TRADE LAB — what we trade, what to skip

check	result	if failing
✅ join_integrity	all returned price keys match the requested keys
✅ price_coverage	all chains priceable
✅ unresolved_rate	2% unresolved over 48h
🟡 ledger_arithmetic	0 pnl mismatches, 8 rows exit before entry	assert exit_ts>=entry_ts and pnl==notional*ret at WRITE time (#431)
✅ freshness	journal 3.5h old
❌ control_arm	only 2 control trades in 7d	without a live control the scoreboard cannot be validated
🟡 dead_fields	1 previously-flagged field(s) still null/empty (already baselined, awaiting triage)
✅ metric_liveness	registered KPI changes and provenance are live

WINDOW

Funnel · seen vs taken

What the system saw, and what it chose

Outcome mix

What happened to what we took

Grey UNRESOLVED = no outcome at all. It is shown, never dropped — a hidden slice here is what made a 57%-unmeasured book look confident for two weeks.

Allocation

Share of capital deployed, by trade type

Where the money went. Compare against the chart on the right — a type taking a big slice here and a small one there is overfunded.

Earning its slot?

Share of capital vs share of profit

My addition, and the chart I'd watch daily. Teal = share of P&L, grey = share of capital. Teal shorter than grey means that type consumes more budget than it returns. This is the allocation question a pie cannot answer.

P&L composition

Where the money actually moved

Slice size = |P&L|, colour = direction (teal profit / red loss). Slices therefore sum to gross money moved — the net is in the centre, because those are different numbers and conflating them is the one way this chart can mislead.

Volume vs result

Capital deployed → what came back

Same slices, sized by capital deployed instead, coloured by whether that capital came back. Read side-by-side with the left: a type that is large here and red there is where the budget is being burned.

Precise signed magnitudes

P&L and expectancy by trade type

The exact figures behind the two donuts above. Bar length = P&L, label = expectancy per trade and sample size. Use the donuts to spot where to look; use this to confirm how big and on how many trades.

Decision quality over time

Is the selector getting better?

Expectancy per day across the window. Flat-and-negative means selection is not learning — which is expected until an entry feedback loop exists.

One paragraph

The whole problem, stated once

We take 165 trades a day and lose -5.9% on each one. The losses are not spread evenly — they are concentrated in trades that die within an hour (-$141,908 of the -$141,038 total). Two mechanical defects cause most of it: a stop that cannot fill where it is set because exits run on a 5-minute cron, and no quality gate on entry — 7% of what we block is blocked for capacity, only 8% for quality. Fix those two and the book moves roughly $110,675 without finding a single new signal.

Defect 01

There is no resting order

-$73,889

Stop set −20%; STOP_GAP trades realise -63.3%.
The stop is not an order — it is a condition tested every 5 min. Code books fill = cur*(1−slip): the polled price, not the level.
Gap = detection lag (fixable) + real on-chain slippage (not). $73,889 is an upper bound, not a forecast — plan on ~half.
Fix: 1-min cron (#185) — measured runtime is only 10s / 291 positions, and prices are one batch call per chain.

Defect 02

No quality gate

-$36,786

864,777 signals blocked — but 7% only because we were full.
Only 69,440 (8%) were refused on merit.
3 sources lose to a random control and cost $36,786.
Fix: gate on tail-rate vs control, not on free slots (queue #187).

What already works

Don't break these

Tail wins carry the book: 7% of trades produce $166,260.
Holding ≥2h is flat-to-positive at every bucket.
ETH and Base are the only chains not bleeding.
Never cut the tail to raise win rate — the tail is the edge.

Skip list

Stop trading these now

Any exit inside 60 minutes that isn't a real stop.
Sources marked CUT: momentum_scalp, sniper, trench.
Hours 03, 10, 17, 22 UTC — worst expectancy.
Regime extreme_poor — -6.7% expectancy.

Grounding

Where every number here comes from

Source: paper_journal.jsonl — 24,327 journalled positions, 93 trading days.
Excluded: 7,224 rows whose outcome is unknowable because the price feed died (the #181 bug booked these as −100%). Counting them would overstate losses by roughly a third.
Excluded: runner-bag legs below $60 — they are a second leg of one decision, not a new decision.
Anomaly: 67 rows have exit_ts < entry_ts (negative hold). They are counted in the totals but kept out of the sub-hour bucket, which they would otherwise inflate from 91% to 95% of losses. Separate clock bug — worth its own ticket.
Control cohort: n=249 random entries, expectancy -18.1%, tail 5.6% — any source that cannot beat this has no demonstrated edge.

The question

Do we enter every trade that comes into the system?

No — but not for the right reasons. Of 889,104 evaluations, 864,777 were blocked. The system therefore looks selective. But when you split the block reasons, 7% are capacity limits — we were already full, so the signal was dropped regardless of quality. Only 8% were rejected because the setup was judged bad.

That means selection is effectively arbitrary with respect to quality: which trade gets in depends on when it arrived and whether a slot was free, not on whether it was good. That is why the book tracks the random control so closely.

Funnel

Where signals go

evaluations	889,104
blocked	864,777
entered	24,327
scored here	15,370

Block reasons

Capacity is doing the selecting, not judgement

reason	n	% of blocks	kind	what it means
strict_gate_reject	367,254	42.5%	DATA	—
no_price	178,951	20.7%	DATA	no price available — cannot enter
mc_below_floor	59,946	6.9%	DATA	—
admission_reject	57,779	6.7%	QUALITY	admission gate judged it bad
mc_resolved_dexscr	47,257	5.5%	DATA	—
quality_tx_unavailable	46,518	5.4%	DATA	—
src_cap	45,177	5.2%	CAPACITY	per-source slot limit hit
total_cap	12,277	1.4%	CAPACITY	global open-position limit hit
decision_reject	11,661	1.3%	QUALITY	decision engine refused
sm_distrib	11,508	1.3%	DATA	smart money selling into it
fomo_large_tx_reject	8,798	1.0%	DATA	—
quality_tx_reject	7,777	0.9%	DATA	—
throttled	3,293	0.4%	CAPACITY	rate limiter
source_cap	2,398	0.3%	DATA	—
shadow_mc_cap	2,220	0.3%	DATA	—
young_defer	956	0.1%	DATA	—
nonpositive_or_invalid_size	821	0.1%	DATA	—
oggy_net_cap	144	0.0%	DATA	—
shadow_mc_deadline	25	0.0%	DATA	—
buys_band	17	0.0%	DATA	—

Taxonomy

Every trade classified by what actually happened to it

verdict	n	% book	expectancy	win	P&L
NEVER_ARMED	8654	56.3%	-19.4%	8%	-$175,913
STOP_GAP	1596	10.4%	-63.3%	0%	-$108,047
ROUND_TRIP	1366	8.9%	-26.8%	0%	-$29,077
CLEAN_WIN	1256	8.2%	+27.5%	100%	$29,602
TAIL_WIN	1050	6.8%	+198.0%	100%	$166,260
BE_STOP_SAVE	982	6.4%	-9.5%	0%	-$9,730
DEAD_ZONE_ARM	443	2.9%	-38.0%	0%	-$13,457
BE_STOP_EARLY	12	0.1%	-40.2%	0%	-$529
TIME_STOP	6	0.0%	-4.6%	0%	-$29
ORDINARY_LOSS	3	0.0%	-38.3%	0%	-$115
SMART_MONEY_EXIT	2	0.0%	-1.1%	0%	-$3

NEVER_ARMED — never even reached +5%. Entry was simply wrong. Biggest single bucket.
STOP_GAP — stop triggered but filled far below the level. Mechanical, fixable.
ROUND_TRIP — went up, gave it all back. Exit-timing problem.
DEAD_ZONE_ARM — armed, then died between arm and trail. Arm threshold too low.
TAIL_WIN / CLEAN_WIN — the entire positive side of the book.

By chain

Where it works

chain	n	exp	win	P&L
sol	7220	-8.4%	20%	-$91,797
bsc	3442	-7.6%	20%	-$35,883
base	1919	-0.1%	21%	-$4,609
eth	1495	+2.3%	19%	-$411
robinhood	1292	-5.5%	16%	-$8,297

By regime

Market state at entry

regime	n	exp	win	P&L
poor	7847	-8.5%	19%	-$94,053
ground	6907	-2.9%	19%	-$41,117
extreme_poor	616	-6.7%	18%	-$5,867

Edge test

Every source vs a random control cohort

A source only has edge if it beats random entry (n=249, expectancy -18.1%, tail 5.6%). Win rate is deliberately not the deciding column — tail rate is, because 7% of trades generate all the profit.

source	n	expectancy	win	tail	vs control	P&L	call
volume_path	2005	-1.5%	18%	1.1%	+16.6pp	-$4,659	WATCH
fomo	1738	-3.9%	20%	6.3%	+14.2pp	-$18,271	KEEP
clu:fomo	1291	+0.8%	23%	3.9%	+18.9pp	-$4,742	WATCH
momentum_scalp	893	-30.0%	12%	3.9%	-11.9pp	-$29,555	CUT
whalemover_rh	815	-3.5%	14%	0.9%	+14.6pp	-$3,423	WATCH
whalemover	767	-2.2%	22%	3.0%	+15.8pp	-$3,584	WATCH
known_conv	765	-5.2%	15%	1.6%	+12.8pp	-$4,460	WATCH
sa_accum	700	-2.5%	19%	3.0%	+15.6pp	-$3,857	WATCH
source_strategy	669	-3.4%	16%	2.1%	+14.7pp	-$3,452	WATCH
oggy_kol	633	-9.7%	19%	4.7%	+8.3pp	-$9,414	WATCH
smartmoney	495	-8.1%	16%	5.1%	+9.9pp	-$6,938	WATCH
cluster	462	-10.0%	20%	6.1%	+8.1pp	-$7,549	KEEP
fomo_curated	417	-14.9%	19%	4.6%	+3.2pp	-$7,505	WATCH
accum_engine	388	-4.0%	26%	3.6%	+14.1pp	-$2,789	WATCH
wallet_intel	370	-1.9%	22%	3.2%	+16.2pp	-$1,488	WATCH
wallet_buysize	322	+11.8%	27%	5.6%	+29.9pp	$1,638	WATCH
trending	310	-13.5%	16%	1.6%	+4.6pp	-$4,722	WATCH
control	249	-18.1%	14%	5.6%	+0.0pp	-$5,185	WATCH
alpha_brief	216	-8.7%	20%	2.3%	+9.4pp	-$2,190	WATCH
oggy_boost	198	-0.0%	31%	10.1%	+18.0pp	-$985	KEEP
cipher	183	-4.8%	25%	6.6%	+13.2pp	-$1,765	KEEP
sniper	163	-27.3%	17%	2.5%	-9.3pp	-$4,795	CUT
clu:dowville	160	-0.4%	26%	2.5%	+17.7pp	-$437	WATCH
volume_spike	141	-5.3%	26%	2.8%	+12.7pp	-$1,107	WATCH
inversion_alpha	110	-13.0%	23%	0.9%	+5.1pp	-$1,741	WATCH
floor_revival	107	-6.4%	21%	6.5%	+11.7pp	-$1,179	KEEP
trench	91	-22.0%	22%	4.4%	-3.9pp	-$2,436	CUT
kol:michacalls	81	+0.4%	25%	4.9%	+18.4pp	-$503	WATCH
clu:sapphy	78	-10.5%	18%	2.6%	+7.5pp	-$1,067	WATCH
accum_sig_gmgn_cluster	55	+12.8%	31%	7.3%	+30.9pp	$440	KEEP
clu:shadow_dowville	52	-4.4%	25%	1.9%	+13.6pp	-$342	WATCH
accum_sig_fresh_swarm	52	+11.9%	33%	7.7%	+30.0pp	$405	KEEP
accum_sig_flow_div	47	+4.7%	28%	4.3%	+22.8pp	$45	WATCH
accum_sig_stealth	46	+7.0%	26%	4.3%	+25.0pp	$131	WATCH
kien_conv	40	-8.4%	20%	7.5%	+9.7pp	-$525	KEEP
leading_conv	40	-3.3%	5%	0.0%	+14.7pp	-$133	WATCH

The single biggest finding

Hold time decides everything

hold	n	expectancy	win	tail	P&L
<1h	7831	-16.9%	12%	1.3%	-$141,908
1-2h	1267	+3.1%	23%	7.4%	-$3,912
2-4h	1083	+5.6%	25%	7.5%	$182
4-6h	579	+13.5%	27%	5.4%	$3,857
6-24h	1995	+1.7%	27%	5.9%	-$8,053
1-7d	1595	+15.4%	31%	7.5%	$13,338
>7d	953	-0.6%	31%	0.8%	-$978

Sub-hour trades are 51% of the book and 101% of the losses.
Every bucket ≥2h is flat or positive. The edge is not in speed.
This is the cheapest fix available: stop taking trades that resolve in under an hour.

Entry hour · UTC

Expectancy by hour of day

Bar height = |expectancy|. Teal above zero, red below. Best: 00:00 (+3%), 09:00 (+1%), 14:00 (+0%). Worst: 03:00 (-13%), 10:00 (-12%), 17:00 (-11%). Treat hour-of-day as a tiebreaker only — the sample per hour is 409–899 trades and these gaps are not yet significant.

Per-trade record

What / why / when — worst offenders in each class

Full journal: paper_journal.jsonl (24,327 entries) · rendered at /journal.html. Below: the two costliest trades in each failure class, each with the machine-written why and the fix it implies.

STOP_GAP

CAPITALISMcontrolsol-200.0%-$220

WHEN 07-11 03:12 UTC · held 5m · peak -2.4% · size $110

WHY Stop set at -40% but filled at -200.0% — gapped 160 points through the stop in a 5-minute polling window.

FIX A 5-min poll cannot honour a -40% stop on this volatility. Either widen the stop to match reality (-200%) and size down, or move exits to a price-triggered watcher/limit order rather than a cron poll.

43oggy_boosteth-200.0%-$220

WHEN 07-14 11:13 UTC · held 388m · peak +3.1% · size $110

WHY Stop set at -45% but filled at -200.0% — gapped 155 points through the stop in a 5-minute polling window.

FIX A 5-min poll cannot honour a -45% stop on this volatility. Either widen the stop to match reality (-200%) and size down, or move exits to a price-triggered watcher/limit order rather than a cron poll.

NEVER_ARMED

TROLLclu:fomosol-103.0%-$85

WHEN 09-20 12:47 UTC · held 7407m · peak +21.8% · size $82

WHY Never traded above +25% after entry (peak +22%). No follow-through at all — the entry had no edge; price went against us from the first poll.

FIX This is an ENTRY problem, not an exit problem. Gate the 'clu:fomo' source on a confirmation bar (require +25% within 10 min of signal before committing full size); enter on a starter clip only.

billioggy_kolsol-59.5%-$71

WHEN 07-08 16:03 UTC · held 14m · peak -1.0% · size $120

WHY Never traded above +25% after entry (peak -1%). No follow-through at all — the entry had no edge; price went against us from the first poll.

FIX This is an ENTRY problem, not an exit problem. Gate the 'oggy_kol' source on a confirmation bar (require +25% within 10 min of signal before committing full size); enter on a starter clip only.

ROUND_TRIP

27oggy_boosteth-133.3%-$116

WHEN 07-29 07:52 UTC · held 120m · peak +91.8% · size $87

WHY Reached +92% (a genuinely profitable peak, breakeven was +42.9%) and still closed -133.3%. Gave back 225 points of open profit.

FIX Bank a scale-out: sell 1/3 at +43% to lock the trade green, then trail the rest. A single all-or-nothing trail converts winners into losers.

14oggy_boosteth-133.3%-$116

WHEN 08-01 11:03 UTC · held 199m · peak +80.4% · size $87

WHY Reached +80% (a genuinely profitable peak, breakeven was +42.9%) and still closed -133.3%. Gave back 214 points of open profit.

FIX Bank a scale-out: sell 1/3 at +43% to lock the trade green, then trail the rest. A single all-or-nothing trail converts winners into losers.

TAIL_WIN

DIMwallet_buysizeeth+3870.0%$3,193

WHEN 07-26 00:53 UTC · held 295m · peak +5270.4% · size $82

WHY Caught a power-law runner: peaked +5270%, banked 3870.0% ($+3192.74). This cohort is where the entire book's positive expectancy lives.

FIX Protect this cohort. Widen the trail once a trade clears +150% so a tail runner is never stopped on normal volatility; consider a 25% moon-bag with no trail at all.

slingclu:fomosol+4116.4%$3,087

WHEN 08-09 13:22 UTC · held 370m · peak +4901.1% · size $75

WHY Caught a power-law runner: peaked +4901%, banked 4116.4% ($+3087.29). This cohort is where the entire book's positive expectancy lives.

FIX Protect this cohort. Widen the trail once a trade clears +150% so a tail runner is never stopped on normal volatility; consider a 25% moon-bag with no trail at all.

DEAD_ZONE_ARM

SOULmomentum_scalpsol-133.3%-$110

WHEN 07-09 23:32 UTC · held 15m · peak +36.1% · size $82

WHY Armed at +36% (profile arms at +25%) but the 30% trail needs a peak above +42.9% to break even. Peaked +36%, so the trailing exit was priced below entry the moment it armed. Realized -133.3%.

FIX Raise arm above breakeven: set arm >= 48% for a 30% trail, or cut the trail to 20%. Never arm below 1/(1-trail)-1.

BRIDGE BRIDGE momentum_scalpsol-133.3%-$110

WHEN 07-12 11:08 UTC · held 14m · peak +29.1% · size $82

WHY Armed at +29% (profile arms at +25%) but the 30% trail needs a peak above +42.9% to break even. Peaked +29%, so the trailing exit was priced below entry the moment it armed. Realized -133.3%.

FIX Raise arm above breakeven: set arm >= 48% for a 30% trail, or cut the trail to 20%. Never arm below 1/(1-trail)-1.

M-22 · FinallyNitin · shadow paper

Scanner arms — coverage before performance

Realized universal-paper exits only; selection lift is graded separately in alpha_feature_scoreboard.json. Decision effect: NONE.

arm	resolved / total	coverage	median return	promotion gate
M22_BURST_PROPENSITY_GE_5PCT	360/382	94.2%	-11%	WAIT_MATCHED_CONTROL_OOS_FDR
M22_NR7	250/272	91.9%	-3%	WAIT_MATCHED_CONTROL_OOS_FDR
M22_PPV10	172/186	92.5%	-3%	WAIT_MATCHED_CONTROL_OOS_FDR
M22_PPV10_NR7	6/8	75.0%	-3%	WAIT_COVERAGE
M22_PPV10_WR20_BREAKOUT	80/85	94.1%	-3%	WAIT_MATCHED_CONTROL_OOS_FDR
M22_SUPPLY_RECLAIM	388/409	94.9%	-3%	WAIT_MATCHED_CONTROL_OOS_FDR
M22_UNRESOLVED_SUPPLY	578/641	90.2%	-3%	WAIT_MATCHED_CONTROL_OOS_FDR
M22_WR20_BREAKOUT	122/128	95.3%	-3%	WAIT_MATCHED_CONTROL_OOS_FDR

The answer

Should there be a system that decides what we enter?

Yes — and the absence of one is the root cause, not a missing feature. Today entry is decided by arrival order and free slots. A real admission brain would score every candidate before it competes for a slot, and spend the day's risk budget on the best N rather than the first N.

What it decides: enter / skip, size, and which exit profile — not just yes/no.
What it scores on: source tail-rate vs control, chain, regime, hour, expected hold. Every one of those is already measured on this page.
How it learns: each decision is written to the journal with its score; weekly, the score→outcome relationship is refit and any feature that stops predicting is dropped. That loop already exists for exits (param_evaluator) — it has never been pointed at entries.
Guardrail: it must be graded against the random control cohort continuously. A brain that cannot beat random is worse than no brain, because it feels justified.
Expected gain: cutting the 3 below-control sources and sub-hour churn is worth ≈$36,786 and $141,908 respectively on this sample — before any new signal is added.

Scoreboard

Measure these, not win rate

Expectancy per trade (the only P&L-linked one)
Tail rate ≥+100% vs control
Payoff vs breakeven payoff
Stop-fill slippage (set vs realised)
% resolving <1h — the churn meter
Capacity-block rate — how often quality never got asked
Void rate — feed failures, so bugs can't hide as losses

Sequence

Order of work — each step makes the next measurable

01 · #181

Stop booking feed outages as −100%. Until this lands every other number is contaminated.

02 · #185

Move exits off the 5-min cron. Recovers ≈$73,889 and makes stop levels mean something.

03 · #187

Tail-rate-vs-control gate. Cuts 3 sources worth $36,786.

04 · new

Sub-hour churn filter — the largest single loss pool (-$141,908).

05 · new

Only then: the scoring brain, graded against control weekly.