Methodology
How the numbers on PiChan's public pages are made, and what they leave out. PiChan reports what it saw. It does not grade launchpads or tokens, and nothing here is a recommendation.
Launchpad Scorecard
The Scorecard counts what happened to the tokens launched through each launchpad on Arc: how many launched, how many creators launched them, whether the tokens reached a pool, and whether they traded. It is computed once a night from PiChan's own collected chain data, with no third-party scores, and written as a static file. The page and the weekly export read that file and nothing else.
Cohorts, horizons and windows
- Launch-day cohort: the tokens whose launch block falls on one UTC day.
- Horizon: how long after each token's own launch a metric looks: 24 hours, 7 days or 30 days. A cohort is only reported for a horizon once every token in it has reached that horizon and every data stream has caught up. There are no provisional values.
- Window: how many cohorts a published row pools: 7, 30 or 90 days, ending at the newest cohort that has matured for that horizon. Rows for different horizons can end on different days; every row prints its own cohort range. One-day rows are building blocks: they are stored but not published.
- One token, one launch. Duplicates from re-reads and backfills are removed before counting.
Rule v3.1 parameters
| Rule | Value |
|---|---|
| Minimum sample | A window needs at least 20 tokens in its pooled sample (the whole window, not each day). Below that the row is Unknown (insufficient n), and the counts seen are still printed. |
| Still trading at a horizon | At least 3 swaps and at least 10 USDC of swap volume in the 24 hours ending at the horizon. Only tokens with a USDC-quoted pool are counted, in the sample and in n. This plain count includes a wallet trading with itself; the farm-adjusted count below does not. |
| Still trading, farm-adjusted | Still trading as above, and at least 3 distinct traders in the same 24 hours, and the largest trader below 80% of the USDC volume of the attributed trades. The creator and wallets PiChan labels (system, contracts) are left out of both counts. |
| Trades and their coverage | A trade's wallet is read from the token transfers of its transaction (join version 1): the wallet holding at least 85% of the trade's side is the trader. A transaction whose wallet cannot be told (no wallet reaches that share, or a buy in the launch transaction) is not attributed. Every value that reads trades prints the share of swap transactions attributed beside its n. |
| Creator sold within 24 hours | The creator's attributed sells within 24 hours of launch reach at least 10% of everything they received in that time, the launch mint included, so the share never exceeds 100% (base received_24h). The creator is taken from PiChan's deployment records and counted only when it comes from the launch event. n is the launches whose creator held the token. A launch whose creator never held it, and whose launch transaction left at least 1% of the minted supply with another address (a launch vault, not the pools), is counted apart and left out of n; what the vault later sends the creator counts like any transfer. |
| Positions, hold time and retention | A position opens at a wallet's first attributed buy in the first 24 hours and follows every transfer of the token from then on. It closes when what is left falls to 0.001 of its peak or less. Hold time is the median time to close of positions closed within 7 days of launch; the share still open at 7 days is printed beside it. Retention: of positions opened in the first hour, the share still open at 24 hours. |
| Fall from the first-hour average | Per token, on its USDC pools: the USDC-weighted mean of the pool price after each swap in the first hour, and the lowest pool price in hours 1 to 24 (after the first hour). The value is how far the low is below that average, in percent (0 when it is not below). The median is over tokens. |
| USDC depth at 24 hours | Per token, the depth within ±2% of the price at 24 hours: the USDC its pools hold for prices down to 2% below, plus the tokens they hold for prices up to 2% above, valued at the price. Every liquidity position counts for the part of its range inside the band. Summed over at most 20 USDC pools per token, those with the most swaps in the first 24 hours. A pool with no swap by then (its price is not known) or with a hook that rewrites the pool's own accounting is left out and counted; a token is Unknown only when no USDC pool can be measured. |
| Graduated (pool after launch) | Not applicable to a window when at least 95% of its launches already had their pool at launch: there is nothing to graduate to. |
| From repeat launchers | The creator made at least 2 other launches on Arc, through any launchpad, in the 7 days before this launch. Only creators decoded from the launch event count. |
| Data gap: lower confidence | A gap of more than 300 blocks (≈ 150 s) in a required stream lowers the row's confidence to medium; the gaps are listed in the row's inputs. |
| Data gap: Unknown | A gap of more than 7,200 blocks (≈ 3,600 s) makes the row Unknown (input gap). |
| Missing chain times | When the share of a row's records with no chain time is above 0.001, its confidence is medium. |
| Private mainnet | Launches and pools before 16 Sept 2026, 00:00 UTC (block 21,068,653) are excluded from every cohort. They are counted in each run's inputs, and the archive keeps them. |
Every parameter in this snapshot
| Parameter | Value |
|---|---|
| STATS_EXCLUDE_PRIVATE_MAINNET | true |
| STATS_MIN_N | 20 |
| STATS_TRADING_MIN_SWAPS | 3 |
| STATS_TRADING_MIN_USDC | 10 |
| STATS_SERIAL_LOOKBACK_DAYS | 7 |
| STATS_SERIAL_MIN_OTHER | 2 |
| STATS_GAP_BLOCKS | 300 |
| STATS_GAP_UNKNOWN_BLOCKS | 7200 |
| STATS_POOL_AT_LAUNCH_NA | 0.95 |
| STATS_MISSING_TIME_MEDIUM | 0.001 |
| STATS_FARM_MIN_WALLETS | 3 |
| STATS_FARM_MAX_TOP_SHARE | 0.80 |
| STATS_CREATOR_SELL_PCT | 10 |
| STATS_BUYER_WINDOW | 24h |
| STATS_FIRST_HOUR | 1h |
| STATS_HOLD_HORIZON | 7d |
| STATS_POSITION_DUST | 0.001 |
| TRADE_JOIN_VERSION | 1 |
| TRADER_DOMINANCE | 0.85 |
| STATS_DEPTH_BAND_PCT | 2 |
| POOL_CAP | 20 |
| STATS_VAULT_MIN_SHARE_PCT | 1 |
| STATS_CREATOR_SELL_BASE | received_24h |
| PUBLIC_MAINNET_FROM_BLOCK | 21068653 |
Seconds are at Arc's nominal block time of 500 ms. Changing any parameter bumps the rule version, and the change is noted here.
Status, confidence and Unknown
- Status is final or Unknown. Unknown always carries a reason. Only an insufficient sample keeps its numbers on the row; any other Unknown means the sample is partial, so no number is shown as the metric.
- Confidence is high when every required stream is complete for the window, medium with a gap between 300 and 7,200 blocks, and low on input gap or input lag rows.
| Reason code | Meaning |
|---|---|
| input_gap | input gap: the window reaches before the data streams start, or a stream has a gap of more than 7,200 blocks |
| input_lag | input lag: a data stream had not reached the end of the window when the job ran |
| lookback_incomplete | the 7-day creator lookback reaches before the earliest collected launch |
| insufficient_n | fewer than 20 tokens in the window |
| not_applicable:pool_at_launch | not applicable: at least 95% of these launches create their pool in the launch transaction |
| not_collected:trader_identity | not collected under rules v1–v2: needed trader identity (computed from rule v3) |
| not_collected:holder_ledger | not collected under rules v1–v2: needed the holder ledger (computed from rule v3) |
| not_collected:liquidity_depth | not collected under rules v1–v2: needed liquidity depth (computed from rule v3) |
| not_collected:creator_in_event | not collected: this launchpad's launch event carries no creator |
| quote_token_unknown | this token's bonding curve is priced in a token PiChan can't confirm as USDC, so its curve trades can't be counted in USDC |
| not_collected:curve_trades | not collected: this launchpad's trades happen on its own bonding-curve contract, which PiChan does not decode yet |
Launchpads, kinds and groups
- Launchpad contracts and versions of one product are reported together. Each launchpad in PiChan's registry has a kind, with dated evidence: public (many independent creators), template (one launcher code deployed as many instances), bot (one operator launching automatically) or candidate (not yet classified).
- Public and template launchpads get their own row. Bot launchpads are pooled in one labelled row, "Automated launchers (single operator)", and candidates in "Unidentified launchers". A group row is computed from all its members' tokens, and it lists every member with its launch count. No launchpad is hidden.
- A launchpad whose identity is not confirmed (no active registry entry) is shown as "launcher" and its contract address, with the registry's evidence on the next line, never by a guessed name.
- The total for the chain includes every launch, grouped or not.
Limitations
- Cohorts before Arc public mainnet (16 Sept 2026 00:00 UTC, block 21,068,653) are excluded from the Scorecard (rule v2); the archive keeps them.
- Swaps on pools not in the pool index (v2 pairs, v3 forks) are not attributed to a token.
- Only USDC-quoted pools count toward volume and 'still trading' (native USDC in v4 and the ERC-20 interface 0x3600…0000).
- 'Still trading' is shown plain and farm-adjusted (rule v3: at least 3 distinct traders and no trader above 80 % of attributed volume, creator and labelled wallets excluded).
- Creators come from PiChan's deployment records (D1); creator metrics count only a creator from the launch event; pads whose launch event carries no creator are Unknown.
- Trader-based metrics read the trade join; each prints the share of swap transactions with an attributed trader. A buy in the launch transaction is not attributed.
- Hold time and retention follow the tokens of each buyer's position from the first buy, by transfers; balances are not read on chain, so reflection tokens drift.
- Liquidity at 24 h is the USDC depth within ±2 % of the price, over at most 20 USDC pools per token; pools without a price by 24 h or with a hook that rewrites pool accounting are left out and counted.
- Creator sold leaves out launches whose creator never held the token and launches minted into a launch vault, each counted.
- Cohorts before 25 Sept 2026 have no trade-join or liquidity values: the swaps and liquidity streams start on 24 Sept.
- 30- and 90-day windows read Unknown (input gap) until collected data reaches back far enough to fill them.
- The snapshot time is when the nightly run started (00:30 UTC); it finishes a few minutes later.
- Gap thresholds are converted to seconds at the nominal block time; the measured block time can differ slightly.
- Launchpad kinds and names come from PiChan's registry. A launchpad not yet identified is shown by its launcher contract.
Weekly export and corrections
- Every Monday the nightly run publishes the week that ended on Sunday (ISO week) as JSON and CSV, linked from the Scorecard page. A week is published only when it is complete.
- A published file is never overwritten. A correction is a new file, <week>-rev<n>, whose header names the file it supersedes; the correction is noted here.
- Every correction to a published number is listed on the Corrections page, with the original.
- Columns, one row per line: chain, week, subject_kind, subject_id, subject_name, pad_kind, metric, horizon, window_days, cohort_from, cohort_to, num, den, value, unit, status, unknown_reason, confidence, rule_version, registry_version, join_attributed, join_trades. Ratios are stored as num and den; den is the n printed beside every percentage. The CSV repeats the JSON header as # comment lines.
Rule version history
- v1, 24 Sept 2026: first published rule set.
- v2, 25 Sept 2026: launches and pools from Arc's private mainnet (before block 21,068,653, 16 Sept 2026 00:00 UTC) are excluded from every cohort.
- v3, 12 Oct 2026: creators from PiChan's deployment records; still trading farm-adjusted, creator sold, hold time and USDC depth within ±2% of the price at 24 hours computed; positions still open at 7 days, first-hour buyer retention and the fall from the first-hour average to the low of hours 1 to 24 added. Plain and farm-adjusted still trading are shown side by side.
- v3.1, 15 Oct 2026: creator sold divides the creator's sells by everything they received within 24 hours, not by their peak balance, so it never exceeds 100%.
Not affiliated with Circle or any launchpad named here. Not financial advice.