A3.8 — Failed Expansion Continuation
a3 / a3-failed-expansion-continuation @ a3-v1 · code 128d930
campaign strategyHOLDOUT REJECTED
Holdout: n=100, expectancy=-0.2200R -- did not confirm the prior evidence.
Lifecycle
completed
Outcome
fail
Phase
Phase A
Research gate
HOLDOUT
Engine
strategy lab
Market
MNQ
Timeframe
MNQ 1-minute OHLC
Side
BOTH
Tracking
Family, lineage and what this record needs nextMechanism family
Volatility expansion / contraction
a3-failed-expansion-continuation
Mechanism parent
None declared.
Phase detail
Scientific validation lane (Phase A).
Next required test
None — this record is complete.
Integrity & certification
Signals as recorded; NOT_PASSED means not passed or not applicableEngine certified
unknown
Execution certified
unknown
Integrity audit
not passed
Shadow verifier
unknown
Technical invalidation
None recorded.
Note
Longest observed losing streak 16 trades -- confirm this is tolerable before any further stage.
Note
Holdout expectancy diverges materially from Validation's own -- treat as a fresh, not fully explained, open question.
Artifacts
Everything committed for this recordCampaign record (A3)
research-output/strategy-lab/campaigns/a3/campaign_state.json
Holdout artifacts
research-output/strategy-lab/campaigns/a3
Stage ladder
Each cell carries the campaign runner's own classification label — hover for the verdict and its reasonIntake
passed
Pre-reg
passed
Lightning
passed
Robust
passed
Exec/stat
passed
Validation
passed
H-eligible
passed
Holdout
failed
Audit
inconclusive
Closed
failed
INTAKE
Registered into campaign A3 at code version 128d930.
PREREGISTRATION
Pre-test passed: distinct, causal, executable, required data available.
LIGHTNING
SURVIVED_LIGHTNING_SCREENAggregate expectancy 0.0635R (PF=1.099) over 3166 resolved trades, 4/4 qualifying years (n>=15) agree with the aggregate sign, top-10 winner concentration 9.9% (below the 50% concentration flag). Survives the Lightning Screen -- NOT validated.
ROBUSTNESS
ROBUSTNESS_SURVIVEDNo single-year dependence, no severe winner concentration (no preregistered neighborhood declared).
EXECUTION STAT
EXECUTION_STAT_SURVIVEDBootstrap CI excludes zero and the edge survives up to 0.5pt adverse slippage.
VALIDATION
VALIDATION_SURVIVEDValidation expectancy 0.0833R holds sign and reasonable magnitude relative to Discovery's 0.0582R.
HOLDOUT ELIGIBILITY
HOLDOUT_ELIGIBLESurvived Lightning Screen, the Robustness Gauntlet, the Execution+Statistical Gate, and the Validation/Walk-Forward Gate against 2496 Discovery and 670 Validation trades.
HOLDOUT
HOLDOUT_OPENED_ONCEDiscovery: n=2496, expectancy=0.0582R, PF=1.091. Validation: n=670, expectancy=0.0833R, PF=1.131. Holdout (this run, sealed bank immediately after A3.3's consumed slice, stop reason=FORMAL_TARGET_100): n=100, expectancy=-0.2200R, PF=0.703.
FINAL AUDIT
a3-failed-expansion-continuation@a3-v1: economic credibility=true, statistical credibility=true, executable=true, stable=true, distinct=true. 2 unresolved risk(s) flagged.
CLOSED
REJECTEDHoldout: n=100, expectancy=-0.2200R -- did not confirm the prior evidence.
Evidence recorded
Quoted from the campaign record — never recomputed hereScreen n (FIXED_2R)
3,166
Expectancy
+0.064R
Profit factor
1.10
Win rate
36.0%
Total R
201
Max DD
32R
Longest L streak
16
Trades / year
883.7
| Split | Window | n | Expectancy | PF | Max DD |
|---|---|---|---|---|---|
| Discovery | 2022-06-03..2025-03-31 | 2,496 | +0.058R | 1.09 | 32R |
| Validation | 2025-04-01..2025-12-31 | 670 | +0.083R | 1.13 | 30R |
| Holdout | sealed bank slice | 100 | -0.220R | 0.70 | 36R |
Discovery → Validation
Expectancy degradation -0.025R (positive means Validation came in weaker than Discovery).
Bootstrap CI
95% CI +0.019R .. +0.105R · P(expectancy > 0) = 100%
Winner concentration
Top 5 5.0% · top 10 9.9% · top 20 19.9% of total R.
Leave-one-year-out
No sign flip when any single year is removed.
Holdout vs prior evidence
Discovery: n=2496, expectancy=0.0582R, PF=1.091. Validation: n=670, expectancy=0.0833R, PF=1.131. Holdout (this run, sealed bank immediately after A3.3's consumed slice, stop reason=FORMAL_TARGET_100): n=100, expectancy=-0.2200R, PF=0.703.
Read from research-output/strategy-lab/campaigns/a3/campaign_state.json
Holdout position
One-shot, sealed, never retriedEligible
Yes — the campaign recorded Holdout eligibility.
Authorised
Yes, 2026-09-16 17:14Z
Authorised by
AJ <meatham07@gmail.com> -- explicit chat authorization ('AJ explicitly authorizes a one-shot sealed Holdout for A3.8 using the next deterministic untouched chronological Holdout slice after all previously consumed Holdout data, with no block shopping, no retry, no tuning, no rule changes'), trade-count-based deterministic Holdout consumption policy, 2026-09-16
Specification frozen
Fully frozen before the Holdout was opened.
Sealed data consumed
Yes — this Holdout is spent and can never be re-run.
Slice consumed
2020-05-21 07:01Z → 2020-06-26 13:31Z · 35,656 bars · 100 trades · stop reason FORMAL_TARGET_100
Preserved unused
682,907 sealed bars left for later strategies.
Validation data consumed
Yes.
Read from research-output/strategy-lab/campaigns/a3/A3_8_HOLDOUT_LEDGER.json · research-output/strategy-lab/campaigns/a3/A3_8_HOLDOUT_RESULT_REPORT.json
Lineage
Declared at intake, carrying any prior negative resultRelation to prior work
new independent mechanism
Frozen specification
Snapshotted at intake — not re-read from a live moduleMechanism hypothesis
A directional range expansion that is followed by a genuinely attempted continuation (a new extreme beyond the expansion bar), which then fails and gives back through the expansion bar's own close, predicts a reversal -- distinct from fading the expansion bar itself directly.
Long rule
Stage 1 (expansion): a closed 5-minute bar's range >= 2x ATR(14, computed over bars strictly before it), closing above its own open (up-expansion). Stage 2 (attempted continuation): within the next 3 closed 5-minute bars, a bar makes a new high beyond the expansion bar's own high. Stage 3 (failure): within the next 3 closed 5-minute bars after that, a bar closes back BELOW the expansion bar's own close. Stage 4: enter SHORT (reversal) at the next 1-minute bar's open.
Short rule
Symmetric for a down-expansion (closing below its own open): attempted continuation is a new low beyond the expansion bar's own low; failure is a bar closing back ABOVE the expansion bar's own close; enters LONG.
Entry trigger
A four-stage sequence: (1) a >=2x-ATR directional expansion bar, (2) within 3 bars a confirmed new extreme beyond it (attempted continuation), (3) within a further 3 bars a close back through the expansion bar's own close (confirmed failure). If either window expires without its condition, the sequence is abandoned with no trade.
Earliest executable entry
Next 1-minute bar's open after the failure-confirming 5-minute bar closes.
Initial stop
The extreme reached during the attempted-continuation stage (structural): that stage's own new high (for a short entry) or new low (for a long entry).
Exit
FIXED_2R only (this seed strategy's declared A3 scope), via the shared TradeTracker; 5-day max-hold expiry as a backstop.
Causality
The ATR baseline for the expansion stage uses only bars strictly before the expansion bar. Each subsequent stage (continuation-attempt, failure) is evaluated only from bars strictly after the prior stage's own confirming bar, in chronological order -- no stage ever looks ahead of its own confirming bar. Entry is deferred to the next 1-minute bar's open after the failure-confirming bar closes. No same-bar entry.
Expected frequency
LOW
Required data
MNQ 1-minute OHLC
Campaign data window
MNQ 2022-06-03..2026-09-15
Parameters
expansionAtrMultiple = 2 (number) — Minimum bar range, as a multiple of the preceding ATR, to qualify as an expansion.continuationWindowBars5m = 3 (integer) — Bars after the expansion bar within which a new extreme must be made.failureWindowBars5m = 3 (integer) — Bars after the continuation-attempt bar within which the failure close must occur.atrLength5m = 14 (integer) — 5-minute ATR lookback used for the expansion-magnitude requirement.
Read from research-output/strategy-lab/campaigns/a3/campaign_state.json
History
Every dated event this record appears in, newest first2026-09-16
Holdout result recorded for A3.8REJECTED
100 trades consumed · stop reason FORMAL_TARGET_100 · 682,907 sealed bars preserved unused · Holdout: n=100, expectancy=-0.2200R -- did not confirm the prior evidence.
research-output/strategy-lab/campaigns/a3/A3_8_HOLDOUT_RESULT_REPORT.json
2026-09-16
One-shot sealed Holdout authorised for a3-failed-expansion-continuation
AJ <meatham07@gmail.com> -- explicit chat authorization ('AJ explicitly authorizes a one-shot sealed Holdout for A3.8 using the next deterministic untouched chronological Holdout slice after all previously consumed Holdout data, with no block shopping, no retry, no tuning, no rule changes'), trade-count-based deterministic Holdout consumption policy, 2026-09-16
research-output/strategy-lab/campaigns/a3/campaign_state.json
2026-09-16
Sealed Holdout slice assigned to A3.8
Evaluated chronologically forward from 2020-05-21T07:01:00.000Z, frozen before evaluation.
research-output/strategy-lab/campaigns/a3/A3_8_HOLDOUT_LEDGER.json
2026-09-16
a3-failed-expansion-continuation@a3-v1 registeredACTIVE_RESEARCH · HOLDOUT
[a3] Holdout-eligible but awaiting explicit human authorization -- automatic routing stops here per spec Sec. A#.6, never opens Holdout on its own.
research-output/strategy-lab/experiment_registry.json