HISTORICAL ANOMALYT+34M
H·6B41
PUMP.FUNSIMILARITY94.8%
COMPARED WITH127 states
REACHED >$1M31%
HISTORICAL BASELINE4.2%
7.4×ABOVE BASELINE
The matched cohort reached $1M at 7.4× the rate of the unconditional Pump.fun population.
HISTORICAL MARKET INTELLIGENCE/ PUMP.FUN
Every pumphas happenedbefore.
REPLAY compares live Pump.fun tokens against historical market states to reveal the launches that looked most similar — and what happened next.
ANALYZE A PUMP.FUN TOKEN
LIVE INDEXING · PUMP.FUN · SOLANA
OR OPEN A SAMPLE STATE
CURRENT STATEPUMP.FUN · SOLANA
AGE42m
MARKET CAP$418.2K
LIQUIDITY$71.4K
24M VOLUME$386.7K
HOLDER VELOCITY+18.4%
BUY PRESSURE64.8%
MARKET DNA MATCH
94.7%
MATCHED AGAINST187HISTORICAL MARKET STATES
CURRENTMEDIAN OUTCOMEINTERQUARTILEUPPER TWINLOWER TWINLOG SCALE · NORMALISED AT ENTRY
AGGREGATE HISTORICAL OUTCOMESN = 187
REACHED $1M
31.4%
REACHED 2×
44.8%
FELL >50%
27.2%
MEDIAN PEAK
+182%
DESCRIPTIVE HISTORICAL OUTCOMES · NOT A FORECAST
PUMP.FUN STATES INDEXED12,481,930
TOKENS TRACKED1,842,103
LIVE SCANACTIVE
LATEST SLOT312,884,190
INDEX STATUSONLINE
INDEX TELEMETRY IS A DEVELOPMENT PLACEHOLDER — NOT YET SYNCED TO THE PRODUCTION INDEX
01MARKET MEMORY
Traders see the present.
REPLAY remembers the past.
Traditional scanners tell you what is happening. Price. Volume. Liquidity. Holders. Those numbers are accurate and they are context-free — they describe a state without telling you what that state has historically preceded.
01
PRICE
Where it is — not where it has gone from here before.
02
VOLUME
How much traded — not what that turnover has preceded.
03
LIQUIDITY
Depth right now — not depth when everyone exits.
04
HOLDERS
How many wallets — not which wallets, or for how long.
REPLAY asks a different question: when tokens looked like this before, what happened next?
TRADITIONAL SCANNER
“What is happening?”
- OUT Current price and market cap
- OUT Volume over the last interval
- OUT Holder count and top wallets
- OUT Liquidity depth at this moment
RESULT — A DESCRIPTION OF THE PRESENT
REPLAY
“When has this happened before?”
- IN The full state vector, not a single metric
- IN Ranked historical states with the same fingerprint
- IN The outcomes that actually followed them
- IN Distribution of results — upside and failure alike
RESULT — PRECEDENT FOR THE STATE YOU ARE TRADING
02MARKET DNA
A market state is
more than a chart.
REPLAY does not compare candlestick shapes. A chart is one projection of a market; the state is the whole object — capital, distribution, structure and time measured together at a specific point in a token’s life.
LIVE STATECOHORT MEANDIMENSION DIVERGENCE
MARKET
Scale, turnover and the tempo of execution.
Market cap$418.2K
Liquidity17.1%
Volume$386.7K
Volatility61 σ
Transaction velocity94 / m
HOLDERS
Who holds the book, and how they got there.
Holder growth+18.4%
Concentration21.8%
Distribution1,204
New-wallet ratio44%
FLOW
Direction, size and quality of capital.
Buy pressure64.8%
Capital inflow$74.1K
Smart wallets14
Accumulation38%
STRUCTURE
How the launch itself was assembled.
Deployer priorREPEAT
Wallet clustering6.2%
LP structureBONDING
Launch progression74%
TIME
Position within the lifecycle, not the clock.
Token age42m
Growth velocity+112%
Lifecycle stageDISCOVERY
Time in stage18m
RETRIEVAL PATH
IN
LIVE TOKEN
A contract address resolves to its current on-chain state and full transaction history.
01
MARKET STATE VECTOR
Twenty-four attributes are normalised against lifecycle stage, so a 40-minute token is compared like for like.
02
HISTORICAL INDEX
Every indexed launch is stored as a sequence of states — not a single snapshot per token.
03
NEAREST MARKET STATES
Historical states are ranked by distance across the full vector and filtered for structural validity.
OUT
FORWARD OUTCOMES
What actually happened after each matched state is measured and aggregated into a distribution.
03THE REPLAY
94.7%
HISTORICAL SIMILARITY
187 comparable Pump.fun market states identified across the index.
PRIMARY MATCH DRIVERS
WEIGHTHISTORICAL COHORT
N = 187MEDIAN MAXIMUM EXPANSION+182%
MEDIAN TIME TO PEAK6h 42m
HISTORICAL OUTCOMES ARE DESCRIPTIVE, NOT PREDICTIVE — THEY DESCRIBE WHAT FOLLOWED SIMILAR STATES, NOT WHAT WILL FOLLOW THIS ONE.
CLOSEST REPLAYS
RANKED BY VECTOR DISTANCE| # | Historical state | State at match | Peak reached | Expansion | Similarity |
|---|---|---|---|---|---|
| 01 | H·4A2F | $392K | $2.7M | +588% | 96.1% |
| 02 | H·9C17 | $441K | $910K | +106% | 94.8% |
| 03 | H·2B8E | $387K | $138K | −64% | 93.7% |
| 04 | H·7D30 | $428K | $1.4M | +227% | 92.9% |
01H·4A2F+588%$392K → $2.7M96.1%
02H·9C17+106%$441K → $910K94.8%
03H·2B8E−64%$387K → $138K93.7%
04H·7D30+227%$428K → $1.4M92.9%
Historical tokens are anonymised in the public demo. Each handle is one market state, not one token.
04REPLAY FEED
The market,
through history.
REPLAY continuously scans active Pump.fun tokens and surfaces states whose historical cohorts diverge sharply from the population baseline — in either direction.
SCANNING· LAST SURFACED 0s AGO
DISTRIBUTION PATTERNT+1H 12M
H·D28A
PUMP.FUNSIMILARITY91.7%
COMPARED WITH89 states
FELL >50% / 12H67%
HISTORICAL BASELINE41%
67%FELL >50% WITHIN 12H
Concentration and new-wallet skew place this state inside a historically distributive cohort.
EARLY MOMENTUMT+19M
H·3F07
PUMP.FUNTOKEN AGE19m
MATCHED COHORT42 launches
MEDIAN EXPANSION+271%
MEDIAN TIME TO PEAK3h 08m
+271%MEDIAN SUBSEQUENT EXPANSION
Small cohort — 42 states. Treat the dispersion as wide and the median as weakly supported.
FEED OBJECTS SHOWN ARE ILLUSTRATIVE AND THE TOKENS ANONYMISED — THE LIVE SCANNER ACTIVATES WITH THE PRODUCTION INDEX.
05HISTORICAL TRAJECTORIES
One market state.
Hundreds of possible histories.
Each faint line is a real historical launch that occupied a state close to this one. The fan is the point: the same fingerprint has produced very different outcomes, and the shape of that disagreement is itself information.
DRAG ACROSS THE CHART TO READ THE DISTRIBUTION AT ANY POINT
CURRENTMEDIAN OUTCOMEINTERQUARTILEUPPER TWINLOWER TWINLOG SCALE · NORMALISED AT ENTRY
COHORT SIZE84states plotted
MEDIAN OUTCOME+182%maximum expansion
UPPER QUARTILE+486%top 25% of cohort
LOWER QUARTILE−38%bottom 25% of cohort
DISPERSIONWIDEoutcomes are not concentrated
06EXIT INTELLIGENCE
A position is only worth what the market can absorb.
Headline market cap is a multiplication, not a bid. What matters is realisable value: the size the book can take before the price you are exiting into moves against you.
STATUSIN DEVELOPMENT · SECONDARY CAPABILITY
POSITION$10,000
ESTIMATED EXIT
$9,230
EST. IMPACT
−7.7%
EXIT SCORE
71 / 100
Whale concentrationMODERATE
Liquidity stressLOW
Holder concentrationELEVATED
STRESS SCENARIOSESTIMATED MARKET IMPACT
TOP HOLDER SELLS 25%
Largest wallet holds 4.6% of supply.
TOP 3 HOLDERS REDUCE
Combined 11.2% of supply exiting into current depth.
COHORT-WIDE 10% ROTATION
Modelled on distribution behaviour in the matched cohort.
Impact estimates are modelled from observable liquidity and holder structure. They are approximations, not execution guarantees — real fills depend on routing, contention and conditions at the moment of the trade. Treat these figures as a comparative measure of exit difficulty between tokens, not as a quoted price.
07WHY REPLAY EXISTS
Markets rhyme.
Memory is the edge.
Every Pump.fun launch creates data — order flow, wallet formation, liquidity structure, the exact shape of a distribution. Most of it disappears into charts nobody revisits.
REPLAY turns that history into searchable market memory. Every launch. Every stage. Every outcome. Indexed so that today’s market can be compared against yesterday’s, at the level of structure rather than shape.
Markets don’t repeat. They leave fingerprints.
EVERY LAUNCH
1,842,103
tokens observed from first block
EVERY STAGE
12,481,930
discrete states captured across lifecycles
EVERY OUTCOME
1,842,103
forward results measured and stored
EVERY COMPARISON
REAL-TIME
queried in milliseconds against the index
The more the market trades,
the more REPLAY remembers.
08METHODOLOGY
Similarity,
not prophecy.
REPLAY converts each market into a multidimensional state at successive points in its lifecycle. Those states are compared against historical observations on normalised attributes, ranked by similarity, and the outcomes that actually followed are measured.
MATCHED COHORT · 31PARTIAL MATCHINDEXED · OUT OF COHORT
REPLAY does not tell you what will happen. It shows you what happened when markets looked similar before.
A chart shows where price went. REPLAY shows when the market has been here before — and how often that ended well.
STATED LIMITATIONS
- Similarity is not causation. A close match describes structural resemblance, not a mechanism.
- Sample size governs confidence. Small cohorts are reported as small cohorts.
- Historical regimes change. Outcomes from a different market environment carry less weight.
- Outcomes are distributions, never single numbers. Dispersion is reported alongside the median.
I
DATA COLLECTION
Launches are observed from the first block. Trades, transfers, wallet formation and liquidity events are captured continuously rather than sampled.
II
STATE NORMALIZATION
Raw attributes are normalised against lifecycle position, so a 20-minute token is never compared against a 20-hour one on absolute values.
III
SIMILARITY ENGINE
Each state becomes a vector. Distance is computed across the full set of dimensions, with attribute weighting derived from historical discriminative power.
IV
HISTORICAL RETRIEVAL
Nearest states are retrieved and filtered — degenerate matches, insufficient forward history and structurally invalid launches are excluded.
V
OUTCOME ANALYSIS
What actually followed each matched state is measured over fixed horizons and aggregated into a distribution, reported with its sample size.
09CALIBRATION
When REPLAY says 30%,
how often is it 30%?
A similarity engine that is never checked against its own record is an opinion. Every probability REPLAY reports is scored against what actually happened, and the result is published here — including where it is wrong.
WITHIN 4 POINTSOUTSIDE 4 POINTSDOT AREA ∝ RESOLVED STATES IN BUCKET
A point on the diagonal means the number meant what it said. A point above it means REPLAY understated the outcome; below, it overstated.
The curve bends toward the middle at both ends — the engine is measurably over-confident at the extremes. That is worth knowing before you trade on a 90%.
RESOLVED STATES18,240Historical states with a full forward window
MEAN DEVIATION2.9 ptsPredicted vs observed, weighted by bucket size
WORST BUCKET6.1 ptsOver-confident at the extremes, as expected
REVIEW WINDOWROLLING 90DRecomputed as the index grows
Calibration figures shown are illustrative until the production index has resolved a full review window. The method is fixed: every reported probability is bucketed, matched against the realised outcome, and published whether or not it flatters the engine.
10REPLAY LIVE
REPLAY analyzes itself.
The first public market REPLAY analyzes is its own. The module below is wired to the same engine every other token passes through, and populates automatically once $REPLAY has an on-chain state to read.
AGE
—
Time since first block
CURRENT MARKET STATE
—
Live vector snapshot
HISTORICAL MATCH
—
Similarity against the index
MATCHED STATES
—
Cohort size at this stage
MODULE POPULATES ON LAUNCH · NO FIGURES ARE IMPLIED OR PROJECTED
11TOKEN
Built on the market it studies.
$REPLAY launches through Pump.fun — the same venue the engine indexes.
AVAILABLE NOW
- Contract address analysis
- Market DNA matching
- Historical cohort retrieval
- Replay Feed (read access)
COMING NEXT
- Deeper historical cohorts
- Advanced filtering
- Live market alerts
- Custom watchlists
RESEARCH
- Exit Intelligence
- Deployer intelligence
- Scanner functionality
- External API access
CONTRACTPLACEHOLDER · SET AT LAUNCH
REPLAYzn4cVXt9kqAy7mHsQe3TgWuPdBfLj2Npump
Ecosystem utility is described as intent, not entitlement. Items outside AVAILABLE NOW are unbuilt and may change or be dropped. Holding $REPLAY does not confer equity, revenue rights, guaranteed or perpetual access to any feature, or any expectation of return.
12SYSTEM EXPANSION
System expansion.
Coverage widens as the index deepens. Each phase is gated on the one before it — nothing downstream is meaningful without the memory layer underneath.
PHASE 01IN PROGRESS
PUMP.FUN MEMORY
- Historical indexing
- Contract address analysis
- Market DNA
- Historical cohorts
- Replay Feed
PHASE 02NEXT
DEEP INTELLIGENCE
- Wallet structure
- Exit Intelligence
- Advanced alerts
- Deployer intelligence
PHASE 03PLANNED
SOCIAL CONTEXT
- X velocity
- KOL propagation
- Narrative origin
- Attention vs capital
PHASE 04PLANNED
ON-CHAIN MEMORY
- Solana-wide coverage
- Additional launch venues
- Public API
- External integrations
Find its precedent.
Paste a Pump.fun contract address and see what history remembers.
ANALYZE A PUMP.FUN TOKEN
LIVE INDEXING · PUMP.FUN · SOLANA
OR
REQUEST INDEX ACCESS
ACCESS OPENS WITH THE PRODUCTION INDEX · NO SPAM, ONE MESSAGE
HISTORICAL INTELLIGENCE. NOT FINANCIAL ADVICE.