0.50	52%	+3.1
TLT ↗	48% · -0.2% n299	33% · -1.7% n9	-14.5 pts	41% · -0.9% λ 0.47	40%	+0.6
ZB=F ↗	40% · -0.6% n43	25% · -1.6% n4 thin	-14.5 pts	35% · -0.9% λ 0.29	38%	-3.0
AGG ↗	55% · +0.1% n281	40% · -0.7% n5 thin	-14.8 pts	50% · -0.1% λ 0.33	53%	-2.9
ZT=F ↗	46% · -0.1% n39	29% · -0.3% n7 thin	-17.6 pts	39% · -0.2% λ 0.41	34%	+4.9
IEF ↗	51% · +0.0% n259	30% · -0.8% n10	-20.6 pts	40% · -0.4% λ 0.50	43%	-2.5
VCSH ↗	64% · +0.2% n351	43% · -0.1% n7 thin	-20.6 pts	55% · +0.1% λ 0.41	63%	-7.6
GOVT ↗	50% · +0.0% n242	25% · -0.7% n8	-25.4 pts	39% · -0.3% λ 0.44	45%	-5.7
EXHD.DE ↗	56% · +0.1% n481	31% · -0.5% n13	-25.5 pts	42% · -0.2% λ 0.57	44%	-2.3
GBL=F ↗	46% · -0.2% n33	20% · -1.4% n5 thin	-25.5 pts	37% · -0.6% λ 0.33	42%	-4.8
EXX6.DE ↗	49% · -0.0% n442	21% · -0.8% n14	-27.7 pts	33% · -0.5% λ 0.58	40%	-6.7
GBX=F ↗	29% · -1.0% n24	0% · -3.2% n4 thin	-29.2 pts	21% · -1.6% λ 0.29	36%	-15.5
ZN=F ↗	43% · -0.1% n37	0% · -0.9% n4 thin	-43.2 pts	31% · -0.4% λ 0.29	39%	-7.9
ZF=F ↗	52% · +0.0% n29	0% · -0.2% n3 thin	-51.7 pts	40% · -0.0% λ 0.23	38%	+1.5
BNDX ↗	55% · +0.1% n207	0% · -0.1% n2 thin	-55.1 pts	46% · +0.1% λ 0.17	59%	-13.1

Hammer (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
GBX=F ↗	43% · -1.9% n21	33% · -1.3% n3 thin	-9.6 pts	41% · -1.8% λ 0.23	37%	+3.7
EXX6.DE ↗	40% · -1.1% n47	0% · -1.6% n1 thin	-40.4 pts	37% · -1.1% λ 0.09	50%	-12.8
TLT ↗	45% · -0.5% n115	— · — n0 thin	—	45% · -0.5% λ 0.00	51%	-6.0

Fade the reversal from the high (short) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
ZT=F ↗	39% · -0.1% n33	100% · +0.7% n5 thin	+60.6 pts	60% · +0.2% λ 0.33	64%	-4.2
MUB ↗	41% · -0.3% n236	89% · +0.3% n9	+47.8 pts	64% · -0.0% λ 0.47	47%	+16.6
ZF=F ↗	54% · +0.1% n37	100% · +0.8% n10	+45.9 pts	77% · +0.5% λ 0.50	68%	+8.9
GBS=F ↗	58% · +0.1% n19	100% · +0.7% n2 thin	+42.1 pts	65% · +0.2% λ 0.17	61%	+3.8
ZN=F ↗	55% · +0.1% n40	90% · +2.0% n10	+35.0 pts	72% · +1.1% λ 0.50	66%	+6.2
BNDX ↗	34% · -0.4% n100	67% · +0.5% n6 thin	+32.7 pts	46% · -0.1% λ 0.38	43%	+3.3
AGG ↗	34% · -0.4% n229	67% · +0.2% n6 thin	+32.6 pts	46% · -0.1% λ 0.38	48%	-1.7
VCSH ↗	29% · -0.3% n253	62% · +0.1% n13	+32.6 pts	47% · -0.1% λ 0.57	38%	+9.1
IEF ↗	40% · -0.3% n215	70% · +0.3% n10	+30.5 pts	55% · -0.0% λ 0.50	57%	-2.3
GBM=F ↗	72% · +0.2% n18	100% · +1.9% n1 thin	+27.8 pts	75% · +0.4% λ 0.09	59%	+15.5
MBB ↗	35% · -0.3% n271	57% · +0.2% n7 thin	+22.0 pts	44% · -0.1% λ 0.41	46%	-1.7
BND ↗	36% · -0.3% n203	57% · +0.0% n7 thin	+21.6 pts	44% · -0.2% λ 0.41	48%	-4.1
EXHD.DE ↗	43% · -0.2% n433	58% · +0.4% n26	+15.0 pts	54% · +0.2% λ 0.72	58%	-4.7
EXX6.DE ↗	47% · -0.3% n455	53% · +0.7% n17	+6.3 pts	51% · +0.3% λ 0.63	61%	-10.8
EXHC.DE ↗	44% · -0.1% n450	50% · -0.0% n24	+6.0 pts	48% · -0.1% λ 0.71	53%	-5.1
TLT ↗	48% · -0.2% n232	50% · +0.1% n6 thin	+1.7 pts	49% · -0.1% λ 0.38	57%	-8.3
EXHB.DE ↗	50% · +0.0% n358	46% · -0.1% n28	-3.9 pts	47% · -0.0% λ 0.74	50%	-2.3
GOVT ↗	39% · -0.2% n191	33% · -0.1% n3 thin	-5.4 pts	38% · -0.2% λ 0.23	50%	-12.1
GBL=F ↗	47% · -0.2% n17	33% · -0.7% n3 thin	-13.8 pts	44% · -0.3% λ 0.23	57%	-13.2
ZB=F ↗	39% · -1.2% n36	25% · -1.2% n4 thin	-13.9 pts	35% · -1.2% λ 0.29	61%	-26.3
GBX=F ↗	53% · +0.0% n17	33% · -0.3% n3 thin	-19.6 pts	48% · -0.0% λ 0.23	63%	-14.5

Memory · DRAM cycle

since 2025-08-01FADING3 of 7 members have test history

The DRAM / NAND up-cycle: AI-driven HBM and enterprise-SSD demand met constrained supply, and pricing power returned to the memory makers.

Canonical setup · pooled across the theme	All history	In the regime	Lift	Edge over regime drift	Breadth ≥+10	Names worse	Admission	Health
Completed reversal from the low (long) ↗	58% · +3.0% n979	86% · +26.6% n72	+27.7 pts	+11.4 pts drift 75%	100% of 3	0	PASS	FADING last 60s: 40% n10 · prior 94%
Hammer (long) ↗	56% · +2.5% n1239	71% · +11.8% n170	+14.4 pts	-4.1 pts drift 75%	100% of 3	0	does not qualify fails: drift	FADING last 60s: 38% n24 · prior 76%
Fade the reversal from the high (short) ↗	42% · -2.4% n850	15% · -23.7% n34	-26.9 pts	-10.5 pts drift 25%	0% of 3	3	does not qualify fails: events, lift, breadth, drift	thin recent sample last 60s: 67% n6 · prior 4%

Completed reversal from the low (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
WDC ↗	58% · +3.1% n324	87% · +26.8% n23	+29.3 pts	78% · +19.6% λ 0.70	68%	+10.0
MU ↗	59% · +2.8% n311	86% · +13.1% n22	+27.6 pts	78% · +9.9% λ 0.69	72%	+6.2
STX ↗	59% · +3.0% n344	85% · +27.5% n27	+26.5 pts	78% · +20.9% λ 0.73	69%	+9.4

Hammer (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
WDC ↗	54% · +1.8% n419	72% · +11.5% n53	+17.8 pts	69% · +10.0% λ 0.84	70%	-1.6
MU ↗	58% · +2.6% n407	72% · +12.5% n58	+14.7 pts	70% · +11.1% λ 0.85	75%	-4.7
STX ↗	57% · +3.1% n413	68% · +10.1% n59	+10.9 pts	66% · +9.1% λ 0.86	71%	-4.5

Fade the reversal from the high (short) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
WDC ↗	44% · -2.1% n312	18% · -26.1% n11	-25.7 pts	30% · -14.7% λ 0.52	35%	-4.4
STX ↗	42% · -2.2% n275	15% · -17.6% n13	-26.4 pts	27% · -10.9% λ 0.57	34%	-6.9
MU ↗	39% · -4.1% n263	10% · -32.0% n10	-28.8 pts	24% · -18.1% λ 0.50	32%	-7.8

Semis · capex chain

since 2025-08-01FADING15 of 42 members have test history

The AI capex cycle broadened from the GPU leaders into the equipment and foundry chain — wafer-fab equipment and the semi ETFs re-rated; the mature analog/handset names did not.

Canonical setup · pooled across the theme	All history	In the regime	Lift	Edge over regime drift	Breadth ≥+10	Names worse	Admission	Health
Completed reversal from the low (long) ↗	59% · +1.8% n5299	68% · +7.2% n293	+9.4 pts	+8.1 pts drift 60%	53% of 15	3	PASS	FADING last 60s: 52% n46 · prior 71%
Hammer (long) ↗	58% · +2.2% n5680	56% · +2.2% n731	-1.9 pts	-4.2 pts drift 60%	13% of 15	5	does not qualify fails: lift, breadth, drift	FADING last 60s: 53% n108 · prior 56%
Fade the reversal from the high (short) ↗	38% · -2.4% n4467	39% · -3.2% n298	+0.1 pts	-1.2 pts drift 40%	20% of 15	7	does not qualify fails: lift, breadth, drift	regime intact last 60s: 52% n44 · prior 36%

Completed reversal from the low (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
SMH ↗	63% · +2.1% n371	86% · +9.8% n21	+22.6 pts	78% · +7.3% λ 0.68	68%	+10.7
LRCX ↗	64% · +3.5% n333	83% · +14.5% n24	+19.3 pts	78% · +11.3% λ 0.71	68%	+10.1
AMAT ↗	56% · +2.2% n333	75% · +9.6% n20	+18.5 pts	69% · +7.1% λ 0.67	68%	+1.0
SOXX ↗	63% · +1.9% n356	80% · +11.5% n20	+17.4 pts	74% · +8.3% λ 0.67	68%	+6.4
INTC ↗	55% · +0.8% n362	69% · +12.2% n16	+14.1 pts	63% · +7.8% λ 0.62	62%	+1.4
QCOM ↗	53% · +0.4% n342	67% · +3.5% n18	+14.1 pts	62% · +2.4% λ 0.64	52%	+10.0
KLAC ↗	60% · +2.4% n381	73% · +4.7% n30	+13.2 pts	70% · +4.1% λ 0.75	66%	+3.5
MRVL ↗	55% · +1.0% n337	65% · +6.9% n20	+10.1 pts	62% · +4.9% λ 0.67	57%	+4.2
TXN ↗	58% · +1.2% n369	67% · +4.8% n15	+8.7 pts	63% · +3.4% λ 0.60	55%	+8.0
AMD ↗	58% · +3.7% n304	67% · +15.0% n24	+8.5 pts	64% · +11.7% λ 0.71	56%	+7.9
ADI ↗	55% · +0.7% n378	59% · +2.1% n22	+3.8 pts	58% · +1.7% λ 0.69	58%	+0.2
NVDA ↗	63% · +3.8% n358	63% · +3.7% n19	+0.6 pts	63% · +3.7% λ 0.66	60%	+3.5
MCHP ↗	53% · +0.6% n370	46% · -1.0% n13	-6.8 pts	49% · -0.3% λ 0.57	52%	-2.4
NXPI ↗	62% · +2.1% n347	53% · +0.0% n19	-9.4 pts	56% · +0.7% λ 0.66	53%	+2.9
AVGO ↗	66% · +2.5% n358	50% · -1.4% n12	-15.6 pts	57% · +0.4% λ 0.55	56%	+0.8

Hammer (long) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
MRVL ↗	56% · +2.5% n431	71% · +11.3% n56	+15.3 pts	69% · +9.9% λ 0.85	58%	+10.9
INTC ↗	56% · +1.3% n349	67% · +7.2% n69	+10.8 pts	65% · +6.5% λ 0.87	64%	+0.9
KLAC ↗	62% · +3.1% n387	69% · +3.5% n48	+7.0 pts	68% · +3.5% λ 0.83	67%	+0.5
TXN ↗	55% · +1.2% n331	58% · +3.3% n48	+3.0 pts	58% · +3.0% λ 0.83	54%	+3.9
AMAT ↗	60% · +3.2% n376	62% · +7.7% n50	+2.2 pts	62% · +6.9% λ 0.83	70%	-8.5
LRCX ↗	58% · +2.2% n379	58% · +8.2% n57	-0.4 pts	58% · +7.3% λ 0.85	69%	-11.1
SOXX ↗	63% · +3.2% n365	60% · +4.2% n50	-3.0 pts	60% · +4.0% λ 0.83	69%	-8.6
NVDA ↗	61% · +3.6% n419	57% · +0.6% n28	-3.8 pts	58% · +1.4% λ 0.74	59%	-1.3
SMH ↗	63% · +3.2% n362	59% · +2.9% n46	-4.3 pts	60% · +2.9% λ 0.82	68%	-8.9
ADI ↗	60% · +2.2% n363	55% · +2.1% n49	-4.7 pts	56% · +2.1% λ 0.83	58%	-1.8
AMD ↗	54% · +1.6% n422	48% · -0.5% n46	-6.2 pts	49% · -0.1% λ 0.82	57%	-7.8
QCOM ↗	52% · +0.9% n349	46% · -4.8% n50	-6.4 pts	47% · -3.8% λ 0.83	51%	-3.8
NXPI ↗	54% · +1.5% n397	44% · -3.3% n36	-9.8 pts	46% · -2.3% λ 0.78	52%	-5.2
MCHP ↗	52% · +1.1% n373	34% · -3.4% n55	-18.0 pts	37% · -2.7% λ 0.85	50%	-12.9
AVGO ↗	62% · +2.4% n377	42% · -1.9% n43	-19.6 pts	46% · -1.0% λ 0.81	53%	-7.3

Fade the reversal from the high (short) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
TXN ↗	37% · -2.2% n308	61% · +2.1% n18	+24.1 pts	52% · +0.6% λ 0.64	45%	+7.5
MCHP ↗	46% · -0.8% n322	68% · +3.5% n28	+22.2 pts	62% · +2.4% λ 0.74	49%	+13.0
QCOM ↗	44% · -1.1% n307	65% · +5.9% n23	+21.2 pts	59% · +3.8% λ 0.70	48%	+10.4
NXPI ↗	38% · -2.8% n322	39% · -1.4% n23	+1.5 pts	39% · -1.8% λ 0.70	48%	-8.9
NVDA ↗	33% · -3.5% n302	35% · -3.4% n26	+1.2 pts	34% · -3.4% λ 0.72	39%	-4.6
AMD ↗	46% · -1.9% n321	43% · -4.1% n14	-2.6 pts	44% · -3.2% λ 0.58	44%	+0.1
AMAT ↗	38% · -3.8% n271	36% · -12.8% n14	-2.7 pts	37% · -9.1% λ 0.58	33%	+3.4
AVGO ↗	31% · -2.8% n289	26% · -1.5% n19	-4.8 pts	28% · -1.9% λ 0.66	45%	-17.0
MRVL ↗	43% · -2.0% n314	37% · -4.3% n19	-6.2 pts	39% · -3.5% λ 0.66	42%	-3.6
SMH ↗	33% · -3.8% n282	27% · -9.0% n15	-6.6 pts	29% · -6.9% λ 0.60	33%	-3.3
ADI ↗	39% · -1.9% n300	32% · -5.8% n25	-6.7 pts	34% · -4.7% λ 0.71	42%	-8.4
SOXX ↗	34% · -3.6% n278	27% · -9.4% n15	-7.1 pts	30% · -7.1% λ 0.60	33%	-3.2
INTC ↗	38% · -2.4% n311	26% · -21.8% n27	-12.0 pts	29% · -16.5% λ 0.73	37%	-7.8
LRCX ↗	39% · -2.7% n268	20% · -7.4% n15	-18.8 pts	28% · -5.5% λ 0.60	34%	-6.1
KLAC ↗	38% · -2.4% n272	18% · -13.3% n17	-20.3 pts	25% · -9.3% λ 0.63	34%	-9.1

Rates · curve spreads 2s30s / 2s10s / 5s30s / 10s30s, US + Bund (adopted 2026-09-18)

since 2026-03-01thin recent sample4 of 8 members have test history

Adopted by Brad 2026-09-18 with the rates regime. Each spread is a DURATION-NEUTRAL curve index (rising = steepening; +1.0% = 10 bp of steepening), not a price ratio — a 2y/30y price ratio is just a 30-year direction trade. Long = a DV01-neutral steepener. Close-only series, so the setups are close-based exhaustion reads. Bund curves run on the ETF proxies; any curve using the 2-year German proxy is flagged WEAK (its proxy tracks the Schatz future at only ~0.6) and is shown but kept OUT of the pooled admission test. The four spreads share legs, so pooled events are far from independent.

Canonical setup · pooled across the theme	All history	In the regime	Lift	Edge over regime drift	Breadth ≥+10	Names worse	Admission	Health
Flattening washout — 21-session low close, 5-session move in its worst decile (long the steepener) ↗	61% · +0.3% n193	82% · +0.7% n11	+20.7 pts	+26.4 pts drift 55%	0% of 0	0	does not qualify fails: events, breadth	thin recent sample last 60s: 100% n4 · prior 71%
Steepening blow-off — 21-session high close, 5-session move in its best decile (short — flattener) ↗	54% · +0.2% n158	100% · +0.8% n6	+46.2 pts	+55.7 pts drift 44%	0% of 0	0	does not qualify fails: events, breadth	thin recent sample last 60s: 100% n2 · prior 100%

Flattening washout — 21-session low close, 5-session move in its worst decile (long the steepener) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
DE_2s10s ↗	41% · -0.3% n103	83% · +1.1% n6 thin weak proxy	+42.5 pts	57% · +0.2% λ 0.38	53%	+3.5
DE_2s30s ↗	47% · -0.1% n96	83% · +1.4% n6 thin weak proxy	+36.4 pts	60% · +0.5% λ 0.38	55%	+5.4
DE_5s30s ↗	56% · +0.3% n95	83% · +1.2% n6 thin	+27.5 pts	66% · +0.6% λ 0.38	51%	+14.7
DE_10s30s ↗	66% · +0.3% n98	80% · +0.4% n5 thin	+13.7 pts	71% · +0.3% λ 0.33	55%	+15.9

Steepening blow-off — 21-session high close, 5-session move in its best decile (short — flattener) — evidence by name

Name	All history	In the regime	Regime lift	BLENDED (default)	Blended base	Edge over base
DE_10s30s ↗	47% · -0.1% n77	100% · +0.2% n3 thin	+53.2 pts	59% · -0.0% λ 0.23	45%	+14.2
DE_5s30s ↗	60% · +0.5% n81	100% · +1.3% n3 thin	+39.5 pts	70% · +0.7% λ 0.23	50%	+20.1
DE_2s10s ↗	46% · -0.1% n78	20% · -0.2% n5 thin weak proxy	-26.2 pts	38% · -0.1% λ 0.33	48%	-10.6
DE_2s30s ↗	59% · +0.3% n86	25% · -0.1% n4 thin weak proxy	-34.3 pts	50% · +0.2% λ 0.29	47%	+2.6

Regime shift watch · semis & memory — buy the spike or buy the dip?

After a stock prints one of its 5 biggest up-days of the past year (see Top-5 Days), two entries compete. Spike: buy the signal close, hold 10 sessions. Dip: wait 5 sessions and buy only if it has pulled back below the signal close, hold to day 21. Each is scored against the same stocks' any-day base over the same holding length, on a rolling 182-day window, counting only trades whose outcome was known by that date. A style "leads" when it beats the other by ≥ 5 points. Display-only: it flags, it does not change sizing or the bucket.

Now: MIXED — neither style leads

as of 2026-10-28Testing bucket entry in step (enters at the signal)

-2.0 ptsspike edge · 128 trades · median -0.25% vs a 51% base

-12.0 ptsdip edge · 67 trades · median -3.58% vs a 51% base

Buy The Spikelast decisive style · Last flip 2026-01-16: buy the dip → buy the spike.

Green line above blue = momentum is paying (enter on the spike). Blue above green = pullbacks are being bought (wait for the dip). The flip is the signal to change the entry style of top-5-day trades; either line near zero means the setup is only matching drift.

Regime 2 · the analog regime — Gulf War oil shock, 1990–1992

A theme regime (Regime 1, above) re-weights evidence for a group. An analog regime is a different object: a past window that looked like now, reduced to a cross-asset signature — how crude's moves lined up with the 10-year, the 2-year and the four steepeners — then tested leg by leg against today. It is a map, never an estimator: nothing here touches ranking or sizing. A leg matches when today's weekly correlation has the same sign and sits within 0.30 of the analog's. FRED daily data through 2026-10-27 (spot WTI, constant-maturity yields).

The chart — today against 1990–92, aligned at day 0

brown = 1990–92 · blue = now · bands = the three 1990 phases

Signature — correlation of crude's return with each rates leg

weekly (5-session sums) · daily in grey

Window	CL ~ 10y yield ↗	CL ~ 2y yield ↗	CL ~ 2s10s ↗	CL ~ 2s30s ↗	CL ~ 5s30s ↗	CL ~ 10s30s ↗
a. Shock 1990-08 → 1990-10 · n63	+0.83 +0.61	+0.49 +0.39	+0.82 +0.45	+0.78 +0.41	+0.72 +0.37	-0.01 +0.06
b. Unwind 1990-11 → 1991-02 · n80	+0.26 +0.43	-0.10 +0.05	+0.63 +0.47	+0.62 +0.52	+0.39 +0.49	+0.17 +0.34
c. Aftermath 1991-03 → 1992-12 · n449	+0.10 +0.07	+0.05 +0.11	+0.04 -0.08	+0.06 -0.09	+0.04 -0.11	+0.05 -0.04
NOW — war regime to date since 2026-03-01 · n166	+0.49 +0.40	+0.51 +0.37	-0.09 -0.01	-0.21 -0.13	-0.33 -0.24	-0.35 -0.28
NOW — last 63 sessions n65	+0.31 +0.28	+0.45 +0.24	-0.16 +0.05	-0.31 +0.02	-0.45 -0.03	-0.60 -0.02
All history 1986–2026 n9430	+0.16 +0.13	+0.09 +0.07	+0.12 +0.10	+0.11 +0.09	+0.07 +0.05	+0.02 +0.02
Pre-Gulf 1988 – Jul 1990 n629	+0.10 +0.03	+0.08 +0.00	-0.00 +0.04	-0.00 +0.02	-0.08 -0.02	+0.01 -0.03

What the table says. The 1990 shock had a signature that exists nowhere else in 40 years: crude ~ 10-year at +0.83 and crude ~ 2s10s at +0.82 — oil up meant the long end sold off while the front end held, so the curve bear-steepened with crude. Before the invasion the same correlations were zero. Today the 10-year leg matches (+0.49 vs +0.83): crude is driving the long end just as it did in 1990. The steepener legs do not (0 of 4): this time crude is also driving the 2-year (+0.51 vs +0.49 then), so oil up has meant bear-flattening. The difference is the Fed: in 1990 the market priced easing into a recession; today it prices a hawkish response.

Signature match vs the 1990 SHOCK phase — leg by leg

Leg	1990 shock	Now (regime to date)	Now (last 63 sessions)	Verdict
CL ~ 10y yield ↗	+0.83	+0.49	+0.31	DOES NOT MATCH
CL ~ 2y yield ↗	+0.49	+0.51	+0.45	MATCH
CL ~ 2s10s ↗	+0.82	-0.09	-0.16	DOES NOT MATCH
CL ~ 2s30s ↗	+0.78	-0.21	-0.31	DOES NOT MATCH
CL ~ 5s30s ↗	+0.72	-0.33	-0.45	DOES NOT MATCH
CL ~ 10s30s ↗	-0.01	-0.35	-0.60	DOES NOT MATCH

The path — what the three phases did, and where we are

Phase	Crude	10-year	2-year	2s10s	2s30s
a. Shock	21.6 → 35.3 max 41.1	8.29 → 8.65	7.85 → 7.77	+44 → +88 bp	+51 → +101 bp
b. Unwind	35.3 → 19.3 max 35.6	8.57 → 8.02	7.68 → 7.04	+89 → +98 bp	+102 → +115 bp
c. Aftermath	19.4 → 19.5 max 24.1	8.12 → 6.70	7.18 → 4.56	+94 → +214 bp	+110 → +284 bp
NOW since 2026-03-01	71.1 → 103.4 max 114.6	4.05 → 4.99	3.47 → 4.70	+58 → +29 bp	+123 → +60 bp

a. Shock: Iraq invades Kuwait (Aug 2). Crude doubles; the long end sells off on inflation while the front end holds - the curve BEAR-STEEPENS with oil. b. Unwind: Crude collapses into and through Desert Storm (Jan 17). Yields fall with oil; curve moves still track crude. c. Aftermath: Oil is flat and irrelevant. Recession + credit crunch: the Fed cuts 8% -> 3% and the curve steepens 120bp on POLICY, not on crude.

What happened next — measured from the 1990 crude peak (1990-10-11)

After the peak	Crude	10-year	2-year	2s10s	2s30s
+1 month	-17%	-43 bp	-38 bp	-5 bp	-7 bp
+3 months	-33%	-76 bp	-85 bp	+9 bp	+14 bp
+6 months	-49%	-83 bp	-104 bp	+21 bp	+23 bp
+12 months	-44%	-143 bp	-208 bp	+65 bp	+90 bp
+24 months	-46%	-240 bp	-399 bp	+159 bp	+243 bp

The lesson of the analog is sequence, not correlation. The great steepener of 1990–92 (2s10s +160 bp in two years) did not come while crude was rising, and in the aftermath it had no correlation to crude at all — it came after the crude peak, led by the 2-year collapsing as the Fed eased. While crude was the driver, the trade was the 10-year, not the curve. One analog is one path: n = 1. This section exists to keep the map in front of us and to say, nightly and unedited, which legs today actually match it.

Proposed regimes — tested, not adopted

A candidate has to clear the same admission test before it governs anything. Nothing here affects ranking or sizing.

The bench is empty — rates and the rates curves were adopted on 2026-09-18. A new candidate lands here first and has to face the same admission test.

Governance

Where it applies. Daily Testing (equities and commodities), Weekly Ideas, Weekly Spreads, Completed Reversals and its feed, and the Ensemble (signal and base rate). Gap-Open and Reversal Band use a hard regime cut by design.
Theme ≠ sector. The evidence is per name. Energy's regime is refiners and products, not the integrateds; semis' is the equipment chain and the ETFs, not the mature analog and handset names. The blend applies per name, so a theme member with no regime effect simply shows none.
Precedence. When a ratio spans two themes the most recent regime governs.
Both directions. A regime that lifts longs usually punishes the opposite trade — the short-side table is part of the evidence, not an afterthought.
Retirement. A regime flagged fading for ten consecutive sessions is reviewed; retiring it means deleting one line — every study falls back to all-history automatically.
What it never does. It never raises the single-name cap, never overrides the requirement that forward odds confirm the direction, and never turns a thin sample into conviction.

Evidence recomputed nightly from each name's own daily history (rc_hist). Setups: completed reversal from the low = low below the prior close, close in the top 10% of the range, net positive; hammer = above-median range, low ≥1% below the prior close, close in the top 40%; fade-the-high = the mirror, read as a short. Forward window 21 sessions. Educational research, not advice.