hotfix15
BASELINE / IMMUTABLE- Fresh hedge
- freeze grid
- Stale hedge fallback
- none
- Structure context
- not consumed
- Recovery lot ceiling
- not explicit
- Feb 23 L2 lifetime
- ~38h 53m
XAUUSD.s · M5 · Feb 16–28 stress replay
Escape the dead zone.
ALAMO v3 keeps hotfix15’s hedge backbone, then adds a bounded recovery controller. The result is more activity and a shorter stale hold—with one unresolved cost: equity drawdown still needs calibration.
MECHANISM: PROVEN
CAP 6.25 RUN: COMPLETE
STRUCTURE ORDERS: 0
PROMOTION: HOLD
CURRENT TEST CAP: 6.25 LOT
01 / MACHINE DELTA
Same engine. Different recovery discipline.
v3 does not introduce a second trade strategy. It wraps existing ALAMO, stop-grid and Skill 6 paths with explicit state, age and exposure permissions.
ALAMO v3
POLICY / OPT-IN WRAPPER- Fresh hedge
- protected budget
- Stale hedge fallback
- deterministic
- Structure context
- advisory veto
- Recovery lot ceiling
- cap ∧ hedge volume
- Feb 23 stale recovery
- < 1 hour
| Observed metric | hotfix15 | v3 cap-8 replay | Delta |
|---|---|---|---|
| Final balance | $149,189.27 | $176,128.01 | +$26,938.74 |
| Canonical fill events | 5,620 | 6,810 | +1,190 |
| Live-hedge inactivity | ~2,333 min | 43 min | −98.2% |
| Max equity / logged DD | 20.15% | 22.70% | +2.55 pp |
| Structure-originated orders | 0 | 0 | unchanged |
The v3 balance and DD values come from the completed cap-8 candidate replay. The active test default is now 6.25 lots and remains a calibration candidate until a full report passes the DD gate.
02 / POLICY INPUTS
Seven controls. One bounded recovery contract.
The values in MT5 define when v3 may recover, how much exposure it may add, and when risk overrides structure. The cap-6.25 default has completed Feb and Mar log runs, but remains research—not a live-trading guarantee.
AlamoV3PolicyEnabled
true means v3 is active.
This enables the state-based recovery wrapper. Set it to false only for hotfix15 parity tests; false deliberately restores the legacy freeze behavior.
Read the boolean literally: ON = v3 policy. The old “Off preserves hotfix15” label described the false case and was easy to misread; the EA input label is now explicit.
StaleBudgetSec1,800sProtect the first 30 minutes. A new hedge gets time to stabilize the basket, preventing instant reversal and whipsaw. After 30 minutes, v3 may consider bounded recovery.
MaxStructureHoldSec10,800sStructure may delay—not imprison. Alignment can keep the hedge protected for at most three hours. This is the absolute escape bound against another ~38h 53m dead zone.
RecoveryCapLots6.25Maximum, never a target. New recovery exposure is clipped to min(6.25, live hedge lots). It sits between cap 5’s long pause and cap 8’s higher 22.70% DD.
UnwindFloor−$100Signal cannot realize any loss it wants. A reversal may authorize an unwind only when basket economics are at least −$100, preventing a structure signal from closing a deeply negative basket.
MaxDrawdownPct20%Stop expanding risk. At the policy ceiling, v3 stops normal recovery permission and routes toward safe exit. The cap-8 replay still logged 22.70%, so execution-path validation remains mandatory.
MinMarginLevelPct150%Capital beats signal. Below this margin floor, v3 must not add recovery exposure—even if age and structure would otherwise permit it.
Why 6.25 now? The completed Feb log improved final balance to $178,452.63 and reduced logged DD versus cap 8, but still reached 21.01%. The preset is evidenced, not promoted; the HTML tester report must still confirm the exact max-equity DD.
03 / FAILURE WINDOW
What changed between Friday night and Tuesday.
Weekend closure is excluded. The comparison focuses on market-open behavior while an ALAMO hedge is live.
FEB 20 · 23:30–23:53
Normal grid activity
Sell-side grid levels and fills continue before the weekend close.
FEB 23 · 01:06
ALAMO hedge fires
Both versions protect the stressed basket. Fresh-hedge freeze remains intact.
FEB 23 · 01:40+
v3 opens bounded recovery
Opposite-side recovery lots are clipped to the configured cap and live hedge volume.
HOTFIX15 · FEB 24 15:59
Legacy hedge finally closes
hotfix15 holds through the long dead zone; v3 has already returned to normal cycles.
04 / PAIN → RESOLVE
Every new rule closes a specific failure mode.
v3 is intentionally narrow: preserve the profitable engine, remove indefinite states, and make every recovery decision auditable.
01
38h hedge freezeGrid and Skill 6 stop while a live ALAMO hedge never resolves.
→
Age budget + absolute boundFresh hedge stays protected; stale hedge enters capped recovery; structure cannot hold past three hours.
02
Blind signal closeA reversal can be technically valid while basket P&L is economically unacceptable.
→
Economic unwind floorDirection is advisory; closure still requires basket net P&L ≥ −$100.
03
Oversized recovery requestA requested lot above the remaining allowance can block progress or exceed intent.
→
Clip, account, re-checkEvery order path uses the remaining gross allowance and never exceeds the live hedge volume.
04
Structure lifecycle noiseRepaint, duplicate alerts and ambiguous terminal outcomes make evidence unreliable.
→
Closed-bar stable IDsM5/M15 snapshots are deduplicated and resolve to explicit outcomes; they never own a trade.
05
Partial Skill 6 pairOne side may open while its counterpart fails, creating accidental directional exposure.
→
Atomic pair or rollbackThe paired cycle completes as designed or cleans up the partial leg.
05 / COST LEDGER
What v3 buys—and what it costs.
The stale-hold fix is real. So are its costs. These are release constraints, not footnotes.
BALANCED VERDICT
Better recovery.
Higher complexity.
v3 trades an indefinite freeze for controlled activity. That improves opportunity and final balance in the cap-8 replay, but it also increases exposure to drawdown, churn, execution timing and parameter sensitivity.
Current status: mechanism accepted; parameter set not promoted.
Drawdown regressed
BLOCKERCap 8 logged 22.70% versus hotfix15’s 20.15%. Faster recovery did not mean lower equity stress.
More fills mean more friction
MEASURE6,810 canonical fills versus 5,620 adds spread, commission, slippage and live-execution sensitivity that a tester can understate.
6.25 completed, not promoted
HOLDThe Feb run ended at $178,452.63, but max logged account DD was 21.01% and the longest live hedge still reached 8h 19m.
A 20% gate is not an instant stop
OVERSHOOTOpen-position loss and execution latency can move equity beyond the decision threshold before closure completes.
Fixed thresholds are regime-sensitive
ROBUSTNESS30 minutes, three hours, −$100 and 6.25 lots may behave differently with volatility, balance, symbol contract or broker conditions.
Structure adds logic, not proven alpha
ADVISORYClosed-bar lifecycle and dedup improve reliability, but sampled outcomes do not justify structure-owned orders.
06 / RISK GATE
More motion is not automatically safer.
The cap-6.25 Feb run is complete. It improves balance and beats cap 8’s logged DD, but 21.01% still exceeds the confirmed 20.15% baseline report.
Drawdown comparison
Percentage of account equity decline observed in the fixed Feb 16–28 stress window.
promotion ceiling · 20.15%
07 / STATE FLOW
Time opens the gate. Risk decides the lane.
The controller evaluates hedge age, structure, gross recovery exposure, basket economics, drawdown and margin on every cycle. Existing ALAMO, Grid and Skill 6 code remains the execution owner.
NO LIVE HEDGENORMAL
AGE ≤ 30 MINHEDGE FRESH
STALE / ALIGNEDBOUNDED HOLD
HEDGE CLOSEDNORMAL
RECOVERY CAPPEDAfter freshness expires, opposite-side recovery may run within remaining cap and hedge volume.
UNWIND READYReversal plus basket net ≥ −$100 permits the existing close owner to unwind.
FORCED SAFE EXITBackbone missing, DD ≥ 20%, margin < 150%, or hold ≥ 3h overrides structure.
08 / CONTROL PATH
Structure advises. Existing modules execute.
No structure signal owns a trade, magic number or P&L path. That boundary is deliberate.
CLOSED BARSM5 / M15 structure snapshot
PURE DECISIONALAMO v3 risk policy
EXISTING OWNERSALAMO · Grid · Skill 6
Hard boundary: the structure module contains no order send, market buy/sell, position close or independent trade ownership calls.
Why not let structure trade? The sampled detector outcomes did not demonstrate reliable independent edge: many setups expired and zero structure-originated orders were required to solve the dead zone. Until outcome evidence improves, structure remains a veto/accelerator—not a strategy.
09 / LOG EVIDENCE
Two completed runs. One clearer release decision.
The June 20–21 agent logs contain ten test starts. Only the completed cap-6.25 runs below are promoted into this evidence layer; interrupted and differently configured runs are excluded.
RUN 09 · FEB 16–28
Primary stress replay
Final balance$178,452.63+$29,263.36 vs hotfix15
Max logged account DD21.01%+0.86 pp vs baseline report
Max live-hedge age8h 19m 32srecovery cap reached
Recovery / exposure4 allows · 6.25max exposure equals configured cap
RUN 10 · MAR 16–31
Out-of-window replay
Final balance$237,937.63separate March window
Max logged account DD10.43%ALAMO fire telemetry peak
Max live-hedge age4h 22m 04scap reached; no forced exit logged
Recovery / exposure3 allows · 6.25max exposure equals configured cap
FEB 23 · POLICY AND CAP ENFORCEMENT
01:06:28Hedge fires
→
01:37–01:39Aligned hold
→
01:40:00Recovery opens
→
01:50:412.64 → 2.06 lot
At 01:51 the logged exposure is exactly 6.25 lots. The next policy state is RECOVERY_CAP_REACHED, proving that the shared cap clips the final request instead of overshooting.
Structure evidence says “advisory only.”
Across both completed cap-6.25 runs: 154 setups, 127 expired (82.5%), 11 TP (7.1%), 13 ambiguous losses, 2 timeouts and 1 SL. All 3 reversal setups expired; none reached TP. The detector lifecycle works, but the sample does not support structure-owned orders.
Operational log finding: Skill 6 correctly blocked the second leg when its market buy failed, but emitted 330 BUY_FAILED blocks in February and 823 in March while the market was closed. Pair integrity is preserved; retry throttling and log deduplication remain improvement work.
Source boundary: run 09 starts at log line 1,840,521 and completes at 2,145,935 on June 20. Run 10 starts at 2,145,961 on June 20 and completes at line 299,139 after midnight in the June 21 log.