MAINLINE Equal core → selective loser reanchor → exact-excess TP → locked equal core → optional pair reset.

N0 EQUAL CORE
  ↓  completed outside-BB displacement + setup confirmed at a check
R0 REANCHOR INTENT
  ↓  next open: setup + loser side + quantity + risk fully revalidated
R1 RESCUE ACTIVE
  ↓  resting exact-excess TP
R2 EXCESS HARVESTED
  ↓
L0 LOCKED EQUAL CORE
  ↓  optional symmetric pair trim/reset
N0 / L0

NO ACTION remains the main decision. Do not start a new action if the account is already in rescue, in directional handoff, if the setup was missed, or if the minimum-sufficient add fails its risk/horizon cap.

Equal core does not mean risk-free: margin, funding, mark-price rules, slippage and latent orders remain separate risks. It only means first-order directional price delta cancels while quantities are equal.

Preferred low-attention path:
N0 → N1 → R0 → R1 → R2 → L0 → optional P0 → N0/L0

Optional winner-SL branch:
N1 → N2 → D0

Quantities, not order status

After add d and close m, excess = d − m. The cores are equal again only at m=d. TP_PARTIAL, an unfilled TP or an unknown acknowledgement are not neutral waiting.

Every setup requires a completed outside-BB impulse, sufficient displacement, a fresh local extreme and evidence that the impulse is no longer accelerating. A BB touch or an old distant average is not a setup by itself.

Separate three clocks: 1m is the execution timeframe; 5/15/30 min are check intervals; 5m/15m context uses completed aggregated bars only. A family name alone does not identify parameters or check phase. Exact profiles are in chapter 13.

1. At the phone-check close, a setup only creates an intent.
2. At the next 1m open, recheck who is still the loser, setup freshness, drift, impulse invalidation, equal-core state and absence of another rescue.
3. Recompute quantity at the actual fill price. Never blindly execute stale quantity.
4. If the minimum-sufficient add exceeds risk/margin caps → NO ACTION.
5. If a fill occurs, disable any incompatible winner conditional and immediately place the exact-excess TP.

A model fill and a manual click are different events

close[t] → open[t+1] is a causal research execution model. A manual click follows observation, response and actual acknowledgement; it does not guarantee the next open. Never use the new bar’s still-unknown high/low/close at its open.

Before committing the add: recheck orders that may have filled in the meantime. Incompatible winner protection must not coexist with new rescue excess; a rejected intent alone is not a reason to silently delete previously valid protection.

Acknowledgements matter: use actual add fills to compute the average and reducing quantity; confirm the correct leg, quantity, price and active TP. Submitted is not acknowledged. The model idealizes this transition; this page does not prove exchange atomicity.

For equal SOL quantity q on LONG and SHORT:

Current price P cancels. If quantities remain equal, price movement alone does not create new directional pair PnL. If previous reanchors produced S > L, the pair contains a positive locked spread independent of current price.

This is the ideal linear model. Funding, slippage, mark-price liquidation, maintenance margin and execution failure sit outside the invariant.

Add d to the loser at P1, creating new aggregate average A. After a retrace, close only the temporary excess d. The remaining equal cores are:

q LONG  @ P0
q SHORT @ A

Their combined PnL is q(A − P0), independent of current price.

4 → 40 example

4 LONG  @ 100
4 SHORT @ 100
price → 105
add 36 SHORT @ 105
new aggregate SHORT average = 104.50

After a small retrace, close the 36-SOL excess in profit. The remaining 4 LONG @100 + 4 SHORT @104.50 contain roughly $18 of locked pair PnL at any later price.

With equal cores at different averages:

q LONG  @ L
q SHORT @ S, where S > L

closing the same amount h from each side realizes:

This realizes locked profit without predicting the next SOL direction. You can then optionally reopen a fresh equal LONG+SHORT core.

The whole cycle: where does profit actually come from?

Let s=+1 for a LONG rescue and s=−1 for SHORT; E0 is the old loser average, Ew the opposite-core average, P1 the actual add fill and T the later closing price. The prior locked spread is B0=s·q·(Ew−E0).

The add itself creates no profit. At the fill, T=P1, total PnL remains B0. Incremental cycle profit comes from a favorable price move on the added quantity; the new average only allocates it between realization and retained core spread.

The trim transfers 4.50 from open spread into realized PnL; the total is not 31.50. Average-cost accounting differs from a virtual “close only the new lots” ledger, but consistent ledgers agree on total economic PnL.

Derived linear model, excluding costs. Locked spread can also be negative; “locked” means no directional delta, not necessarily profit. A full trim leaves 0+0 FLAT, not an open equal core.

An accounting scenario, not a signal or recommended size. Supports both rescue legs, distinct starting averages and partial TP. All calculations exclude fees, spread/slippage, funding and liquidations.

Required bounds: q>0, all prices>0, 0≤m≤d, 0≤h≤q; trim only after all excess is closed.

IM proxy is not a venue margin or liquidation calculation. Adverse loss covers only remaining net excess under the specified move, not all risks. Zero net delta does not prove safety.

If the old loser gap is D bp and the post-reanchor target requires an adverse-to-favorable move of R bp with small lock g:

A smaller target can make rescue faster but makes the add explode. The objective is the smallest add that gives a reasonably reachable target and passes the risk budget, not the smallest possible retrace.

Risk by unattended horizon

v0.1.2 uses research planning proxies around 110 bp (5 min), 160 bp (15 min), and 285 bp (60 min). These are not guaranteed tail bounds. A longer unattended horizon should generally reduce permitted excess.

Under the idealized 100× interpretation, +3% ROE is roughly a +3 bp favorable price move.

• 20–30 bp: faster rescue, often requires a large add.
• 30–50 bp: smaller add, more waiting.
• 50–100 bp: only when wallet and unattended interval support a longer rescue.

The table is an approximation for g=3 bp. Corrected under its stated equation: D300/R30 → M11.11, add10.11; D500/R50 → M10.64, add9.64. Original BBSH1/BBSH3 values are recorded in the change ledger.

Exact geometry for a specified target

With g as a fraction of the average (3 bp = 0.0003), s=+1 for LONG and s=−1 for SHORT:

A* must lie between old average E0 and fill P1; if the existing book already meets the condition at T, required add is 0. If the target requires an unreachable average or a zero denominator, no finite positive solution exists. Apply lot size and constraints separately, then recheck actual A and achieved retrace.

M≈D/(R−g) is a small-move explanation, not universal order sizing. Do not use it when R≤g or it produces a negative add. 75 bp is the desired retrace in the visible STALL/REVERSAL examples; some fills achieve a different retrace under declared slack. G5/G10/G20 mean initial gross / wallet, not net exposure or wallet ROI.

Here x is net rescue-excess notional as a multiple of wallet and a is adverse movement in bp.

The reentry-to-middle/new-extreme window was short: median 3 minutes, p75 6 minutes, p90 9 minutes.

1. Missed setup = NO CHASE.
2. Act only when a valid setup happens to be visible at the check.
3. After an add, place the exact excess TP immediately.
4. A winner floor can remain conditional only if its trigger leaves an acceptable account state.
5. Sleep state: no rescue excess and no latent order that can create a large naked book.

Three distinct durations

p95 is not a stop or a guarantee. x·a/10,000 gives loss at a specified move, not the worst future loss. A quantile from the original event set does not automatically become the quantile of selectively traded rescues. A TP without a fill does not reduce excess.

A future arena should separately test waiting, time limits, excess reduction and controlled unwind from matched states. This page does not promote a new timeout or stop-loss threshold as a validated rule.

MAINLINE OFF A full winner-only exit converts a neutral book into directional handoff, so it is no longer used as the base profit-lock mechanism.

DEFAULT:
  winner_mode = NONE
  keep winner core open
  seek profit through loser reanchor + exact-excess TP + locked spread

PREFER BEFORE WINNER-ONLY EXIT:
  1. exact-excess TP
  2. symmetric pair trim/reset
  3. partial coordinated winner action (challenger)

FULL WINNER EXIT:
  separate DIRECTIONAL_HANDOFF challenger only,
  with explicit repair/exit plan and its own risk budget.

Collision rule: an intent alone must not silently delete previously valid protection; before an actual rescue fill, incompatible winner conditionals must no longer remain active. Mainline uses winner_mode=none. If a stop has already filled, re-evaluate the add from the new state.

DEFAULT:
  leave the winner alone

ARM / TIGHTEN PROFIT FLOOR:
  only after profit is large enough and BB structure has materially weakened

FULL WINNER EXIT:
  explicit directional-handoff branch,
  preferably after a completed 1m middle-BB close against the old direction
  and when 1m + 5m no longer strongly support the old move

PARTIAL WINNER EXIT:
  challenger with a smaller net-quantity change than full exit; overall safety remains unproven

After adding to the loser at P1, weighted average moves toward market. When retrace reaches the explicit target, close only the excess. Original core stays.

Once quantities are equal again, price delta re-locks. If spread between averages is large enough, realize part of it without directional prediction:

Partial TP and failed transitions

A partial fill reduces only the quantity actually closed. Remaining d−m is still directional; amend the TP to the remainder rather than the original d. Mainline symmetric trim follows confirmed equality. Reduce-only describes a closing constraint, not a fill guarantee.

The symmetric-trim identity assumes both legs fill at the same price. Different fill prices yield h·[(S−L)+(PcloseL−PcloseS)] before costs. If only one side closes, the exit is temporarily non-neutral. Partial fills and unknown acknowledgements require reconciliation, not a “complete” label.

11.1 Edge → middle / edge → opposite edge

Across 236,161 real SOLUSDT 1m bars, the intuition is broadly supported, but more slowly than a moving-band chart can visually suggest. Using static event-time BB levels:

Moving BB levels can move toward price, so a visual “return to middle” is not necessarily a return to the original middle price.

11.2 Flat/mixed channel is more mean-reverting

Use 5m mainly as a veto/size modifier, not as another indicator stack.

11.3 Large 1m candles are often retraced

11.4 SOL impulse-retrace lookup

Corresponding p95 adverse excursion: ~110 bp (H5), 157 bp (H15), 282 bp (H60), 489 bp (H240), and 1093 bp (H24h).

[S1] Originally reported BB statistics from BBSH1, SOLUSDT 1m 1 Jan–14 Jun 2026. This review did not recalculate them from raw bars. Full event denominators, uncertainty intervals and cohort definitions are not included on this page; these tables are descriptive context, not calibrated profit probabilities for an add. An unreached target and a right-censored endpoint are distinct outcomes.

OPTIONAL RESEARCH SPECIALIST A separate MDL×DCC study uncovered a simpler adaptive filter based on price position inside the BB. On P30 it improved the combination of PnL, drawdown and completed equal-core state on the continuous later replay; it was not generally better on P05/P60.

This is a P30 × specific BB geometry × already-studied-history specialist. A manual visual equivalent of the learned filter is not validated, so it is a challenger/veto rather than a mandatory manual rule.

What actually worked in this experiment

BB-position is a simpler ablation without MDL model selection or DCC control. It emerged within the MDL×DCC research programme; its effect is not automatically attributed to the complete MDL×DCC governor. A useful specialist discovery still matters when the simpler component wins.

The table uses 1+1 SOL, W10,000 USDT, P30, phase 0, 1 Apr–14 Jun 2026 at 00:00 UTC: 74 days and one minute. The filter compares five BB-position bins using 1,024 matured five-minute observations against the unconditional baseline. It is not a visual threshold already validated for manual use. [S4]

At phase +10 min the filter makes +69.06 versus baseline +75.64; at +20 min DD reaches 131.07. Among 20 actual random filters, 3 match or exceed +63.68 PnL, while the BB filter has lower DD in that set. This is an interesting combination of properties, not proof of universal PnL dominance.

These are intermediate capacity findings, not final validation. All examples are 1 Apr–31 May 2026, W250k, causal next-open, winner_mode=none, and end equal-core in the shown cell.

The same four cases: time, size and identity

All at W250,000 USDT; 61 days; both OHLC orders match on PnL/DD. Phase is the actual top-level phone_offset_min, not the default offset in the nested policy. [S3]

STALL’s longest rescue is 4 days 2 hours 9 min; REVERSAL G10’s is 19 hours 42 min. These are upper bounds from touched 1m bars. A clean 61-day endpoint does not prove quick rescues. STALL in April: +28,451.44 USDT including −915.87 open PnL and OPEN_RESCUE_TAIL; May starts separately. Preserve an open endpoint as a specialist diagnostic: do not hide it or automatically declare the entire idea a failure.

All 30 detailed exported rows are W250k: ranking absolute PnL is not a wallet frontier. Net maxima are not recommended sizes. June is already-examined chronological diagnostics, not an untouched holdout.

work_unit_id: 5a2608181acb45b41d6839e220f574a55ef6774550b2bab334926ef80958a6a6
capacity_policy_id: 1e39dd50ac387d5165a719092a06391d32456ddec7d5ee14c2c7e3ead17a5722

work_unit_id: 04bb16e636c9867ae88f23438f68970dbaf7d9de86f9453c40c32c897974d587
capacity_policy_id: eff2c29659da2b0386dbee038a1cc8cbdfec61a3526496b5597a5d062ed4c471

work_unit_id: 1a601a5de5a0586082767c211898a0046e14634fd015025feb5b1b695acacf99
capacity_policy_id: af6f89db8e73096be5287f76725c65151a1665dffd31536ad9b995dee0732ca9

work_unit_id: 56f7da206cb465b13c15da66104f10c31ecc5199fbce480db86c3bb0648c6519
capacity_policy_id: bcd9b2e0e5e38438491730250f553d6aeaf83c1cd795f2044966572efa0e6bc8

A. State first

If RESCUE_ACTIVE or DIRECTIONAL_HANDOFF:
  do not start a new unrelated setup.
  Manage existing tail / TP / repair first.

If EQUAL_CORE and no pending orders:
  proceed to the setup gate.

B. Setup gate

NO ACTION if:
- only outer-BB touch or only reentry
- impulse is still accelerating
- no completed large outside-BB displacement
- no fresh extreme + stall/reversal/selective confirmation
- setup was missed
- minimum-sufficient add exceeds risk/margin cap

REANCHOR INTENT if:
- completed displacement is large enough
- fresh extreme exists
- STALL / REVERSAL / SELECTIVE gate is valid
- continuation veto does not require rejection or smaller size

C. At the next open

REVALIDATE EVERYTHING.
If loser/setup/size/risk no longer hold → CANCEL.
If they hold → fill minimum-sufficient quantity → immediately place exact-excess TP.

D. After TP

If quantity is equal again:
  measure locked spread.
  pair trim/reset only when spread is large enough.
  otherwise NO ACTION.

NO ACTION:
- only an outer BB touch
- only a reentry into BB
- 1m and 5m still strongly support the winner direction
- a useful stop would sit inside normal noise
- trigger would leave an oversized naked loser

TIGHTEN / ARM:
- profit is large enough for a structurally positive stop
- 1m channel has flattened/turned
- or completed middle-BB close against winner direction
- resulting directional exposure is acceptable

Prefer partial winner action or neutral pair trim before full winner-only exit.

NO ACTION:
- only a BB touch
- impulse is still accelerating
- setup window was missed
- minimum-sufficient add exceeds wallet/horizon cap
- no new adverse displacement after a prior rescue

ARM, not automatic fill:
- large completed outside-BB displacement
- fresh local extreme / stall / reversal structure is visible
- 1m + 5m do not demand a full-size continuation veto
- candidate size is mathematically useful and wallet-feasible

ON FILL:
- immediately recompute aggregate average
- immediately place partial TP for the exact excess
- target by explicit price, not stale pre-add ROE
- add again only after new adverse displacement + new event

Mathematical identities and current execution choices: equal-core accounting, weighted-average reanchor, next-open revalidation, exact-excess TP, return to equal quantity, pair-reset identity, NO CHASE and NO ACTION as a legal main decision.

Still evolvable: STALL/REVERSAL/SELECTIVE ranking, 30–75+ bp target geometry, optimal core/wallet, adaptive BB-position veto, 5m/15m/30m attention, maximum allowed rescue duration and any future winner/directional repair.

Not yet proven: 8-hour sleep, live venue execution, funding/impact, sealed holdout, durable profitability or an optimal wallet. The separate v0.1.3 research line is in development/testing and may change parts of this policy after new evidence.

The best current interpretation is not an active BB scalping system. It is a low-frequency neutral-lock system.

small equal core buys time and reduces the need to predict direction
+
BB identifies rare displacement opportunities
+
loser reanchor converts part of displacement into a locked average spread
+
excess TP restores equal quantity
+
symmetric pair trim/reset realizes profit without choosing direction
+
winner-only SL remains an optional directional overlay
+
NO ACTION dominates most phone checks

The research supports the individual mechanisms, but does not yet prove that the full manual policy is live-profitable or optimal for a specific wallet. The next synthesis layer is the wallet/core/excess matrix from $100 through $250k+ and concrete scenario cards.

Basis: BD BB Reanchor Lock, BD BB Hybrid Dual-Book, Q1s-DETS, BD Trading Arenas, and fresh BB-only SOLUSDT 1m analysis, 2026-01-01 → 2026-06-14.

Mathematics is directly checked in this revision. Market results come from the sources below; no new trading tests are run here. Working policy choices are not empirical guarantees.

BBSH4 changes: validated calculator, LONG/SHORT and partial fills; conservation ledger; corrected sizing-table rounding; legacy winner instructions separated from mainline; three time horizons; precise evidence and attribution; multi-hit search, mobile contents, deep links and printing closed chapters.

Publication: one standalone file; server filename bbsh.html under /bd/. No external libraries, analytics, market APIs or live orders. Language/font preferences remain in your browser and are not sent anywhere.

noindex,nofollow,noarchive remains. An unlisted link is not access control: anyone with the exact address can open the page. Noindex is an instruction to cooperating search engines, not a password. Actual privacy requires server-side access control. Google must be allowed to read the page to see noindex. [W1]

[W1] Google Search Central — noindex · [W2] MDN — print events

BBSH1.html
ed29dd58eceabe0983353a618c8c396ac6194f708661ad6172de324b48792560

BBSH2.html
7fe107f355af7709fd25c282df15093c729b7fbbe4694efb84555d8a3559f495

BBSH3.html
789fe150506102c453c5cca1ee8e710022703f70759ab4efa6e8433b980eb79f

BD_SOL_V012_LIVE_REVIEW_20260906_SL.md
bce2ad377e6adcf10f6472454ca59af696668a27d61f37466f46721c6ebfa9ee

BOUNDED_RESULT_TAPES.json
d260ad8fb6923c34b9aa89f1dc867752edd626139363b43cd076c867a1da3380

KEY_CONTINUOUS_COMPARISON.csv
928f25d4e2ec43de582ea2716ca97409c3c591f79fff0a2dcb5a4d753a843d13