📥 Accumulation

Chris core thesis — learn the patterns/indicators of accumulation-then-pump. 34 signals in last 7d · accum_engine v1.15.0 · generated 2026-10-12 05:31 UTC

🫀 Engine health & live rules

Status

🟢 OK

last run 3m ago · thread 92561

Picks / 24h

top DOPE 39

Candidates

this run

Capture lag (24h)

median chain→seen · p90 6m · n=1010
every signal starts this far behind

Live rules

alert ≥17 · cooldown 1h (re-fires if score +6) · coverage ≥40% · 24h window

cooldown 12h→1h + escalation (260726) — never mute strengthening accumulation

🏆 S/A-tier wallets accumulating

graded alpha wallets building over 24h · 🤝 ≥2 independent clusters · 🐋 one S-tier loading heavy · size-gated (≥1% mcap or 0.3 ETH)
snapshot 44m old (scan v1.2.0 · 04:47 UTC) — 0 LIVE · 5 WARM · 0 STALE. AGE is the last graded buy's on-chain time, not when we rendered this. BURST = span from first to last buy. LAG = how long after the on-chain buy our collector saw it.

	Token	Age	Burst	Lag	S/A	Repeat	Graded $	%mcap
🐋	$POD base	🟡 10.9h	8.6h	+1m	1S / 0A	25	$18,084	–
🐋	$尤栗 bsc	🟡 7.1h	5.9h	+42m	1S / 0A	14	$5,346	–
🐋	$STFY base	🟡 5.6h	15.0h	+2m	1S / 0A	12	$2,540	–
🐋	$Basecat base	🟡 5.6h	18.9h	+1m	1S / 0A	61	$1,756	–
🐋	$STONKEX base	🟡 5.6h	17.4h	+2m	1S / 0A	17	$1,048	–

🟢 LIVE ≤4h · 🟡 WARM ≤12h · 🔴 STALE — shown as evidence, but only LIVE/WARM are digested to Telegram and paper-traded. ⚡BURST = every graded buy inside 60 minutes: measured token-mean -22.7% vs a -10.0% baseline over 201 distinct tokens (permutation p=0.014, holds out of sample). It is a label, not a filter — split by arming the effect is p=0.0077 among never-armed but p=0.42 among ARMED, so it is absent where it would have paid. The lever that actually moves this book is ARMING (+28.6% vs -36.4% token-mean), not shape.

🔬 Which indicators predict wins

signature edge — win-rate vs 42% baseline over 1962 outcome-tracked signals (#225 phase-1 internal harvest)

✅ real edge

SUPPLY-CONTROL	48%	+6pp · n253
SQUEEZE	46%	+4pp · n240
FLOW-DIV	44%	+2pp · n437
FRESH-SWARM	43%	+1pp · n1827
GMGN-CLUSTER	43%	+1pp · n1827
CLUSTERED-SUPPLY	42%	+1pp · n807

❌ drag — down-weight

VOL-ANOM	28%	-14pp · n80
ABSORB	35%	-7pp · n155
BUNDLED-RISK	36%	-6pp · n209
REACCUM	38%	-4pp · n96
ASSOC-SUPPLY	42%	-0pp · n1734

📖 Method — WHAT / WHY / HOW / WHEN / WHOM

Canonical definition of accumulation and how the engine separates it from buy-then-dump. Spec: knowledge/accum_method_spec.md · content source: public/accum_method.json.

What it detects

≥3 independent buy events on the same CA within 24h, from ≥3 distinct source detectors, with price NOT rising proportionally — buying absorbed without markup, not a spike-chase.

min_fires=3 within a 24h window; the events must come from different detectors (accum, wallet_intel, smartmoney, stealth, trench, cipher) so it is a pattern, not one trader. MC/price staying flat while buying rises is the tell.

Why it matters

Cabals/whales accumulate at the bottom quietly, leave the chart to 'die', then it explodes. Asymmetry: enter at flat MC before the crowd, exit at the pump.

Chris thesis. SPACEMOON: peaked $197K/25 funders ~30d ago, decayed to $9K/9, never fired under an old window-only scan — hence the STEALTH-WAVE historical pass.

How it separates accumulation from buy-then-dump

Four independent defenses (STEALTH, RANGE, ABSORB/NET-IN, WASH/FARMING gates) plus flow structure confirmations. Buyers-minus-sellers is answered directly by ABSORB and NET-IN; wash and churn are gated out.

Mechanism	Checks	Net-vs-sell answer
STEALTH	≥4 fires, MC drift <1.5×	buying stays flat = hidden accumulation, not a run-up that then dumps
RANGE	fires across ≥3 distinct MC bands	re-buying the range, not one spike
ABSORB	buy_sell_ratio≥1.5 and flat MC	DIRECT buys-minus-sells ratio
NET-IN	net_buy_vol_1h_mcap ≥ 0.05	net inflow relative to mcap
WASH	wash_ratio>8000 → −6, ⚠️	strips fake buy/sell churn
FARMING-RISK	connected wallets sold ≥60% back → −5	labels churn as farming, not accumulation
NO-SELLS	no tracked wallet exits on the CA	watchlist shows no exits
FLOW-DIV	OBV/CVD up, 1h price ≤ +5%	flow in, price flat
HOLDER-GROW	holder +≥4%/hr, MC drift <1.3×	hands accumulating
LADDER	buys_ge_1k≥5, median buy <$800	size-tiered buying
ASSOC-SUPPLY	weighted % buy-$ from graded/serial-funder wallets	connected supply, not random

⚖️ Signatures with missing data go UNKNOWN/'blind' and are EXCLUDED from the score — never scored as absent. conviction is normalized over evaluable signatures only; coverage is reported alongside, never added in. 'coverage 100%' = every signature family had data.

When it fires

engine_score ≥ 17 (alert) and coverage ≥ 40%. Cooldown 1h with escalation override — re-alerts inside cooldown when score rises ≥ 6 (accumulation strengthening). 24h live window + 60d STEALTH-WAVE historical scan for wave-then-decay shapes.

Cooldown only prevents duplicate spam; it never mutes intensifying conviction, because market bottoms are where whales accumulate. Escalation override fires when a new graded cluster joins.

Whom it attributes the buying to

S/A-graded priority wallets + manually tracked side-cluster wallets, fresh/dormant/smart typed swarms, serial funders, and whalemovers. siren_level 5 when Sapphy + repeat-or-cross-cluster accumulation converges.

Side-cluster association is retained as intelligence even when volatile stats grade low. Siren is a challenger rule, outcome-tracked, not auto-traded.

⚖️ Measured edge — honest

Baseline real peak 287% over 1962 outcome-tracked signals. The signatures that sound purest (ABSORB, SUPPLY-CONTROL) currently measure worst at predicting pumps; BUNDLED-RISK/ASSOC-SUPPLY/STEALTH measure best by lift. Ranked by real-peak lift from accum_signature_edge.json (2026-08-27). The purest net-buy signatures (ABSORB −60, SUPPLY-CONTROL −74) currently measure WORST at predicting pumps. Loop exists; weights not yet fully reconciled. No signature is validated until reconciliation + FDR proof lands.

Method performance (paper)

Paper trades

402

Net EV $/trade

-8.6

CI low -12.5% · CONFIDENT

Payoff

1.59

win size ÷ loss size

p95 peak

188%

entry catches the pump

Win rate

29%

⚠️ The composite lane is measured negative (-8.58/trade over n=402, CONFIDENT) — big peaks (p95 188%) with payoff 1.6 are not being converted. The edge, if any, is in specific SIGNATURES (ranked below), not the composite score. Exit: SWING_ARM21 champion, promoted 260925.

Signature RANKING — real paper performance

Every accumulation signature, ranked by ACTUAL forward return on graded fires (not fire-frequency or setup-score — real money). This is the answer to “which signals actually work.”

Signature	Independent N	Net EV	CI low	Win rate	Payoff	Grade
REPEAT×40REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	5	+17.0%	-2.8%	80%	83.70	UNRELIABLE
REPEAT×50REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	5	+12.9%	+4.4%	100%	0.00	UNRELIABLE
REPEAT×49REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	5	+11.7%	+8.3%	100%	0.00	UNRELIABLE
REPEAT×32REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	5	+4.8%	-8.1%	80%	0.97	UNRELIABLE
REPEAT×31REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	9	+3.5%	-0.6%	67%	2.38	UNRELIABLE
REPEAT×58REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	4	+3.2%	-4.4%	100%	0.00	UNRELIABLE
REPEAT×39REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	7	+3.1%	-3.7%	71%	2.05	UNRELIABLE
REPEAT×41REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	9	+2.2%	-9.4%	33%	5.24	UNRELIABLE
REPEAT×4REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	189	+1.1%	-1.3%	50%	1.69	CONFIDENT
REPEAT×37REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	33	+1.1%	-1.0%	61%	1.63	LOW
REPEAT×56REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	5	+0.9%	-3.0%	80%	0.00	UNRELIABLE
REPEAT×17REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	26	+0.8%	-1.8%	69%	1.95	UNRELIABLE
REPEAT×29REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	3	+0.5%	-4.8%	67%	1.92	UNRELIABLE
REPEAT×55REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	5	+0.2%	-3.8%	80%	0.00	UNRELIABLE
REPEAT×34REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	3	-0.1%	-2.3%	67%	2.59	UNRELIABLE
SUPPLY-CONTROLSUPPLY-CONTROL — ≥15% supply in one cluster + insider active net: high cluster control = the mechanism behind 7-9-fig runners (measuring WORST (−74 lift) — control alone doesn't predict the pump)ⓘ full method →	253	-0.5%	-2.3%	48%	1.41	CONFIDENT
REPEAT×33REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	8	-0.6%	-11.7%	62%	0.72	UNRELIABLE
REPEAT×61REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	8	-1.0%	-3.3%	88%	2.31	UNRELIABLE
REPEAT×12REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	81	-1.4%	-2.9%	52%	2.58	LOW
REPEAT×7REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	41	-1.5%	-5.9%	27%	2.75	LOW
REPEAT×19REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	16	-2.2%	-6.4%	44%	1.54	UNRELIABLE
REPEAT×62REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	6	-2.2%	-6.8%	33%	7.28	UNRELIABLE
REPEAT×15REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	10	-2.2%	-9.6%	30%	2.56	UNRELIABLE
REPEAT×35REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	51	-2.4%	-3.7%	41%	0.84	LOW
REPEAT×59REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	4	-2.6%	-10.4%	50%	1.29	UNRELIABLE
CLUSTERED-SUPPLYCLUSTERED-SUPPLY — top cluster ≥5% and size ≥5 net: coordinated wallets holding real % (predict-pump lift +9)ⓘ full method →	807	-2.6%	-3.5%	42%	1.08	CONFIDENT
REPEAT×38REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	25	-2.7%	-6.5%	28%	1.96	UNRELIABLE
REPEAT×45REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	45	-2.9%	-4.4%	29%	0.90	LOW
HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →	304	-2.9%	-4.0%	42%	1.11	CONFIDENT
REPEAT×10REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	60	-3.1%	-4.7%	45%	1.12	LOW
REPEAT×54REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	87	-3.3%	-4.7%	45%	0.67	LOW
REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	273	-3.4%	-4.3%	45%	0.53	CONFIDENT
GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	1827	-3.4%	-3.9%	43%	1.02	CONFIDENT
FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →	1827	-3.4%	-3.9%	43%	1.02	CONFIDENT
STEALTH-WAVESTEALTH-WAVE — 60d wave-then-decay historical shape net: 'let it die, then it explodes' (SPACEMOON) (≈ baseline (−0.2))ⓘ full method →	1952	-3.4%	-3.8%	42%	1.03	CONFIDENT
ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	1734	-3.5%	-4.0%	42%	0.99	CONFIDENT
STEALTHSTEALTH — ≥4 fires while MC drifts <1.5× net: buying absorbed with no markup — hidden accumulation, not a run-up that dumps (predict-pump lift +13)ⓘ full method →	1631	-3.5%	-4.1%	42%	1.08	CONFIDENT
REPEAT×36REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	17	-3.6%	-6.2%	18%	1.51	UNRELIABLE
FLOW-DIVFLOW-DIV — OBV/CVD up while 1h price ≤ +5% net: flow in, price flat — no markup yet (measuring −10)ⓘ full method →	437	-3.7%	-4.5%	44%	0.89	CONFIDENT
REPEAT×16REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	21	-3.7%	-8.3%	48%	0.68	UNRELIABLE
REPEAT×6REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	14	-3.7%	-4.9%	14%	1.01	UNRELIABLE
ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →	155	-3.8%	-6.0%	35%	1.16	CONFIDENT
REPEAT×8REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	57	-3.9%	-8.3%	39%	1.31	LOW
REPEAT×107REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	4	-4.0%	-4.7%	0%	0.00	UNRELIABLE
SIDE-CLUSTERSIDE-CLUSTER — whale side-wallets in, primary not yet net: side wallets front-run the main — copy the earliest (—)ⓘ full method →	8	-4.0%	-6.1%	25%	0.41	UNRELIABLE
REPEAT×42REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	10	-4.2%	-13.2%	20%	2.60	UNRELIABLE
SQUEEZESQUEEZE — bb_squeeze true net: volatility contraction before markup (measuring −12)ⓘ full method →	240	-4.2%	-5.1%	46%	0.55	CONFIDENT
REPEAT×43REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	18	-4.3%	-5.9%	28%	0.44	UNRELIABLE
REPEAT×131REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	3	-4.4%	-7.4%	0%	0.00	UNRELIABLE
REPEAT×13REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	34	-4.4%	-6.5%	47%	1.30	LOW
REPEAT×44REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	6	-4.5%	-9.5%	50%	0.51	UNRELIABLE
REPEAT×47REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	7	-4.6%	-7.8%	43%	0.20	UNRELIABLE
REPEAT×21REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	9	-4.6%	-8.6%	44%	0.67	UNRELIABLE
REPEAT×5REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	55	-5.1%	-8.6%	45%	0.73	LOW
REPEAT×46REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	3	-5.1%	-5.3%	0%	0.00	UNRELIABLE
REPEAT×53REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	4	-5.1%	-14.2%	50%	0.39	UNRELIABLE
REPEAT×25REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	189	-5.6%	-6.8%	38%	1.07	CONFIDENT
BUNDLED-RISKBUNDLED-RISK — ≥500 bundlers + ≥3 rat traders net: bundled/risk supply (RISK flag) (predict-pump lift +47 (but it's a risk, not a buy signal))ⓘ full method →	209	-5.6%	-6.4%	36%	0.62	CONFIDENT
REPEAT×11REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	294	-6.0%	-6.7%	40%	0.55	CONFIDENT
REPEAT×27REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	28	-6.5%	-10.5%	32%	0.70	UNRELIABLE
REPEAT×9REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	15	-7.0%	-8.9%	20%	0.63	UNRELIABLE
REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →	96	-7.4%	-11.1%	38%	0.73	LOW
REPEAT×22REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	8	-7.5%	-12.4%	12%	1.66	UNRELIABLE
VOL-ANOM	80	-7.6%	-10.1%	28%	0.64	LOW
REPEAT×18REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	47	-8.1%	-11.0%	32%	0.56	LOW
REPEAT×48REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	7	-8.3%	-19.5%	29%	0.53	UNRELIABLE
REPEAT×14REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	17	-9.3%	-11.8%	12%	0.22	UNRELIABLE
REPEAT×30REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	8	-9.7%	-15.9%	12%	0.04	UNRELIABLE
HOLDER-SURGEHOLDER-SURGE — 24h holder delta over threshold net: new hands taking supply (tail, realized-negative in paper)ⓘ full method →	3	-10.7%	-22.9%	0%	0.00	UNRELIABLE
REPEAT×20REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	8	-10.9%	-12.1%	0%	0.00	UNRELIABLE
WM-BUYWM-BUY — tracked whalemover bought net: a whalemover is in (weight raised to 8)ⓘ full method →	9	-15.9%	-29.7%	11%	0.11	UNRELIABLE
REPEAT×23REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	7	-16.4%	-27.9%	29%	0.10	UNRELIABLE
REPEAT×26REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	3	-18.5%	-35.6%	0%	0.00	UNRELIABLE
RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →	10	-19.0%	-24.8%	0%	0.00	UNRELIABLE
REPEAT×28REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	5	-20.6%	-40.9%	20%	0.70	UNRELIABLE
NET-INNET-IN — net_buy_vol_1h_mcap ≥ 0.05 net: net inflow sized to mcap (thin sample)ⓘ full method →	4	-21.6%	-41.2%	0%	0.00	UNRELIABLE

Emerging (<3 graded fires, too thin to rank yet): REPEAT×134, REPEAT×60, REPEAT×24, REPEAT×133, REPEAT×132, REPEAT×130, REPEAT×51, REPEAT×124, REPEAT×106, FRESH-VELOCITY, REPEAT×52, REPEAT×136, REPEAT×135, REPEAT×57, REPEAT×129, REPEAT×127, REPEAT×126, REPEAT×125, REPEAT×123, REPEAT×108, REPEAT×104, REPEAT×87, REPEAT×86, REPEAT×85, REPEAT×84, REPEAT×77, REPEAT×64

Signature leaderboard — the accumulation indicators

Which accumulation indicators fire most in the last 7d, and the avg engine-score / conviction of the setups they appear in. Higher avg score = a signature that tags the strongest accumulation.

Signature	Fires	Avg score	Avg conviction
REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	30	36	29%
REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →	29	38	30%
GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	27	40	31%
FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →	17	41	33%
RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →	16	41	34%
HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →	15	40	32%
ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	14	41	33%
ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →	6	35	27%
SQUEEZESQUEEZE — bb_squeeze true net: volatility contraction before markup (measuring −12)ⓘ full method →	5	33	26%
REPEAT×4REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →	4	34	28%
STEALTHSTEALTH — ≥4 fires while MC drifts <1.5× net: buying absorbed with no markup — hidden accumulation, not a run-up that dumps (predict-pump lift +13)ⓘ full method →	4	34	28%
LADDERLADDER — buys_ge_1k≥5, median buy <$800 net: size-tiered accumulation (thin sample)ⓘ full method →	4	42	30%
FRESH-VELOCITYFRESH-VELOCITY — fresh-wallet count accelerating between snapshots net: fresh supply velocity up (thin sample)ⓘ full method →	4	40	29%
CLUSTERED-SUPPLYCLUSTERED-SUPPLY — top cluster ≥5% and size ≥5 net: coordinated wallets holding real % (predict-pump lift +9)ⓘ full method →	3	21	20%
FLOW-DIVFLOW-DIV — OBV/CVD up while 1h price ≤ +5% net: flow in, price flat — no markup yet (measuring −10)ⓘ full method →	2	44	32%
VOL-ANOM	1	50	36%

Recent accumulation signals

Token	Chain	Score	Conv	Cov	MC	Signatures	Rec TP/SL	Age
DOPE	SOL	39	34%	72%	$1.26M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →	—	1h
DOPE	SOL	39	34%	72%	$1.26M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →	+20%/-15%	2h
DOPE	SOL	39	34%	72%	$1.26M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →	+20%/-15%	4h
DOPE	SOL	23	21%	86%	$1.26M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →	+20%/-15%	5h
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.4d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.5d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.6d
TikTok	SOL	50	36%	100%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →VOL-ANOMASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.6d
TikTok	SOL	47	33%	100%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →LADDERLADDER — buys_ge_1k≥5, median buy <$800 net: size-tiered accumulation (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.7d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.7d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.8d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.8d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.9d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	2.9d
TikTok	SOL	44	35%	83%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	3.0d
TikTok	SOL	28	30%	69%	$1.24M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →RANGERANGE — fires across ≥3 distinct MC bands net: re-buying the range, not one spike (thin sample)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	3.1d
1F916	BASE	34	27%	83%	$1.78M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →SQUEEZESQUEEZE — bb_squeeze true net: volatility contraction before markup (measuring −12)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	5.6d
1F916	BASE	18	19%	69%	$1.78M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →SQUEEZESQUEEZE — bb_squeeze true net: volatility contraction before markup (measuring −12)ⓘ full method →ASSOC-SUPPLYASSOC-SUPPLY — weighted % buy-$ from graded/serial-funder wallets net: connected supply via graded, funder, or serial-funder ties (predict-pump lift +14)ⓘ full method →	+20%/-15%	5.7d
CLAUS	ETH	46	33%	100%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →LADDERLADDER — buys_ge_1k≥5, median buy <$800 net: size-tiered accumulation (thin sample)ⓘ full method →SQUEEZESQUEEZE — bb_squeeze true net: volatility contraction before markup (measuring −12)ⓘ full method →FLOW-DIVFLOW-DIV — OBV/CVD up while 1h price ≤ +5% net: flow in, price flat — no markup yet (measuring −10)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-VELOCITYFRESH-VELOCITY — fresh-wallet count accelerating between snapshots net: fresh supply velocity up (thin sample)ⓘ full method →	+20%/-15%	5.9d
CLAUS	ETH	43	30%	100%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →SQUEEZESQUEEZE — bb_squeeze true net: volatility contraction before markup (measuring −12)ⓘ full method →FLOW-DIVFLOW-DIV — OBV/CVD up while 1h price ≤ +5% net: flow in, price flat — no markup yet (measuring −10)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-VELOCITYFRESH-VELOCITY — fresh-wallet count accelerating between snapshots net: fresh supply velocity up (thin sample)ⓘ full method →	+20%/-15%	5.9d
IMD	ETH	27	22%	83%	$16.90M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →	+20%/-15%	5.9d
ZC	ETH	21	19%	86%	$10.88M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →CLUSTERED-SUPPLYCLUSTERED-SUPPLY — top cluster ≥5% and size ≥5 net: coordinated wallets holding real % (predict-pump lift +9)ⓘ full method →	+20%/-15%	5.9d
CLAUS	ETH	33	26%	83%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-VELOCITYFRESH-VELOCITY — fresh-wallet count accelerating between snapshots net: fresh supply velocity up (thin sample)ⓘ full method →	+20%/-15%	6.2d
IMD	ETH	27	22%	83%	$16.90M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-SWARMFRESH-SWARM — ≥300 fresh_wallets typed net: CZ pattern — 100s of small/fresh wallets (predict-pump lift +8)ⓘ full method →	+20%/-15%	6.3d
ZC	ETH	17	18%	69%	$10.88M	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →CLUSTERED-SUPPLYCLUSTERED-SUPPLY — top cluster ≥5% and size ≥5 net: coordinated wallets holding real % (predict-pump lift +9)ⓘ full method →	+20%/-15%	6.3d
CLAUS	ETH	40	28%	100%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →LADDERLADDER — buys_ge_1k≥5, median buy <$800 net: size-tiered accumulation (thin sample)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →FRESH-VELOCITYFRESH-VELOCITY — fresh-wallet count accelerating between snapshots net: fresh supply velocity up (thin sample)ⓘ full method →	+20%/-15%	6.3d
CLAUS	ETH	37	26%	100%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →LADDERLADDER — buys_ge_1k≥5, median buy <$800 net: size-tiered accumulation (thin sample)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	+20%/-15%	6.6d
CLAUS	ETH	30	24%	83%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	+20%/-15%	6.7d
SKY	ETH	25	23%	86%	$9961.59M	REPEAT×4REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →STEALTHSTEALTH — ≥4 fires while MC drifts <1.5× net: buying absorbed with no markup — hidden accumulation, not a run-up that dumps (predict-pump lift +13)ⓘ full method →ABSORBABSORB — buy_sell_ratio≥1.5 and flat MC net: DIRECT buys-minus-sells when flow data exists (measuring WORST (−60 lift) — the purest net signal is empirically weak at predicting pumps)ⓘ full method →HLHL — higher_low true net: Wyckoff staircase (predict-pump lift +9)ⓘ full method →SQUEEZESQUEEZE — bb_squeeze true net: volatility contraction before markup (measuring −12)ⓘ full method →CLUSTERED-SUPPLYCLUSTERED-SUPPLY — top cluster ≥5% and size ≥5 net: coordinated wallets holding real % (predict-pump lift +9)ⓘ full method →	+20%/-15%	6.7d
CLAUS	ETH	37	30%	83%	$317K	REPEAT×4REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →STEALTHSTEALTH — ≥4 fires while MC drifts <1.5× net: buying absorbed with no markup — hidden accumulation, not a run-up that dumps (predict-pump lift +13)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	+20%/-15%	6.7d
CLAUS	ETH	37	30%	83%	$317K	REPEAT×4REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →STEALTHSTEALTH — ≥4 fires while MC drifts <1.5× net: buying absorbed with no markup — hidden accumulation, not a run-up that dumps (predict-pump lift +13)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	+20%/-15%	6.8d
CLAUS	ETH	37	30%	83%	$317K	REPEAT×4REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →STEALTHSTEALTH — ≥4 fires while MC drifts <1.5× net: buying absorbed with no markup — hidden accumulation, not a run-up that dumps (predict-pump lift +13)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	+20%/-15%	6.8d
CLAUS	ETH	30	24%	83%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →GMGN-CLUSTERGMGN-CLUSTER — smart_wallets≥15 or renowned≥20 net: GMGN typed smart-money cluster present (predict-pump lift +8)ⓘ full method →	+20%/-15%	6.8d
CLAUS	ETH	22	23%	69%	$317K	REPEAT×3REPEAT — ≥4 fires, same CA / 24h across sources net: independent systems agreeing = pattern, not one trader (baseline)ⓘ full method →REACCUMREACCUM — prior RUNNER_3X+ firing again net: stage-2 re-accumulation (weight raised to 8)ⓘ full method →	+20%/-15%	6.8d

↳ accumulation_dashboard.py — signals from accum_engine (signatures = the indicators), performance from paper_scorecard (ACCUMULATION family). Backtest-gate any threshold change. Exits use the live SWING_ARM21 champion, promoted 260925.