The last run could not have demonstrated an edge either way, and nothing in the
log said so. Four changes so it does.
1. THE ZERO-SKILL LINE DIVIDED BY THE WRONG DENOMINATOR. m_oosWinLongTotal resets
every era; m_oosSamples only resets on a full model reset. So 'always-long %'
decayed as ~1/era: a run whose true rate is 37% printed 1.2% at era 33 and
0.0% at era 2219. This is the SAME bug already found and fixed for
logBuyPredPct thirty lines above ('era-15 Buy:2% that was really ~30%'), left
in the one line whose whole job is to be the reference every other number is
read against. Correct at era 1, wrong everywhere after - including the '62%
zero-skill' figure in the 2026-08-16 notes. Now per-era, and always-short is
finally readable.
2. THE DEPLOY GATE STATES ITS OWN BAR. 'edge -1pp' era after era cannot separate
'short by a hair' from 'short by an amount no strategy could cover'. The era
line now prints the required win rate, the SE, the effective n and the
lifespan it was deflated by; above 100% it says UNREACHABLE. At 4,738 OOS bars
and L=75.6 there are ~63 independent observations, putting the bar near 66% at
typical coverage.
3. LIFESPAN MEASURED PER RUNG. The first-passage cache already stores touch ages
at every ladder level, so each candidate geometry's resolution time is
readable without training on it - L-vs-width becomes a measurement across the
whole ladder in ONE run rather than a second chart. Each rung reports L,
n_eff, min provable edge and min provable EV.
4. SCALE OBJECTIVE IS PHASE-AWARE, defaulting to MEASURE. Width and detectability
are opposed: labels overlap by L, L grows like m*k = width^2 at fixed ratio,
so min provable EV ~ width^2 while the cost saving from width is only linear.
Doubling width quadruples the smallest EV you can prove. DEPLOY (widest that
clears reachability) is right once an edge is known; MEASURE (narrowest that
keeps round-trip spread under BARRIER_MAX_COST_FRACTION_PCT) is right while it
still has to be shown. The direction does not depend on the exponent, and
item 3 makes the exponent checkable.
Fixed in review: m_lastRungLifespan is cleared on every LadderWinShare entry or a
rejected rung reports the previous rung's lifespan as its own; per-rung
detectability is labelled IS-sample based (the deriver may not see the holdout),
so absolute figures are optimistic while the ranking is unaffected.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Self-review of 1540ba8 against the FULL 6,930-era log rather than the first
three minutes of it. Three corrections.
1. EffectiveSampleSize() clamped in the wrong order. MathMax(2, MathMin(eff,
rawN)) returns 2 when rawN is 1 - an effective sample LARGER than the raw
one, shrinking the SE in exactly the direction the function exists to
prevent. Floor first, cap at rawN last.
2. The horizon cap rejected on the CEILING only, and said so as though that
made the label untruncated. It does not: the horizon ladder also snaps DOWN,
so a pair needing 317 bars is granted 256 and is silently truncated without
ever being flagged CLAMPED. Added SnapHorizonToLadder() / GrantedHorizonBars()
and the scale ladder now reports "needs N gets M" per rung. Rejection stays on
the ceiling alone - matching ReportGeometryExpectancyScan's '!' exactly, which
was the point - because rejecting on the snap-down would select rungs for
landing just above a ladder point rather than for anything about the market.
ComputeBarrierHorizonBars' private copy of the ladder is gone; there is now
one copy, which is the whole reason RequiredHorizonBars was factored out.
3. RETRACTED THE JUSTIFICATION IN 1540ba8's COMMENTS. That commit claimed the
overlap correction was needed because the operating point's null-of-the-
maximum gate fired on 47/73 Perceptron eras (64%) where a family-wise test
should fire on ~5%. Those 73 fits were the first three minutes of a
six-and-a-half-hour run. Over the full run:
PAI 47/3214 = 1.5% HYB 30/1200 = 2.5%
CONV 4/63 = 6.3% LSTM 75/915 = 8.2%
All at or below the null. The gate from 7414570 is working as designed and
PAI's 47 clears were a cold-start transient never repeated in 3,141 later
fits; its threshold over the run's second half has sd 0.01. The overlap
correction is still right - sqrt(p(1-p)/n) on overlapping labels is the wrong
formula - but it fixes no observed failure, and it costs nothing today
because no model is near the deploy line.
WHAT THE FULL RUN DOES CONFIRM, unchanged: the geometry ran away exactly as
described (2.00/6.00 h128 -> 3.49/6.99 h256 -> 4.86/9.71 h384, three passes,
stopping at q90 because the quantile ladder ended), the label stayed long-skewed
at Buy 42.9% / Sell 22.6%, and no checkpoint on any of the four models ever
cleared the deployability floor. Pooled declustered win rates: PAI 31.70%,
HYB 31.02%, LSTM 32.69%, CONV 31.57% - every one 4-6pp below the 37% always-long
chance rate and 1-2.7pp below the 33.7% cost-adjusted break-even.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Three defects, all surfaced by the 2026-08-17 SP500 H4 run that shipped
stop 4.86 / target 9.71 (width 14.57*ATR, horizon 384).
1. EVERY STANDARD ERROR ASSUMED INDEPENDENT SAMPLES. Triple-barrier labels
started one per bar overlap by the label's lifespan, so n calls are worth
~n/L independent observations (Lopez de Prado, AFML ch. 4 - sample
uniqueness). All three sqrt(p(1-p)/n) sites divided by the RAW count.
The tell: the operating point's null-of-the-maximum gate is family-wise and
should fire on ~5% of eras under the null. Measured fire rates - PAI 47/73
(64%), ConvLSTM 9/24, LSTM 8/21 (38%), CONV 4/62 (6%). CONV, the only model
whose margin distribution admits few bins, sat on the null; the rest cleared
a bar that was too low by ~sqrt(L). PAI's deployed threshold consequently
alternated between the ENDS of its own range era to era (0.10 -> 0.88 ->
0.86 -> 0.66; coverage 16% <-> 73%).
TripleBarrierLabel now records when each label became KNOWABLE - the first
winning touch, or both stops, or the timeout - and the prebuild accumulates
the mean. EffectiveSampleSize() feeds the operating point, the member deploy
gate and the ensemble vote gate. Conservative by construction (n/L is an
upper bound on the damage); gates get harder, never easier.
2. THE SCALE LADDER RAN AWAY, again. Horizon scales as swingMedian*sl*tp, and
since 4d8cb08 reachability is measured OVER that horizon - so a wider rung
buys itself the time that makes it look reachable. Same target -> horizon ->
reach -> target loop the excursion window is kept short to avoid; fixing the
window confusion reopened it through the other door. It walked 128 -> 256 ->
384 bars and stopped at q90, the widest rung there is, with every rung
reading 39-48% against a 20% floor. A floor nothing fails selects nothing.
Rungs whose required horizon exceeds BARRIER_HORIZON_MAX are now rejected -
the same rule ReportGeometryExpectancyScan already applied. It was printing
the shipped pair as CLAMPED and disqualified ('h384!') two lines under the
deriver that chose it: two subsystems, one geometry, opposite verdicts.
3. THE RUNG SNAP DESTROYED THE RATIO IT WAS COMPARING. Both legs snapped
independently to the coarse first-passage grid, re-rating each candidate:
q90 4.86/9.71 -> 5.00/10.00 (2.00), q85 4.07/8.14 -> 5.00/10.00 (IDENTICAL
measurement), q75 3.07/6.13 -> 4.00/6.50 (1.63 - a nearer target). So the
ladder compared win shares taken at ratios from 1.63 to 2.17 and read the
differences as scale. It is why the reach column came out non-monotone in
width (q75 48.5% above q90 42.9%). The stop now snaps to its nearest rung in
log space and the target follows the ratio off it; the pair actually measured
is returned and logged, so a collision reads as a collision.
Also: LadderWinShare guarded against the conditional (fractal) geometry path,
which fills n from m_fracLegCount while leaving idxList empty - a latent
out-of-bounds on a currently-dead path.
New log lines: mean label lifespan and effective n on the label-cache line, the
required-vs-available horizon per rung, and the grid pair the reconciliation
actually measured (its tolerance now scales with the grid skew instead of a flat
5pp).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
RECONCILED: the derivation reported "target reached on 17.7% of bars" while the
label cache reported Buy on 35.9%. Nothing was broken. They measure different
windows, and both are correct:
EXCURSION window ~12 bars (the SWING MEDIAN) - what m_excUpCache accumulates
over. Deliberately short: sizing a barrier off travel
measured over a horizon that itself scales with the barrier
is circular, and it ran away to 14-31*ATR on EURUSD/USDCAD
in 2026-08-07. That guard is correct and stays.
BARRIER horizon 64 bars - what the LABEL walk and the first-passage ladder
run over, and how long the EA actually holds the trade.
So `up >= target` is a 12-bar question and `label == Buy` is a 64-bar one, and
the second can freely exceed the first. TripleBarrierLabel gates the excursion
accumulation on `idx - t <= excWindow` while the barrier walk and the ladder run
the full horizon - the split is explicit and intentional.
THE BUG IS MINE. bc57aca's scale ladder tested reachability with `up[i] >= tp`,
i.e. it asked the 12-bar question about a 64-bar trade. That understates
reachability by ~2x, which is why EVERY wide rung was rejected and the geometry
fell back to the tightest rung at 1.61/3.21. The data supported considerably
wider; the test was just asking the wrong question.
FIX: LadderWinShare() reads the answer off the first-passage ladder - target
touched strictly before the stop, over the full horizon, tie to the stop. That
is the identical question the label walk asks, so the ladder share and the Buy
rate should now agree to within rung discretisation. Both legs snap to the
SMALLEST rung at or above the requested multiple (harder target, harder stop) so
the floor stays conservative.
Expect the scale ladder to select a WIDER rung on the next relabel. On this
data the excursion test read 17.7% at q50 where the true full-horizon share is
35.9%, so rungs that scored 8.1% and 2.8% were likely well above the floor.
ALSO:
- Window reconciliation now PRINTED every derivation: excursion travel share,
ladder win share, and the label cache's Buy share side by side, with the
ladder-vs-label gap flagged if it exceeds rung discretisation. Those two must
agree; if they ever stop agreeing, one of them is wrong and the line says so.
- Renamed tpReach/slReach -> tpTravel/slTravel and relabelled the log line. They
describe the EXCURSION window and are near-tautological there (a q50 stop is
exceeded by ~50% of bars); calling them "reached within the horizon" is what
made the two quantities look like one.
- BARRIER_MIN_TP_REACH_PCT is now BARRIER_MIN_REACH_FRACTION_OF_BE (0.60) x
break-even instead of a hardcoded 20.0. Break-even for 1:RR is 100/(1+RR), so
the absolute floor silently tightened as RR rose - 0.60x at RR=2 but 0.80x at
RR=3, penalising the user for asking for a bigger target. Evaluates to exactly
20.0% at the shipped RR=2, so this is a no-op today and correct if the knob
moves.
NOT COMPILED - user compiles.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The derivation read the stop from q75 of ADVERSE travel and the target from q50
of FAVOURABLE travel. Over one horizon those distributions are broadly the same
shape, so q75 > q50 MECHANICALLY - the target came out smaller than the stop no
matter what the market did. SP500 H4 shipped stop 3.07 / target 1.70: a 0.55:1
payoff needing 64.3%. That was never a measurement, it was two mismatched
constants.
The reachability line printed beside it - "target on 50.0% of bars, stop on
25.0%" - is exactly 1-q50 and 1-q75. Tautological. It cannot disconfirm
anything, and it read as validation.
WIDTH AND RATIO ARE INDEPENDENT AND ONLY ONE PAYS. EV = edge x width;
ratio is EV-neutral (a driftless walk reaches +m before -k with probability
k/(k+m), which IS break-even). Width is what buys cost efficiency: the spread
is a fixed 0.047*ATR here, so the shipped 4.77*ATR width paid it 21 times per
unit of travel. So:
RATIO = policy. BARRIER_TARGET_RR = 2.0 (user's 1:2). Break-even 33.3%.
SCALE = measured. The stop quantile is chosen from a ladder, WIDEST FIRST,
taking the first rung whose implied 2x target is still reached often
enough to be a trainable class.
That last clause is the difference from the min-reward:risk raise removed in
2026-08-09, which forced target = 2 x stop with NO reachability test, landed on
6.66*ATR reachable on 3.3% of bars, and trained the model to predict something
that essentially never happened. Same ratio; the scale now retreats until the
data says the target is attainable. Every rung is logged.
LADDER CEILING REMOVED. BARRIER_LADDER stopped at 5.00 and the expectancy scan's
"best resolvable pair on width alone" came back as stop 5.05 / target 4.95 - it
pinned to the top rung. A recommendation landing exactly on the edge of its own
search space is a boundary, not a finding: it cannot tell "5 ATR is optimal"
from "5 ATR is all we allowed". Extended to 20*ATR (8 -> 14 rungs). Nothing else
needs editing - every consumer is parameterised by BARRIER_LADDER_COUNT - and
the horizon constraints (decided >= 60%, reachability floor) now bind instead of
a constant.
THE SCAN COULD NOT SEE THE SHIPPED GEOMETRY. ReportBarrierGeometryScan looked
the configured pair up in its integer grid, and DeriveBarrierGeometry produces
CONTINUOUS multiples (3.07/1.70) that can never equal a grid point - so
cfgExcess stayed at its -1.0 sentinel and the report printed "configured 3:2
scores -1.00000", which reads as a catastrophic score and actually means "never
evaluated". Worse, the grid skipped target<stop entirely because it "inverts the
trade's whole premise" - while the derivation was shipping exactly that. The
incumbent is now always scored as a peer (never crowned; it is already in force
and is not an enum pairing the scan could adopt).
BREAK-EVEN NOW INCLUDES THE SPREAD. Every report quoted the frictionless
SL/(SL+TP). On SP500 H4 that read 64.3% while the MEASURED zero-skill rate was
62.1% - a 2.2pp gap that IS the cost, and that made every model look 2.2pp
better than it was. CostAdjustedBreakEvenPct() prices a win at (TP - spread) and
a loss at (SL + spread), matching the expectancy scan's convention exactly so
the two reports cannot disagree.
It also feeds FitDirConfThreshold, which is the correctness half: the operating
point subtracts break-even from precision, so the frictionless figure made every
candidate threshold look better by the width of the spread - 2.2pp against a
measured edge of 2.3pp, i.e. very nearly all of it.
Era line now carries both: "break-even 64.3% frictionless, 66.6% AFTER SPREAD".
Forces a full relabel and retrain. Requested.
NOT COMPILED - user compiles.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
BN_MIN_STD = 1e-4 caps the per-unit gain at 1/1e-4 = 1e4, and the comment above
it states that as though it were a safety property. It is not. A unit whose
running variance is ~0 is a CONSTANT feature carrying no information, and
dividing its rounding noise by 1e-4 hands the next layer an activation of
several hundred. BN's contract is "output has ~unit variance"; a unit that
cannot supply that must contribute nothing, not the largest signal in the layer.
MEASURED, 2026-08-17 SP500 H4, four topologies on identical separate charts:
model spread Neutral CHOSE Neutral TIED rail
CONV 0.386 0.68% 0.10% 0.48%
LSTM 0.392 0.63% 0.00% 0.00%
HYB 0.376 1.79% 0.00% 0.01%
PAI 0.192 0.09% 80.63% 99.99%
bn1's cached nx normed 1.38e4 over 800 units. PAI's SIGMOID head was on its
rails on 99.99% of bars, with Buy and Sell landing on the SAME rail so they
compared exactly equal, and ApplyClassificationSoftmax()'s strict-majority rule
reported that tie as Neutral on ~80% of bars.
So the long-running "PAI is heavily biased toward Neutral" was never a
class-prior problem: the net CHOSE Neutral on 0.09% of bars. It was float
equality on a saturated head. The 331ab29 counters answered it on their first
run.
PAI-only because it is the one topology whose FIRST batch norm sits on the raw
800-dim input vector - CONV/LSTM/CONVLSTM all have a conv or LSTM stage in
front, so their first BN sees a learned representation with no degenerate
units. That asymmetry was already on file as a suspicion; this is the mechanism.
FIX, mirrored in both backends (host NeuronBatchNorm.mqh and device Network.cl):
forward nx = clamp(delta/sd, -BN_MAX_NX, +BN_MAX_NX), BN_MAX_NX = 8
backward if the forward bound this unit, the output stopped depending on the
input, so d(nx)/dx = 0 and NO gradient passes
The backward half is not optional. g is divided by the same sd the forward
multiplies by, so a degenerate unit gets its GRADIENT amplified 1e4x too - the
"receives gradients divided by sqrt(var) ~ 500" pathology already noted in
Network.cl's Adam kernel. Bounding only the forward would move the explosion
downstream.
8 sigma is inert on anything healthy (|nx| > 8 is a ~1e-15 event under
normality); it binds only on degenerate units, which is the entire point. Same
clamp-to-range idiom the activation derivatives beside it already use.
SelfCheckBnForward/SelfCheckBnHiddenGrad already prove host against kernel, and
BN_OPT_NX was already consumed in the backward for the gamma gradient, so the
new read adds no lifetime assumption.
Expect PAI to change behaviour and CONV/LSTM/CONVLSTM not to (their rail rate is
~0%, so the clamp never binds). No .nnw format or fingerprint change.
ALSO: print the zero-skill reference on the era line. m_oosWinLongTotal and
m_oosWinShortTotal have been accumulated for a long time and NEVER printed,
which is why three separate topologies all sitting at 62% read as a mysterious
coincidence rather than the obvious base rate. It is not a coincidence: with the
target (1.70 ATR) nearer than the stop (3.07 ATR), BOTH sides win on 24.5% of
bars, so winLong+winShort covers ~124% of them and a no-edge caller collects
62.1% whichever way it calls - against a 64.3% break-even. Derived from this
run's own label counts: (11329 - 2776 + 2*2776) / (2*11350) = 62.14%.
Every win rate on that line must be read against this, not against 50%.
NOT COMPILED - user compiles.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
ApplyClassificationSoftmax() requires a STRICT majority over both rivals and
sends every tie, 2-way or 3-way, to Neutral. So "OOS recall Neutral:100%" is
two completely different events sharing one label:
CHOSE - the net genuinely ranks Neutral highest. A class-prior/label problem.
TIED - the top two are EXACTLY equal, so the net expressed no preference and
the tie-break reported Neutral. A SATURATION problem: the head is
SIGMOID, and a saturated sigmoid returns exactly 0.0f or 1.0f in the
DLL's float32, so two classes pinned to the same rail compare equal
and the bar is silently discarded.
Nothing in the logs could tell them apart, and the fixes point opposite ways.
Eras 1-25 of the 2026-08-17 solo PAI run read "Neutral 100%" at spread avg 0.99
- fully saturated - and broke out at era 27 as the spread fell to 0.75. That is
consistent with EITHER story. The user reports the Neutral phase on most runs,
so it is worth four longs to stop guessing.
Four per-era counters on the pass 3 OOS walk, reported as:
| Neutral CHOSE 12.4% / TIED 38.1% (of which B=S 1204) | rail 61.2%
m_oosNeutralStrict - Neutral strictly highest
m_oosNeutralTie - no strict winner; the tie-break produced Neutral
m_oosTieBuySell - the costly subset: Buy and Sell tied AT the top, i.e. a
DIRECTIONAL reading thrown away by float equality
m_oosRailBars - any raw output sitting on a sigmoid asymptote, the
saturation that makes exact ties possible at all
Read on the RAW logits, before ApplyClassificationSoftmax() overwrites TempData
in place. Legitimate because softmax is strictly monotone: it cannot change the
ordering and cannot break a tie either, so the raw reading and the decision
always agree. Placed alongside the existing min/max/spread capture so all the
output diagnostics describe the same values.
Measurement only - no decision path reads these.
NOT COMPILED - user compiles.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two defects behind the "training is highly unstable" report, from 101 eras of
SP500 H4 PAI logs. Neither was the optimizer.
1) LayerLearningReport's dW/W for BN layers divided by the WHOLE packed block.
getWeightsBN concatenates the outgoing dense matrix, gamma/beta, the running
mean/variance, the Adam moments AND BN_OPT_NX - the forward-pass scratch copy
of the normalized input. At era 101 bn1's dense matrix normed 15.1 against a
block norm of 15430.3, of which NX alone was 15429.3: the weights were 0.098%
of their own denominator, a 1022x inflation. NX is also near-constant between
era-end reports (same last forward pass), which pins the numerator down too,
so the layer read "bn1:0.000%" for 101 consecutive eras and was diagnosed as a
frozen first layer. It was the ruler that was broken. The ratio now covers
trainable parameters only (dense matrix + gamma + beta); mean/var/NX/Adam are
excluded. NX is reported separately because it is a health signal in its own
right - bn5 read nx 6.8e6 over 16 neurons, ~1.7e6 per unit against a healthy
~1.0, which is what a near-zero running variance in the denominator looks like.
NO historical dW/W reading on a bn* layer is admissible evidence that a layer
did or did not train. That includes every such claim in this repo's notes.
2) FitDirConfThreshold took a bare argmax of coverage x (precision - breakEven)
over 50 bins. Measured across 98 consecutive fits:
correlation(chosen threshold, win rate at it) = -0.056 over 0.00..0.74
win rate stdev across fits = 1.32pp
binomial SE of that win rate at ~1430 calls = 1.25pp
The correlation is zero - the margin does not rank trades - and the era-to-era
spread IS its own sampling error to within 0.07pp. So the objective was
coverage x (3.4 +/- 1.3) and the argmax over ~37 eligible bins returned
whichever bin drew the luckiest sample. The threshold teleported
0.42 -> 0.04 -> 0.74 in three eras, swinging OOS coverage 0% -> 39%, leaving
the era win rate measured on 1-5 calls and swinging 0% <-> 100%. That is the
entire reported instability.
The argmax is now adopted only if it beats a DETERMINISTIC fallback - the most
selective bin still clearing the coverage floor, chosen from the margin
distribution alone and never from a win rate - by more than a best-of-N
maximum could manage on noise, sqrt(2 ln N) standard errors. Same null-of-the-
maximum correction the deploy gate already applies to model selection.
A plain one-standard-error band was tried first and is NOT sufficient: its
edge is bestScore - bestSE, and with a 2.3pp edge against a 1.25pp SE that
edge is itself +/-50%, so the admitted set would still wander by half its own
width every era. The fallback has to be independent of the noisy quantity.
Simulated on the observed numbers: falls back every era at the current 2.3pp
edge (stable), adopts the argmax once a real edge reaches ~5pp.
NOT COMPILED - user compiles.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
NOT COMPILED - user compiles.
(1) THE RANKING SLICE IS GONE. It reserved 20% of the OOS window so the
pattern-DB backfill would read bars the deployed checkpoint was not SELECTED on.
That objection stands; carving a new region to answer it did not. The calibration
band already has every property the slice was buying:
never trained on | never graded by pass 3 (which walks [0, oosCutoff) and so
never reaches it) | never seen by the deploy gate | purged by a full label
horizon on BOTH sides | and larger besides - 1,684 bars vs the ~970 carved
So the backfill now walks [calibLo, calibHi) and pass 3 goes back to grading the
entire OOS window, exactly as before any of this. The gate gets its full sample
back (~10% of a sigma), the split loses a region, and the failure mode found an
hour ago - a reserved region silently blanking ~10 months of chart arrows,
because arrows are only drawn on bars pass 3 grades - becomes impossible.
One impurity, stated in the completion log rather than hidden:
m_dirConfThreshold is FITTED on that band and the walk applies it to decide which
bars fired, so coverage there is mildly optimistic. One scalar under a coverage
floor, against checkpoint selection over hundreds of eras.
This backfill IS the deploy-time warm-up: it runs right after FinalizeTrainRun()
restores the deployed weights, so it scores with exactly what is about to trade.
(2) EVERY CALL WAS TIER 0, AND IT WAS ARITHMETIC. ConfidenceTier() quartiles
[floorConf, 1] where floorConf = 1/3 - the lowest magnitude a 3-way softmax
winner can hold. But it was fed CalibratedConfidenceMagnitude(), which multiplies
by m_confidenceCalScale, clamped to [0.3, 1.5]. That lower clamp is BELOW 1/3.
Whenever calibration bottoms out, t goes negative and MathMax(0, ...) pins every
call to tier 0.
Which is what the live run does. m_confidenceCalScale is EMA'd toward
empiricalAccuracy / avgClaimedConfidence; with the model over-calling Neutral,
3-class agreement sits near 10% against a claimed confidence near 0.9, so the
ratio is ~0.11 and clamps to 0.3 every era. Logged:
tier prec T0:72%(828) T1:n/a(0) T2:n/a(0) T3:n/a(0)
828 calls, one bucket - the four tier weights and the entire per-tier pattern-DB
ranking reduced to a single number. The backfill was feeding a mechanism that
structurally could not rank.
Tiering now reads the RAW head magnitude, which genuinely lives on the
[1/3, 1] range these bounds were written for. Calibration keeps its real jobs -
AIConfidence() for MM sizing and SignedAIConfidence() for the vote are unchanged.
STILL OPEN, deliberately not touched here: the calibration TARGET itself.
empiricalAccuracy is 3-class agreement, which is the wrong quantity to scale a
DIRECTIONAL confidence against - it counts a Neutral class that is 0.19% of
labels. The honest target is the win rate on the calls the confidence describes
(directional precision), with the claimed-confidence average taken over those
same called bars. That needs a new accumulator and it interacts with the Neutral
over-calling being fixed elsewhere, so it wants one clean run first.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
NOT COMPILED - user compiles.
User: "there is quite some trading going on, but absolutely nothing on the recent
area of the chart, like there is a hard wall starting around november 2025."
That wall is 7caf2f6's ranking slice, and it was placed at the wrong end. Chart
arrows are only ever drawn on bars pass 3 GRADES, and the slice reserved the
NEWEST 20% of the OOS window plus a label-horizon purge. At the live sizing -
~4,860 OOS bars, 128-bar horizon - that is ~1,100 H4 bars withheld from grading,
about ten months back from today, exactly where the wall appears.
The invisible cost was worse than the visible one: it handed the deploy gate the
OLDEST 80% of the OOS window and withheld the most recent regime from the single
decision that has to generalise forward.
Both fixed by putting the reserve at the oldest end instead:
[0, oosScoreHi) OOS - graded by pass 3 (NEWEST, arrows restored)
[oosScoreHi, rankLo) purge - one label horizon
[rankLo, oosCutoff) RANKING - backfill only, graded by nobody
[oosCutoff, calibLo) purge
[calibLo, calibHi) CALIBRATION
... IS
Of the three consumers competing for those bars, recency is worth least to the
ranking: it is an ORDERING of confidence tiers, far less regime-sensitive than an
absolute win rate, while the gate's power and the operator's read of the chart
both want the newest data. The slice keeps every property that made it worth
carving - never graded, never selected on, never seen by the gate, purged on both
sides - so the backfilled rows are still honestly out-of-sample.
RankSliceHiIndex is replaced by RankSliceLoIndex + OosScoreHiIndex; pass 3 now
excludes the slice at the TOP of its walk and descends to 2 as it always did.
The backfill walks [RankSliceLoIndex, oosCutoff) via a new m_dbBackfillStopIndex,
clamped at both ends so a degenerate slice yields an empty walk rather than one
that wanders into graded bars. Verified no reference to the old helper survives.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
NOT COMPILED - user compiles.
Root cause of the Neutral collapse. Logit adjustment (Menon et al. 2020) makes a
classifier Bayes-optimal for BALANCED error by subsidising rare classes. It was
wired here when Neutral was the DOMINANT class - the "big move up / big move down
/ nothing much" era, where the correction pulled the model off the majority.
The triple-barrier relabel (b4a704d) inverted the distribution. The barriers are
now the EA's own SL/TP, so ~89% of bars RESOLVE and only timeouts are Neutral.
Measured on SP500 H4, from the EA's own log:
measured priors Buy 48.26% Sell 41.13% Neutral 10.61%
log-prior spread 1.52 | tau 1.00 CAPPED to 0.79
Neutral became the RAREST class, so the correction started subsidising it - by
tau*(log pB - log pN) = 1.20 logits. With no directional edge to overcome that
(direction is closed at best-of-999, p=1.0000), the model took the free lunch:
OOS recall Buy:1% Sell:0% Neutral:100%
OOS raw out spread avg 0.9993 (softmax saturated, near one-hot)
dW/W bn1 0.000% bn3 2.0% bn5 6.5% (input weight block frozen; head twitching)
The anti-collapse mechanism was the collapse. The recall gate needs >=40% on all
three classes, so nothing could ever deploy and the plateau ladder burned eras.
Present in both runs today (b6b5 froze bn1 by era ~719, 17ae by ~169), so it
predates this week's work.
FIX: the correction now spans the DECIDABLE classes only, Buy against Sell,
centred on their midpoint, with Neutral pinned at offset 0. Neutral is the
ABSTAIN outcome and abstention already has a better owner - m_dirConfThreshold,
refitted every era on the held-out calibration band against a coverage floor and
the measured break-even. Subsidising the abstain class does that job twice and
spends the whole correction suppressing the only decisions that can pay.
What still gets corrected is real: a trending symbol resolves more long barriers
than short, and uncorrected the model inherits that as a standing directional
bias. Here it is log(0.4826)-log(0.4113) = 0.16, so the offsets are tiny - the
correct answer, not a broken one. The two traded classes were already balanced;
the old spread of 1.52 only ever described how rare a timeout is.
Everything is derived from the measured distribution, as requested - offsets from
the priors, cap from the resulting spread. tau itself is deliberately NOT fitted:
tuning it against the same data that selects the checkpoint would add another
search dimension to a project that has been burned by exactly that. tau=1 is the
theory value and the cap (now ~9.5x looser at spread 0.16) will rarely bind.
Log line now reports both spreads and, when the abstain class is the rarest, says
how much the old form would have boosted it. Fingerprint |LA:<tau> -> |LA:<tau>:BS
so models trained under the all-three form re-key instead of resuming.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
b197999 gated the repaint on "this era beat the best checkpoint". That was not
what was asked for and it changes policy rather than fixing a bug, so it is
reverted - behaviour is now byte-identical to before b197999. Only a comment
recording what was verified remains.
What the check found: there is NO repaint or erase at the start of an era. With
NMS on, passes 1, 2.5 and 3 only RECORD predictions into m_arrowSignalCache -
pass 1's own comment says "NMS on: record only - the era-end sweep is the SOLE
renderer, so no raw (un-declustered) arrow is ever drawn mid-era" - and
PruneDirectionalClusters runs once, at the end of pass 3, which is the end of the
era. The requested behaviour was already the implemented behaviour.
So the arrows vanishing at the era boundary is not a timing fault. That sweep
DELETEs the arrow on any bar the era scored Neutral, and the model is currently
scoring Neutral on 98-100% of bars (see the logit-adjustment finding: Neutral is
the RAREST class at 10.6% and the imbalance correction is subsidising it by ~1.2
logits). The chart is reporting the model accurately; the model is the problem.
One thing that CAN clear arrows at era 0, and did on the first attach after the
|ALTW re-key: ClearPersistedChartSignals("fresh topology at era 0 - arrows belong
to a previous model"). That fires once per fresh model, not per era.
Compile-verified: 0 errors, 0 warnings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
User report: "as soon as the next era training begins the chart signals are
erased, they should persist for as long as they are accurate."
Cause: PruneDirectionalClusters ran unconditionally at the end of pass 3, and it
DELETEs the arrow on any bar the CURRENT era scored Neutral. A model exploring
away from its best therefore wipes the chart every era even though the best
checkpoint still calls those turns. On the run that prompted this the model sat
at OOS recall Neutral:100% for 40+ consecutive eras, so essentially every arrow
was deleted at every era boundary.
The render is now deferred to the era-end block - the first point that knows
whether the era beat the best checkpoint - and only a new best repaints. On eras
that did not improve, the previous best's arrows stay untouched. Two exceptions
keep the chart from ever showing nothing: before the first checkpoint exists
there is no best to preserve, so early eras still paint; and a finishing run
repaints unconditionally, because FinalizeTrainRun is about to restore the
deployed weights and the chart must describe THOSE.
Recorded for ensemble members too. A member's own best era is not the deployable
one (the joint checkpoint decides that), but it is still the most accurate thing
that member has drawn, and the alternative is a chart that empties itself.
Also verified against the log, since two other symptoms were reported alongside:
era cadence PAI-b6b5 (before these changes) 3.76 s/era
PAI-17ae (after) 3.53 s/era
topology both "2 dense from 16 units | input 800 (16 bars x 50)"
calibration both fitted on 1699 held-out bars
So training speed is unchanged - the ~6% is the ranking slice removing 20% of
pass 3's bars. It only FEELS fast because this is a single PAI chart taking the
whole 120 ms budget, not four ensemble members sharing it behind an era barrier.
The Neutral collapse is also pre-existing, not new: b6b5 ran at Neutral 94-100%
with 1-5% directional calls for all 723 of its eras, before any of this work.
That is the known neutral-collapse/recall-gate failure mode, and it is what
"barely drawing signals" actually is. Worth watching, separately: b6b5 reached
best-bal 34.2% by era 723 while 17ae is at 13.6% after 77 - too early to read,
but it is the number to check once 17ae has run comparable eras.
Compile-verified: 0 errors, 0 warnings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
User's call before deploy: "I would rather avoid lagging so the NN finds
accurate patterns." Correct instinct, and it picks the conservative variant.
110b384 deduplicated the WHOLE window, so every distinct reading survived at one
slot. The flaw is which slot: it depends on where the calendar-day boundary falls
inside that particular window, and on H4 that boundary cycles through ~6 phases.
A dense layer holds a separate weight per (slot, feature), so a given lag would
have landed on a different coordinate from one window to the next - turning a
stable lagged input into a moving one.
Now it blanks only bars carrying a BYTE-IDENTICAL copy of the anchor's reading
and stops at the first bar that differs. An as-of lookup into a daily file is a
step function in time, so those copies are exactly the contiguous run of bars
sharing the anchor's calendar day. Everything older keeps its natural replicated
run, in the same slots it always occupied - whatever the net learned to read
there, it still reads there.
Why the anchor's reading is the right one to isolate: the window's newest slot IS
the bar being predicted (BuildFeatureWindow's final iteration lands on r, and
pass 3 grades that same index), so it is the reading contemporaneous with the
decision - and the only one the alt screens ever validated. They measured the
CURRENT reading's MI against forward range and never tested lags, so the lagged
content is unproven, which is a reason to leave it undisturbed rather than a
licence to rearrange it.
What is still fixed: the anchor's reading reaches the first layer on one
coordinate instead of once per bar of its day, removing the ~16x gradient
upweight for the validated signal. And this is IDENTICAL to full dedup exactly
where replication was worst - on M15/H1 the whole window sits inside one calendar
day, so the anchor's run is the whole window - and a no-op on D1, where the bar
before the anchor is already a different day and the loop breaks immediately.
The two differ only on middle timeframes, and there this is the safe side.
Fingerprint |ALTW:1 -> |ALTW:2 so nothing trained under the hour-old full-dedup
semantics can silently resume under these.
Compile-verified: 0 errors, 0 warnings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
aba9bd2 kept only the newest slot's copy of the external block. That is right on
every intraday timeframe and WRONG on D1: there the 16 window bars are 16
distinct calendar days, the daily alt file returns a different row for each, and
blanking 15 of them destroyed real information instead of a copy of it. Caught
while extending the measurement to the other instruments.
Now compares values instead of slot positions: walk newest -> oldest, keep the
last DISTINCT reading, blank a slot only when it repeats one a newer slot
already carries. Exact on every timeframe with no timeframe test, and it also
handles weekends, holidays and publication gaps, where a window spans fewer
distinct rows than calendar days. How much collapses falls out of the data:
M15 x 16 bars = 0.17 calendar days -> 1 distinct row -> 15 of 16 blanked
H1 x 16 bars = 0.67 calendar days -> 1 distinct row -> 15 of 16 blanked
H4 x 16 bars = 2.67 calendar days -> ~3 rows -> ~13 of 16 blanked
D1 x 16 bars = 16 calendar days -> 16 rows -> NOTHING blanked
OTHER INSTRUMENTS - the question that prompted this. Per-symbol exports for
USDJPY/XAUUSD/EURUSD are not on disk (written only when that chart is attached),
but they are not needed: CAltData reads a DAILY file for every symbol, so bars
sharing a calendar day are byte-identical by construction everywhere. What
varies per symbol is only WHICH sources, and the catalog (AltDataFetch.mqh
AddSpec rows) gives:
SP500 13 = risk(3) + cot_spec_net(1) + eia(3) + mac(6)
EURUSD 15 = cot(3) + risk(3) + eia(3) + mac(6)
USDJPY 15 = cot(3) + risk(3) + eia(3) + mac(6)
XAUUSD 14 = risk(3) + ivol(2, GVZCLS) + eia(3) + mac(6)
Measured observation-date gaps in the raw sources on disk: VIX, USD index,
DGS10, T10Y2Y, T5YIE, DFF, ECBDFR all 1 day; COT and EIA 7 days; CPI and UNRATE
31 days. NO per-bar source exists anywhere in the catalog - the tick-activity
survivors from the flow screen are an in-terminal feature block, not alt data,
and are untouched by any of this. So the redundancy is universal across
instruments; only its magnitude varies, and by timeframe rather than by symbol.
Compile-verified: 0 errors, 0 warnings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Measured on the live SP500 D1 export (6073 rows, 13 features, 5888 simulated
16-bar windows):
distinct values per feature per window : 1.7 - 2.7 of 16 slots
variance in the first 13 PCs : 96.5 - 97.0%
components for 95% / 99% : 12 / 17-19
effective rank (entropy) : ~11.5
208 inputs carrying about 12 dimensions. Only 6 of the 13 features move daily
(VIX complex, USD, the rates trio); 5 are weekly (COT, EIA, output gap) and 2
monthly (CPI, unemployment). The lookup is as-of by bar open time into a DAILY
file, so bars sharing a calendar day are byte-identical by construction.
The cost is NOT overfitting capacity - collinear copies span ~12 directions,
not 208, so an earlier claim that this wasted 26% of the model overstated it.
It is GRADIENT WEIGHTING. Batch norm standardizes each of the 208 coordinates
independently; that rescales the copies without decorrelating them, so one
factor arrives on 16 unit-variance coordinates, each weight takes a full-size
step, and the factor's aggregate coefficient moves ~16x faster than a per-bar
price feature's. The network was biased toward the external block by a factor
of the window length - and pointing the wrong way, since these features cleared
only a marginal incremental screen while price is the base signal.
Zeroed at WINDOW ASSEMBLY, not in BufferTempData: that output is cached PER BAR
and a bar sits at slot 15 of one window and slot 0 of the next, so a
slot-dependent value there would poison the cache or force a recompute per slot.
The cache keeps true values; only this window's copies are cleared. Width
contract untouched - same count, same positions - so conv/LSTM/HYBRID keep their
bar-major rectangle unchanged and the block arrives at the newest bar, which for
the LSTM is the final timestep. Zero-variance coordinates are safe through batch
norm (divisor is MathMax(MathSqrt(var + BN_EPSILON), BN_MIN_STD)).
Fingerprint gains |ALTW:1 when alt data is on. Same width and same .cfg, so
nothing else would have caught a model trained under the replicated layout
resuming under this one. Conditional append per the existing rule: configs
without alt data keep their fingerprints and their trained models.
NOT the concat branch. CNet is a strictly linear stack (CLayerDescription has no
input-source field; NetBuild wires i to i+1 and stores layer L's weights on
L-1), so a real two-tower model needs a new multi-input layer type across
WarriorCPU, WarriorDML and the OpenCL kernels plus an .nnw format change - the
highest-risk change in this repo, in the code that produced the transposed dense
gradient, the Adam second-moment bug and the reversed LSTM window. This captures
the part of that idea the measurement actually supports, at no engine risk.
Compile-verified: 0 errors, 0 warnings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
TOPOLOGY - reverts the two constants and drops CausalHiddenLayerFloor.
The MQL5 article's 30%-per-layer cut and floor of 20 are load-bearing on ITS
first-layer width of 1000 (1000->300->90->27 needs a floor to stop). This
codebase MEASURES that width, and on the live SP500 H4 config it is 16 units -
already floored, with the budget printing "11360 estimated in-sample bars
cannot support a 800-wide input ... roughly 1.1 weights per training bar -
expect overfitting". At 16 units a floor of 20 makes lastHidden >=
m_initialNeuronsCount, so ComputeHiddenLayerCount returns on its first branch
and the width taper - the only part derived from this symbol's data - became
dead code on all four ensemble members, with depth (2 -> 4) set entirely by
counting feature domains. ComputeLayerWidths had already rejected this exact
pair of constants in its own comment.
The causal floor's premise does not hold either: layers are not inference
steps. The "1 layer linear / 2 nonlinear / 3 multi-connected" result is
Lippmann 1987 and is about hard-threshold units; with sigmoid/ReLU, Cybenko
1989 and Hornik 1991 give universal approximation from a single hidden layer.
Depth buys parameter efficiency for compositional functions, not reasoning
hops. ForceHiddenLayers remains for measuring depth directly.
RANKING SLICE - the backfill no longer reads the window it is judged on.
The deployed checkpoint is CHOSEN as the best-scoring era on the OOS window,
so win rates measured back over it are selection-inflated, and the backfill
was writing exactly those into the table filter weights rank on: the
selection set consumed twice, beside a deploy gate that applies a Sidak
correction for that effect. The newest RANK_SLICE_PCT_OF_OOS (20%) of the OOS
window, plus a label-horizon purge, is now reserved and graded by nothing -
not pass 3, not checkpoint selection, not the gate. The backfill reads only
that. The gate keeps ~80% of its measurement (power goes as the square root,
so ~10% of a sigma), and the slice is the newest data, which is the regime
about to be traded. RankSliceBars returns 0 when no honest slice fits and the
backfill then REFUSES and says so, rather than falling back to the scoring
window and looking like a success.
SHRINKAGE - per-tier win rates are shrunk toward the filter's own pooled rate
by MIN_TRADES_FOR_WIN_RATE pseudo-trades before becoming weights. The raw
ratio at the minimum sample count carries a ~15pp standard error, so a tier
that went 8-2 was handed weight 80 and outranked a tier measured over
hundreds of calls at 55 - the ranking was being driven by which small tier got
lucky. Opt-in per call site (priorWeight 0 keeps the raw behaviour).
Compile-verified: 0 errors, 0 warnings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Four defects in 64c5dd5/1a05e63, found by review + a baseline compile.
Goals 1-8 of that session are unchanged; this makes 6 and 8 actually reachable.
1. HEAD DID NOT COMPILE - 6 errors. CControlPanel::Minimize/Maximize were
declared `virtual bool ... override`, but CAppDialog declares both as
`virtual void` (Controls\Dialog.mqh). errors 265 + 404 on each, plus 151
on `bool ok = CAppDialog::Minimize()`. Return type is void now; there was
never a success flag to forward. Verified: 0 errors, 0 warnings.
2. THE BACKFILL COULD NEVER ADVANCE, and neither could the OOS continual
simulation (that one has been dead since it was written). Both are armed
at the instant convergence is declared, and both advance only from inside
Train(), one chunk per call. But ScheduleTrainingIfNeeded's only per-tick
ArmStudyEvent site sits in the `else` of a branch taken whenever
m_trainingComplete is set and m_trainRunActive is clear - which is exactly
the state FinalizeTrainRun() leaves behind one line before they are armed.
Train() was never called again, so the walks sat at their start index
forever: no "simulation complete" line, and not one row written to the DB
this feature exists to fill. Only a manual Resume/Retrain unstuck them.
Both flags now keep the model schedulable.
3. IN AI_HYBRID - the mode this ships in - the backfill was never even armed.
Ensemble members deploy at Train() ENTRY and return immediately (so no era
is wasted), which skips the era-end block the backfill was started from.
All four members were a no-op for a second, independent reason. Armed on
the ensemble deploy path too, from m_resumeBars/m_resumeOosCutoff.
4. RE-RUNS DUPLICATED ROWS. RegisterSignal inserts unconditionally - no key,
no duplicate check - and m_dbBackfillDone is in-memory, so every later
attach that retrained to convergence wrote a second full set of rows for
the same bars. The ranking would count one bar once per model that ever
deployed, weighting superseded opinions as heavily as the live one. A
.dbfill marker stamps the deployed era; written only on completion (an
interrupted walk redoes itself rather than ranking a partial window) and
deleted with the other sidecars on reset-weights.
Also: WarmBlocking's timeout was silent, which restored the exact silent
pin failure it was added to prevent - it now says so in the journal, and
returns true for "no reference pairs to wait for" so the warning stays rare
enough to be read.
Not addressed, needs a decision: the backfill scores the OOS window with the
checkpoint that was SELECTED as best on that same window, then writes those
win rates into the table filter weights rank on - the selection set consumed
twice, undiscounted, while the deploy gate right next to it applies a
family-wise correction for exactly that effect. The rows are also simulated
triple-barrier outcomes at today's spread sharing a table with realised
fills. The completion log line now states both plainly.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The unit of evaluation in ensemble mode becomes the vote, because the
vote is what trades (user: "at the end of the day they will vote
together during live trading so that would make sense").
Four decisions move from the member to the ensemble:
* which era is "best" -> the era whose COMBINED VOTE scored best
* what is checkpointed -> a JOINT snapshot: every member's weights
at that one era
* when the run gives up -> one shared plateau ladder
* whether it may deploy -> family-wise gate on the vote
WHY THE JOINT CHECKPOINT IS THE POINT: per-member selection picks each
net's own best era, and those eras differ. The resulting quartet was
never measured together at any instant, so the vote it casts live is a
configuration no OOS number ever described. Capturing all four at the
era whose vote won makes the deployed ensemble exactly the measured one.
Correct because of the era barrier (b77e7b4): Train() runs at most one
era per call and a member that finished era N is held until every member
reaches N, so when the last member scores the vote no member's weights
have advanced past end-of-era-N. That makes the deferred simultaneous
capture a guarantee rather than a race. Each snapshot is era-STAMPED and
deploy requires every stamp to equal the winning era - otherwise a member
whose capture failed would still hold an older snapshot and the deployed
quartet would again be one nothing measured. Partial capture rolls the
era back out of "best" so the search continues instead of freezing
behind a checkpoint that does not exist.
Statistics mirror the per-member gate one for one - same coverage floor
(MIN_COVERAGE_FRACTION_OF_BASE_RATE), same always-call-one-direction
chance reference, same EDGE_MIN_SIGMAS margin, same Sidak correction over
the eras ranked (DEPLOY_FAMILY_WISE_ALPHA). Only the population differs:
the bars the VOTE fired on, at Min_Vote_Open, rather than the bars one
member called. Two-sidedness is required of the vote itself - a vote that
never goes short IS the always-long model the chance reference prices in.
Members keep their own per-era statistics and their own learning-rate
dynamics (regression restore, eta decay); those are per-net training
mechanics, not deployment decisions. The shared ladder is mirrored onto
each member so per-era log lines report the state that actually governs
them. Solo charts are untouched on every path.
Verified: full MetaEditor compile, 0 errors 0 warnings.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Four user-reported/requested items, one root cause chain:
1) DEAD CONTROL PANEL in AI_HYBRID mode. All members posted custom event
id 1 and handled id 1001, and CExpertCustom broadcasts every chart
event to every filter - so each posted event ran a train chunk in ALL
N members (N*N chunks per round) and the chart thread never idled
long enough to deliver clicks/drags. profiling.csv: 99.45% of time in
OnChartEventHandler. Fix: per-instance study-event ids
(STUDY_EVENT_ID_BASE + construction order, offset above the Controls
library's ON_* codes - id 1 was also ON_DBL_CLICK, so panel
double-clicks fired training chunks). ArmStudyEvent() is the single
post site; lost-event watchdog replaces the accidental
sibling-clears-my-flag rescue.
2) WARM-UP DUPLICATION. The auto-tune sweep is deterministic over
identical features/labels, and it ends in the full MI diagnostic
suite, which the MI-share gate never intercepted on the sweep path -
four members ran four identical ~36s sweep+report blocks. First
member publishes outcome (g_ensembleChartTuneDone/Installed/Settings);
the rest apply it and skip both.
3) DEINIT STRANDED PANEL+ARROWS (user repro 18:52). Root cause from the
log: the 4,500ms budget runs from MetaTrader's stop REQUEST - a heavy
autosave in flight ate it, OnDeinit got ~430ms and died in the first
member's arrow persist ("Abnormal termination" 432ms in). Fix: early
visible-UI sweep (native prefix deletes for status/panel/dialog)
right after ClearStatusLabel, and a fast path for still-training
models - their arrows are re-rendered every era, so they get one bulk
purge instead of scan+atomic-write in the death window.
4) ENSEMBLE FEATURES (user requests): era BARRIER - members advance era
by era together; a member ahead of the slowest still-training member
declines Train() calls and its chunk budget is donated
(TRAIN_TIME_BUDGET_MS = 120/activeTrainers, UI headroom constant).
COMBINED-VOTE OOS SCORE - each member's pass-3 scan contributes its
adjusted per-bar decision (0.0 on abstain) to a shared row buffer;
the last member to finish the era scores the averaged vote vs the
mirrored Min_Vote_Open against the same target-before-stop outcomes
members grade themselves on, publishing an "Ensemble vote" line on
the aggregated panel. Member headlines now carry their lifetime win
rate with break-even.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The ensemble chart UI had a shared-namespace defect that answered the user
question "what do the arrows represent?" with "a bug": all four members drew
arrows under the same WarSig_<bartime> object names, so the chart showed
whichever member rendered LAST, one member Neutral deleted another member Buy
at the same bar, each member init sweep wiped the arrows the previous member
had just restored, and SaveChartSignals - which rebuilds the sidecar by
SCANNING the chart - persisted every other member arrows into its own history
(the exact cross-model laundering its own header warns about, now happening
BETWEEN ensemble members).
Arrows are now namespaced per member (WarSig_PAI_, WarSig_CONV_, WarSig_LSTM_,
WarSig_HYB_): draw, delete, restore, prune, member init sweep, destructor
purge and the sidecar scan are all member-scoped, and the tooltip names the
model. Global purges keep matching the bare WarSig_ prefix, which covers all
member namespaces plus old-format leftovers from earlier builds.
Labels: the ensemble panel header no longer says "HYBRID ensemble" (HYBRID is
one member; the header is the ensemble) and the CONVLSTM member displays as
ConvLSTM instead of Hybrid. Its SHORT id stays HYB deliberately - it names the
model folder and changing it would orphan every model trained under that path.
Deinit: the alt-data mapping dialog namespace (WarriorAltMap_) joins
WarriorChartPrefixes, so both the OnInit purge and the deinit final sweep now
cover it - it was in neither list, so a dialog starved of its own Destroy()
left its controls on the chart permanently.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- PeriodMA/MA_Type/PeriodRSI: input -> const seeds (closing the set: every
indicator parameter is now tuner-owned)
- Variables\TunedPeriods.mqh: chart-level tuned-period state. A gated
install writes TunedPeriods_{SYM}_{TF}.cfg; next attach reads it BEFORE
the DB fingerprint and classic-signal config, so classic votes, DB key,
and tuner seeds always describe the same indicators regardless of
classic/AI/hybrid use. Restart-grained adoption by design (no mid-run
handle churn); new periods re-key the signal DB (semantics rule).
- EnableAltData input in AI Input Features (consumption gate only;
collection keeps running); |ALT DB-fingerprint token; opt-out on an
alt-trained model correctly starts fresh via the width compare.
- Defaults: all four classic votes OFF (AI-first; WARRIOR_MARKET_BUILD
branches collapsed with the marketplace pivot), order-flow/Wyckoff NN
features OFF (alt data is the default information diet; toggles stay).
Compiles 0 errors / 0 warnings.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The 18 inputs added 2026-08-08 (when the tuner defaulted off and the
values needed an operator path) become compile-time aliases of their own
defaults - same names, zero consumer churn, byte-identical values. The
tuner is now the only path by which these values move: it defaults ON
(the 08-08 off-flip was measured against the direction target's flat
landscape; the objective is now RANGE, which has signal), searches from
the seeds under the Sidak family-wise gate, and persists winners in the
.nnw beside the weights. ADP fingerprint token retired (deviation now
impossible by construction; tuned values were never its job).
Menu shrinks 102 -> 84 inputs. Compiles 0 errors / 0 warnings.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- MI_TUNE_TARGET = MI_TARGET_EXC_RANGE: the coordinate sweep scored
candidates against the barrier label - measured noise - so it climbed a
flat landscape and the gate rightly rejected every winner. It now
selects indicator settings for MI vs realised RANGE (4x null, positive
control), the channel the excursion head consumes these features for.
Winner gate re-tests on the same target. Barrier-label report unchanged.
- AltDataFetch 4014 handling: Alert popup + once-per-session walkthrough
with the two whitelist URLs on their own journal lines (copy-paste
ready); hourly-backoff retry instead of a permanent latch, so the
whitelist fix takes effect without re-attaching.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
System\AltDataFetch.mqh: the EA backfills missing alt-data history at
attach and keeps appending forward while deployed - online learning never
depends on an external process. CFTC Socrata API (no key, 2006->now, one
GET per symbol; ES name variants verified, max-OI dedupe) + FRED (VIXCLS/
DTWEXBGS, key from AltData\keys.txt). Identical publication stamps and
fixed a-priori transforms as research/altdata/export.py; rebuilds the
same {SYM}_D1.csv files, so Python and EA interoperate on one format.
OnTimer hook (30-min staleness check, in-memory compares when current;
never in tester - cache files serve there) + AltDataReload() on signals.
Classic-signal removal CANCELLED per user (vote experiment later).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The derived stop/target were quantiles of EVERY bar''s excursions over a
fixed horizon - q75 adverse gave a 2.6-3.5*ATR stop against a ~1.7*ATR
target (user: "looks limiting"). That pooled measurement was correct
when direction was dead (any subset of bars had the same distribution)
and is provably mis-sized now that the gate certifies the label carries
information: the bars the model trades are the labeled bars, and their
excursions differ from the pool.
FractalDirectionLabel now records, for every Buy/Sell-labeled IS bar
during the prebuild, the favourable and adverse travel in ATR units
over exactly the LEG the label points at - entry close through the next
fractal extreme (user request: "from a fractal to the next for maximum
accuracy"). DeriveBarrierGeometry reads the same q75-adverse/q50-
favourable quantiles off that conditional sample instead of the pool,
with a logged fallback to pooled when fewer than the minimum legs
exist. Quantiles kept over averages deliberately: a mean MFE is
dominated by runaway legs and would set an unreachable target.
No circularity: the fractal label does not depend on SL/TP (the barrier
label does - this path must never feed it). Recording stops the moment
geometry is derived and pinned, so pass 2 relabels and later bars
cannot silently re-shape a certified pair.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two user-reported ensemble regressions, one cause each:
- "getting ready is very long": every member ran the full MI diagnostic
suite (headline MI, positive control, alignment, lag profile,
geometry scan + winner test - ~200 permuted draws per line) on
IDENTICAL features and labels, reporting the same numbers four times.
First member runs it, the rest adopt with one log line. Documented
caveat: if the geometry scan ever ADOPTS a winner under its gate
(it never has), the adoption becomes donor-only and the gate must be
revisited.
- "panel not responsive": four members chunks queue back-to-back on the
one chart thread - 4 x 120ms = 480ms worst-case click latency, the
exact regime the 200ms note in Training.mqh already documents as
broken. Ensemble members now use a 30ms chunk budget, restoring solo
UI latency at slightly higher dispatch overhead.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
All four ensemble members previously wrote their full multi-line panels
to the SAME global label objects - an ensemble chart would flicker
between four stacked panels covering the chart side (user request:
aggregate). Every AI-side SetStatusLabel call site now routes through
CExpertSignalAIBase::PublishStatus - solo charts draw the full panel
exactly as before; an ensemble member claims a slot and contributes
only its HEADLINE to one combined block ("HYBRID ensemble - N models",
then one line per model; the live line leads with the model current
signal). The combined render skips unchanged text and enforces its own
minimum redraw interval so four publishers cannot multiply
ChartRedraw() cost.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The user is right that no special combination logic is needed: the AI
signals are ordinary voting filters, and the aggregate already has
union semantics - abstaining filters do not dilute the average, so an
ensemble chart trades whenever ANY deployed member clears the vote
threshold and disagreeing members net out. What the ensemble preset
actually adds:
- AI_CHOICE value 4 renamed AI_CONVLSTM (the name says the front-end);
enum VALUES stable, CSignalHYBRID class and State\HYBRID\ folder kept,
so saved configs and trained models keep their identity.
- New AI_HYBRID = 6: enables PAI+CONV+LSTM+CONVLSTM together on one
chart - replaces four separate charts of the same symbol. Each member
trains and self-gates independently; only certified members ever vote.
- |ENS1 fingerprint token on every member, so an ensemble member's
weight files can never collide with a solo model of identical
settings on another chart of the same symbol (the duplicate-chart
guard would otherwise correctly fight over one .nnw).
- Private default AIType = AI_HYBRID: one D1 drop now yields every
topology's gate verdict for that symbol.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The first-ever family-wise gate pass (SP500 D1 PAI, +10.4pp, p=0.0081)
certifies a win rate measured on HOLD-TO-RESOLUTION outcomes: entry,
then the measured SL or TP decides. Live, three vote-driven exit routes
could close earlier - the averaged-vote close, the AI early-exit route
(both in CheckClosePosition), and CheckReverse - and the fractal
target's vote flips at swing-marker cadence (~3-5 bars), far inside the
barrier's typical travel time (median 7-8 D1 bars to target). The user
observed exactly this: an opposite arrow near an entry, trade cut,
price kept going.
On a fractal-target chart with a live direction model, all three routes
are now suppressed (m_holdToBarrier, set in InitializeSignal, loudly
logged): positions run to their broker SL/TP. Risk guards and trailing
are deliberately untouched - account protection is not signal opinion.
Barrier-target models keep the vote exits: their label is the vote's
own horizon, so for them the routes are semantically consistent.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
User direction (2026-08-15): back to predicting swing turns, D1 charts,
fractals over ZigZag pivots (their call - balances classes, matches the
reference library target, and a 5-bar fractal confirms 2 bars after its
extreme so labels resolve nearly to the present with no repaint embargo).
- TRAINING_TARGET enum + TrainingTarget input: TARGET_BARRIER (Market
default - existing models keep their meaning and fingerprints) or
TARGET_FRACTAL (private default).
- FractalDirectionLabel (Labels.mqh): per-bar 3-class label = direction
from the bar close to the next confirmed strict 5-bar fractal extreme,
costs charged in the same bid-series convention as the barrier label,
Neutral when the move cannot clear max(2 spreads, 0.10 ATR) or on an
outside bar (both-extreme bars are unorderable within OHLC).
- The barrier walk still runs in full: measured SL/TP geometry, the
expectancy scan, excursion caches and the era gate all keep scoring
what a trade at the EA's own stop/target actually collected - only the
TRAINING label changes. NOT the pre-b4a704d "is this bar the pivot"
form; that target's 31:1 imbalance stays retired.
- Fingerprint token |TGT:FRA1 so switching targets trains a separate
model; AI_META unaffected (guarded setter).
- Private defaults: AIType back to AI_HYBRID (direction topology needed)
+ TrainingTarget=TARGET_FRACTAL = drop-on-D1-chart workflow.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The user should not need a tester corpus run per symbol. Every pattern
condition in Signals\Signal{MA,RSI,MACD,Ichimoku}.mqh anchors its reads on
`int idx = StartIndex()` with zero hardcoded indices (verified), so a
name-hiding StartIndex override + EvalShift(i) on CExpertSignalCustom makes
the EXACT live ladder code answer "what would you have fired at bar i" -
the silent-divergence trap that justified the DB corpus does not exist on
this path, and neither do the GMT-offset ambiguity, the DB row caps, or
the wipe procedure.
- CExpertSignalCustom: m_evalShift + StartIndex()/EvalShift() +
SweepPrepare(bars) (deep-resizes the shared price series); the four
classic signal classes override SweepPrepare to deep-resize their own
indicator buffers.
- CSignalMETA::BuildCorpusBySweep: per bar x per source filter, run
Direction() shifted, harvest the per-side pattern slots + netVote into
the same corpus arrays the DB loader fills; entry=bar open so
MetaPrepareEra's resolution matches at offset +0 with zero price error.
DB corpus remains the fallback when classic filters are disabled.
- Warrior_EA.mq5: META gets the enabled classic filters as candidate
sources (family ids match the descriptor one-hot).
- UseDatabaseRanking default false -> true (user request): a META chart
journals + ranks out of the box.
Workflow per symbol is now: attach ONE chart with AIType=META (optionally
Meta_ExportDataset=true for the offline pool) - candidates, labels,
training and export all happen in place, ~10 seconds of sweep instead of a
tester run.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A resumed META model hot-looped pass 1 (0->100% scan oscillation, silent for
3 minutes until the stall reporter fired) because EVERY window failed at the
first AD/Wyckoff feature: the init-time param adoption called
ReInitADIndicators unconditionally, destroying five freshly-calculating
indicator instances to recreate them with BYTE-IDENTICAL params (verified by
parsing the .nnw header - the MI tuner had kept the configured settings), at
process start, on a box with 1 GB free of 31. The replacements sat cold for
6+ minutes while full-history resweeps starved the indicator threads harder.
- AdoptIndicatorParams: installs a loaded param set into the tuner and
rebuilds handles ONLY when the set actually differs from what the live
indicators run. Both call sites (resume init + panel reload) use it.
- Resumed models get the same 3 warm-up passes as fresh ones. The skip was
the shared root cause of the cold-ATR (ba13eef), cold-AD (2026-08-11) and
this incident - custom indicators recompute from scratch every process
start regardless of what the .nnw proves.
- Cold-sweep backoff: a pass-1 sweep in which every window failed on a
TRANSIENT cause arms a 5s era-start pause instead of an immediate
full-history resweep, so the retry loop stops consuming the CPU/memory the
warming indicators need. The stall reporter names the backoff branch.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
350-era S2 verdict on SP500 H1: the meta head carries REAL ranking skill
(+1.0-1.3pp mean over base, 101/350 eras clear their own 2-sigma bar, traded
subset wins 66.1% at <30% coverage vs 64.5% base) but 0/350 eras produced a
positive cov x (p - BE): the candidate stream sits 3pp under the derived
geometry's 67.5% break-even and ~2.6pp of recovered skill cannot bridge it.
Skill plateaued by mid-run (1.28pp -> 1.05pp), so more eras only buy
multiplicity, and the deploy gate correctly shipped nothing.
The aggregate can hide a deployable subset (one family/side clearing BE
blended with junk), so the META era line now decomposes the SAME traded
population into MA/RSI/MACD/Ichimoku x LONG/SHORT cells, each as
traded/candidates base->traded win rate. 32 cells is a best-of-N search by
construction - any candidate cell faces the family-wise rule before belief.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The NN now has a target that is not per-bar direction (closed, best-of-999
p=1.0000): P(win | this journaled candidate, at the EA's own SL/TP, net of
cost). One net for all 52 pattern-sides, AIType=AI_META.
- NetForward.mqh: the host-side softmax+CE gradient generalized total==3 ->
2||3 on both backprop paths; a 2-class softmax IS a logistic head, and no
compute backend changes.
- SignalMETA.mqh (new): corpus loaded read-only from the LARGEST signal DB on
disk (decoupled from the config fingerprint that burned four S1 runs); the
GMT->server offset is measured PER ROW against entryPrice vs bar open
(DST-immune, histogram logged); a window-span regime filter drops the
pre-2017 daily-backfill rows; 31-feature setup descriptor appended at the
input (26 one-hot + side + tanh netVote + SL/TP ATR + spread/ATR).
- Training.mqh: candidate-queued pass 1, binary-target pass 2, per-candidate
calibration (2.5) and OOS (3) walks. Counter mapping win->Buy / loss->Sell
lets checkpoint selection, the edge floor, the plateau ladder and the
family-wise deploy gate run UNCHANGED: precision reads as win rate among
traded candidates, chance as the base win rate, recalls as sensitivity/
specificity. Era-end META line: coverage x (p - break-even) vs the null.
- Labels are the side-conditional triple-barrier win caches - never the DB's
stop-and-reverse outcome. Logit adjustment deliberately skipped (~40% base
rate). Live inference + online learning guarded off until S3.
- Fingerprint: conditional |TGT:META1; State\META\ folder + 2-output filename
slot keep meta models fully separate from direction models.
Compiles clean (0 errors, 0 warnings). S2 run = attach a chart with
AIType=AI_META; S3 wires the votes via the per-side hooks.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The warning lived inside the VerboseMode-gated corpus report, so a
forgotten wipe silently voided an entire 18-year corpus run - the
outdated-row guard rejected the whole replay against leftover rows
and the run appended 35 rows instead of building a corpus. The check
now runs unconditionally at tester OnInit (MetaCorpusStaleCheck): 52
quiet one-row newest-key probes vs the test start, with a loud stop-
wipe-rerun instruction when the DB is newer than the test. Absent
tables probe quietly via FetchNewestTimeKey''s new quiet flag.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The historical 1000-row cap existed for a real reason: ProcessSignal
pulled BOTH full tables into MQL struct arrays on every buffered
signal, and UpdateSignalsWeights pulled all 52 per cycle -
materializing thousands of string-bearing structs per event is the
practical limit the cap protected against (SQLite itself has none).
Raising the cap for an 18-year meta-label corpus build would have
made runs crawl; sharding across databases would re-read the same
rows and inherit the same cost.
Every question is now answered inside SQLite, one row or one number
per query, flat in table size:
- FetchOpenTradeEntry: the open (NA) trade''s entryPrice for
pattern+direction, LIMIT 1
- FetchNewestTimeKey: newest row''s yyyymmddhhmm via max ROWID
(rows insert chronologically) - the duplicate/outdated guard
- FetchWinLossCounts: COALESCE''d SUM aggregates with the
before-now bound applied in SQL, replacing the tester-only array
trim (now also active live, where it is harmless by construction)
ProcessSignal semantics preserved exactly: prune -> close opposite
(stop-and-reverse still registers its own row) -> duplicate/outdated
-> one-open-trade -> register. CalculatePatternWinRate''s array walk
becomes WinRateFromCounts; the private FetchTradeRecords wrapper and
ShouldDeleteOldestEntry are gone. DB_MaxRowsPerTable=20000 is now
cheap at any table size.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The first corpus build produced 3,681 rows, all 2026, from an 18-year
backtest: ProcessSignal''s outdated-row guard rejects any registration
older than a row its table already holds (correct for a live stream),
so a tester run starting before the leftover rows'' dates silently
registers nothing for the overlap. The corpus report now prints a
loud WARNING when running in the tester with DB rows newer than the
test''s start: corpus builds start from an empty DB.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Implements stage S1 of Meta_Labeling_Design.md, superseding the
original "training-time ladder sweep": the per-side journaling from
652bf81/195be20 already produces the exact candidate stream a sweep
would compute - every pattern instance the live ladders fire, both
sides, uncensored, with netVote and touchable entry price - so the
corpus is READ from the DB instead of re-implementing 26 ladder
conditions in training code. That eliminates the silent-divergence
trap outright: the corpus is by construction identical to live
behaviour. Accepted costs are documented in the module and the doc:
coverage equals the populating backtest, and sampling is one
candidate per fire-stretch (the right dedup for training anyway).
- Expert\AIBase\MetaCorpus.mqh: CMetaCorpus reader (52 tables ->
SMetaCandidate rows) + VerboseMode OnInit report: volume/closed/
S&R-win-rate per family, span, and the GMT->server bar-offset
match table (offsets +0..+3h) that S2''s label plumbing pins to -
measured, not assumed.
- DB_MaxRowsPerTable input (default 1000 = old MAX_TABLE_ROWS): a
corpus build raises it (e.g. 20000) so a 15-20 year backtest
isn''t pruned; wired through CExpertSignalCustom::MaxTableRows().
- Report-only stage: nothing downstream consumes the corpus yet.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The DB system logs objectively; the decision layer reads it to compute
win rates and adjust weights. The journaling path still had one
decision-layer tendril: rows were only written when the root''s
OpenLongParams()/OpenShortParams() succeeded. Those calls validate
ORDER PLACEMENT (broker stops-level, ATR warm-up, entry-mode
rejection) and their failures cluster in volatility/spread conditions,
so the gate non-randomly censored exactly those bars out of every
pattern''s win-rate sample - the same censoring class 652bf81/c8ef478
removed, one layer down. The ledger never needed placement to be
possible: entries are marked at the touchable side of the spread and
exits are same-pattern reversals, not broker fills.
Also documents netVote for what it is: a record of the decision
layer''s state at log time (per-pattern weights inside it drift as
ranking updates land), not an objective measure - the objective part
of a row is pattern/direction/price/result.
SIGNAL_DB_SEMANTICS_VERSION 4 -> 5: row populations gain the
previously censored bars, so the database re-keys.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Verification of 652bf81 on a fresh 7-month backtest DB surfaced the
last one-sided mechanism: ProcessSignal absorbed a reversing signal as
the exit of the opposite trade and skipped registering it. For pure
EVENT patterns that strictly alternate (MACD model 3, the zero-line
cross), every reversal was consumed and the whole ledger landed on
whichever side fired first - 60 Buy rows, 0 Sell rows - so the silent
side never earned a win rate and UpdateSignalsWeights() weighted the
pattern from one side only. State patterns escaped by re-firing one
bar later.
The reversal now closes the opposite trade AND registers its own row;
the existing duplicate/outdated/open-trade checks still bound the
table at one open trade per pattern+side. Row populations change
meaning, so SIGNAL_DB_SEMANTICS_VERSION 3 -> 4 re-keys the database
(the v3 file is orphaned, not wiped - schema is unchanged).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The labelMatchesVote gate compared a single last-writer-wins label
(LongCondition then ShortCondition) against the net vote sign, which
structurally censored the pattern tables: a long event co-occurring
with any short-side state model lost its label to the later writer and
was dropped, while the mirrored short event journaled fine. Ichimoku
models 0/3 and MA model 1 could not produce a row at all by
construction (MA model 1 was "revived" in 8710240 yet still could
never journal - its weight-10 vote is exactly cancelled by the
opposing Pattern_0 state), and every pattern's win rate was measured
on a with-trend-only subset - the exact statistic
UpdateSignalsWeights() feeds back into the weights, self-sealing:
no rows -> no win rate -> default weight -> still censored.
- Direction() now evaluates the two ladders separately and snapshots
each ladder's matched pattern into its own side slot; each side that
matched journals its own row. The flat-vote poisoning the old gate
fixed stays fixed: a label can no longer contradict its side.
- The filter's net vote (raw pattern-weight units) is stored as a new
netVote column - data, never a drop filter. Snapshot is keyed on the
ladder setting a label, not on its weight, so a 0%-win-rate pattern
keeps journaling and can recover.
- SIGNAL_DB_SEMANTICS_VERSION is folded unconditionally into the DB
filename fingerprint: pattern-definition changes (b2069bc, 8710240)
re-key the database instead of blending incompatible Pattern_N
populations under one key, which the input-hash fingerprint cannot
see. 7 months of mixed-semantics rows shared one file because of it.
- dbVersion 2.0 -> 3.0: schema changed, and inserts carry the new
column, so the version-mismatch folder wipe is the migration.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The reference-pair set was re-discovered from Market Watch on every
build, so adding or removing a terminal symbol silently changed what a
trained model's six cross-asset features meant - the last open
train/serve parity gap from the 2026-08-11 audit. The set a model's
FIRST successful build actually used is now stamped into its .cfg
(append-and-length-guard, adopt-don't-compare - the derived-barrier
pattern) and every later build constructs the panel from exactly that
list; a pinned pair that is temporarily unavailable is skipped, never
substituted.
Also warms SymbolSelect/SeriesInfo for every reference symbol at
InitNeuralNetwork, so the terminal's ~minute of async cross-symbol
download starts at init instead of when the first Build() trips over
an unselected symbol - the source of the startup 'only 0 usable
reference pairs' console failures.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
On a CFD whose base and quote currency match (SP500 -> USD/USD) the FX
encoding degenerated: base and quote strength were the SAME series twice
and the divergence feature collapsed to the symbol's own 20-bar return.
Index mode re-encodes the six slots: denomination-currency strength
(fast/slow), a risk-proxy currency's strength (JPY by fixed preference
order - deterministic across rebuilds), and divergence as own move minus
what the denomination alone implies. FX-pair symbols are untouched.
Fingerprint gains :IDX2 for base==quote symbols only, so index models
trained under the degenerate encoding re-key while FX models keep their
filenames. FORCES RETRAIN on index/CFD charts.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
passTrail demanded m_excTrailScored >= EXCURSION_MIN_SCORED (500), but since
e2c9593 the trail race only scores DISJOINT bars: m_excTrailScored is bounded
by m_excScoredD (~OOS/horizon ~= 256 on SP500 H1) minus the post-ring-clear
warm-up (~8), so every chart failed "[trailing incumbent not warm enough to
race]" at 247-248 of a possible ~256 forever - observed live 2026-08-11 on
all four charts. The counter's statistical population is the same disjoint
sample passDj gates on, so it now takes the same minimum
(EXCURSION_MIN_DISJOINT, 200), reachable with margin after warm-up.
Compile: 0 errors, 0 warnings.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three findings from the 2026-08-11 audit:
1. The excursion head's trailing-quantile ring was deliberately never cleared
between eras ("a rolling estimate of the market, not of the era") - but
pass 3 re-walks the SAME OOS window every era, so at each walk's restart
the ring still held the outcome masks of the newest OOS bars from the
previous walk: the chronological FUTURE of the bars about to be scored.
For the first ~window+horizon pushes of every era the "trailing" incumbent
was partly a leading one - conservative for the gate (an informed incumbent
is a harder hurdle) but exactly the self-made-artifact class 06d4785 hunts.
The ring now clears at era-score reset; the warm-up bars simply don't score
the trail race, which the m_excTrailN gating already accounts for.
2. skillTrail compared the head's FULL-block Brier (pro-rated by coverage)
against the incumbent's subset sum - valid only if head skill is uniform
across the OOS walk, while the trail-scored subset systematically excludes
each era's warm-up bars. The audit also found m_excBrierHeadD/BaseD/
m_excOosHitsD declared, zeroed and never accumulated (dead since e2c9593
made every scored bar disjoint). The dead trio is replaced by
m_excBrierHeadT: the head's Brier accumulated only on the bars the warm
incumbent also scored, so the race now compares both predictors on an
identical bar set.
3. The AD/Wyckoff feature blocks read GetData with no EMPTY_VALUE guard; a
cold (still-calculating) indicator returns EMPTY_VALUE everywhere, the
sanitize loop rewrote that to 0.0, and the bar SUCCEEDED - so
BufferTempData cached an all-zero Wyckoff block as a success for the whole
bar frame: the one path the f6150ee only-cache-successes rule cannot see,
because it never fails (the ba13eef class, arriving through values that
never fail; a resumed model's era-0 prebuild starts milliseconds after
OnInit). ADIndicatorCold() probes the NEWEST bar - EMPTY_VALUE there means
async warm-up (transient reject, retried), while deep bars beyond the
buffered depth keep the sanitize loop's neutral-fill so degraded history
still trains. Also fixed m_featureCacheValid's declaration comment, which
still described the pre-f6150ee cached-miss semantics.
Compile: 0 errors, 0 warnings.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
