track record · robinhood chain

Every verdict is timestamped before the outcome.

This page is the product. Every table is recomputed every night, from every passed token in our own ledger and from every token, every day, since each token’s first pool. Nothing here is a dated study, and the tables stay up whether they flatter us or not.

7 sections, listed on the left. Every term is in the glossary. Gray rows have fewer than 30 rows behind them: a hint, not a fact.

October 18, 2026 is a gap in this record. Observations written from October 17 07:42 to October 18 02:00 UTC were lost and won’t be backfilled; here’s what that means.

loading the latest numbers…

What happened to the tokens that passed our checks.

We check the contract and the pool against a short list of facts; a token that holds on every one passes, and we write down its price. 24 hours, 7 days and 30 days later we read it again. Every figure is the share of passed tokens worth more at the later read, with n behind it. A pass means the structure looked clean at that moment; it is not a buy signal, and the figures say so.

If this card is all you see, the live store didn't answer. Try again in a minute.

The same passes, split by what we knew at the moment of the pass.

Each table takes every pass with a 7-day read and groups it by one fact recorded at the pass: where the price sat, how spread out the holders were, the market cap, the pool, the token's age, a model's guess at the odds it halves this week, the launcher, what X was saying, the kind of pass. A group tells you something only when it differs from the all-passes row.

Every token, every day, from its first pool.

The pass ledger only knows tokens that passed. The every-token record has no such filter: one row per token per UTC day from the hourly candles, pass or fail, with the facts a price series carries. Every rule line that reads a candle fact is tested here on every token, every day.

Does any fact at the pass predict a better week?

For every fact we store at the moment a token passes, we split the passes into 4 equal groups from lowest to highest and show the share of each group worth more 7 days on. The groups are set on the first half of the passes and read again on the second half, so a pattern has to show up twice to count. Nothing here changes a rule; it sits beside the rules so you can see which ones the data supports.

What's worth another look.

Research triggers are being tested privately. They’ll appear here with their n and dates once one holds up.

Why is there no list for October 18, 2026?

On October 18, 2026, production was restored in place from the October 17 07:44 UTC physical backup, returning its data to approximately October 17 07:42 UTC. Observations written from October 17 07:42 to October 18 02:00 UTC were lost, including the inputs and original 01:35 UTC snapshot for the October 18 p2026-10-12 publication. In the lost-write interval, 18 hourly runs had succeeded, but their observation rows are unavailable. The October 18 publication is permanently unavailable and will not be backfilled from another source. Real observation collection resumes from October 18 02:33 UTC. Missing observations are unknown, not affirmative absence, zero returns or evidence of no activity.

Nightly listWhat’s there
2026-10-18unavailable: lost in the restore and not backfilled

Noted 2026-10-18. Every earlier list keeps its own date, n and 36-hour shelf life; none is relabeled as October 18. The tables above are recomputed from what survived, each with its own date and n.

Why is there no list for October 16, 17, 19, 20 and 21, 2026?

For October 16, 17, 19, 20 and 21, 2026, the retained all-chain reads on 2026-10-23 found zero publication rows and zero jobs for these days.

Every line we use, and where it came from.

None of the 26 lines below carries a frozen measurement from our own record yet; every line was set by hand, is dated, and says why. Every line is tested on the live views of this page, every day, on every pass and on every token, every day, since its first pool. Nothing here rests on a dated study. Current as of 2026-10-10.

What it decidesValueUnitBasisSet onWhere it came from
Conviction tier: the 10 largest wallets hold this much or less30% of supplyset by hand2026-07-23A Midsummer Meme's Dream: Investigating Market Manipulations in the Meme Coin Ecosystem, arXiv 2507.01963 (posted 16 Apr 2025, revised 2 Jan 2026), section 5.1.3: the 10 largest holders owning over 30% of the supply was the most common anomaly, present in 87.14% of the anomalies detected among high-return tokens. Read July 2026. https://arxiv.org/abs/2507.01963 This is a published finding, not our own measurement: n and date are the paper's, and the line is tested daily on the record page.
Above this the token is watch-only and never sized45% of supplyset by hand2026-07-23Set by hand as a buffer above the 30% line
Conviction needs a pool at least this deep; thinner pools are scout size50,000USD in the poolset by hand2026-08-01Set by hand, lowered from 100,000
Conviction size: this share of the pool, in dollars1% of the poolset by hand2026-10-06Set by hand at 1% on 2026-07-23. Kept on 2026-10-06 as the whole of the size rule for conviction: the $1,000 and $100 dollar caps that sat beside it (hand-set 2026-07-23, no source) came out, so size is a share of the pool by tier and nothing else. The record tests the share.
A deployer still holding more than this fails the token5% of supplyset by hand2026-07-23Policy, set by hand on 2026-07-23 at 5% of supply: a deployer still holding more than this fails the token. Rationale: published research on token risk read in July 2026; no citation on file, so this is a policy line, not a measured one.
Off: no holder-count floor on the sizing check (0 = the check does not run)0walletsset by hand2026-10-06Set by hand at 100 on 2026-07-28, the one holder figure in the July 2026 research. Taken off on 2026-10-06: the record cannot test it (7 passes ever had fewer than 100 holders, and the token-days carry no holder count), so it stood on no measurement while deciding whether a day-0 token is sizeable. Concentration (top10.*) is the sourced rug line; size is capped at 1% of the pool. A holders slice joins the record once the daily holder counts gathered since 2026-10-05 reach 7 days.
Up more than this today fails the sizing check (up more than this today)30% price change in 24hset by hand2026-07-28Set by hand.
Off: no volume floor on the sizing check (0 = the check does not run)0USD traded in 24hset by hand2026-10-06Policy, set by hand on 2026-10-06. What it is: a 24-hour volume floor for the sizing check; 0 means the check does not run. Set at $10,000 on 2026-09-22 as a tradability floor and taken off on 2026-10-06 after a read of the live record page on that day; no cut was frozen and no n recorded, so that reading is not cited here. The record page recuts the line daily and the live table is where its figures live.
Off: no turnover floor on the sizing check (0 = the check does not run)0% of market cap traded in 24hset by hand2026-10-06Policy, set by hand on 2026-10-06. What it is: a floor on 24-hour turnover as a share of market cap for the sizing check; 0 means the check does not run. Set at 5% on 2026-09-22 as a tradability floor and taken off on 2026-10-06 after a read of the live record page on that day; no cut was frozen and no n recorded, so that reading is not cited here. The record page recuts the line daily.
A pool over $100,000 doing less than this reads as fabricated or dead0.2% of the pool traded in 24hset by hand2026-07-28Set by hand
Off (0) since 2026-10-06: conviction no longer needs a decline from peak0% below all-time highset by hand2026-10-06Policy, set by hand on 2026-10-06. What it is: the decline from peak a conviction entry once needed; 0 means the check does not run and the record still shows it as a retired line. Set by hand to replace the base-proximity block, taken off on 2026-10-06 after a read of the live record page on that day; no cut was frozen and no n recorded, so that reading is not cited here. The record page recuts the line daily.
A pool drained more than this fails the token70% below its recorded peakset by hand2026-09-09Set by hand
At base band ends here30% above the 30-day lowset by hand2026-09-04Policy, set by hand on 2026-09-04. What it is: the band label 'at base', within 30% of the 30-day low. Chosen on 2026-09-04 after a read of the live record page on that day; no cut was frozen and no n recorded, so that reading is not cited here. The record page recuts the bands daily and shows what each did.
Just off base band ends here100% above the 30-day lowset by hand2026-09-04Policy, set by hand on 2026-09-04. What it is: the band label 'just off base', 30 to 100% above the 30-day low, the second band in order. The record page recuts the bands daily and shows what each did; no cut was frozen for this line and no n recorded.
Well off base band ends here; beyond is already ran500% above the 30-day lowset by hand2026-09-04Policy, set by hand on 2026-09-04. What it is: the band label 'well off base', 100 to 500% above the 30-day low; above 500% is 'already ran'. Chosen on 2026-09-04 after a read of the live record page on that day; no cut was frozen and no n recorded, so that reading is not cited here. The record page recuts the bands daily.
Two ranges for the model's guess (Jev): under 50% (unlikely to halve this week) and 50% and over (may well halve)0.5probabilityset by hand2026-10-06Policy, set by hand on 2026-10-06. Rationale: Decided 2026-10-06: two ranges a reader can use, under 50% (unlikely to halve this week) and 50% and over (may well halve). What each range did is cut daily from the record. The figures in this rationale were read from the live record page on 2026-10-06 and are not a frozen cut (no n recorded); the record page recuts the line daily.
No age floor: our hourly read judges a token from its first pool0days since the first poolset by hand2026-10-06Policy, set by hand on 2026-10-06. What it is: the age past the first pool a token needed before our hourly read judged it; 0 means no floor. Inherited at 14 days from the first configuration of the hourly scan and set to 0 on 2026-10-06 after a read of the live token-day table of the record page on that day; no cut was frozen and no n recorded, so that reading is not cited here. The gate judges age-related risk through its own checks.
No pool floor: our hourly read judges every priced token from its first pool0USD in the poolset by hand2026-10-06Policy, set by hand on 2026-10-06. What it is: the pool depth a token needed before our hourly read judged it; 0 means no floor. Inherited at $20,000 from the first configuration of the hourly scan and set to 0 on 2026-10-06 after a read of the live record page on that day; no cut was frozen and no n recorded, so that reading is not cited here. Pool depth now sets how often a token is re-judged (KEELAGE_REJUDGE_DEEP_USD), never whether.
A factory launch joins the universe only once its deepest pool holds at least this much; below it the pool check records no pool (rechecked once after 7 days)1,000USD in the poolset by hand2026-10-10Policy, set by hand on 2026-10-10. What it is: the smallest pool a launch needs before the pool check counts it as having a pool. Was 'any liquidity above zero' from the first pool check (2026-10-05); set to 1,000 because a pool under that cannot fill the smallest position the trading check would size, and dust pools inflated the universe count. No cut was frozen and no n recorded. Separate from sweep.minPoolUsd (the judge floor, 0): this line decides whether a launch is in the field, not whether it is judged.
No market-cap floor: our hourly read judges tokens at every market cap0USD market capset by hand2026-10-05Decided 2026-10-05: a new service includes every token at every market cap; the inherited $50k floor came out.
No market-cap ceiling: our hourly read judges tokens at every market capnoneUSD market capset by hand2026-10-05Decided 2026-10-05: a new service includes every token at every market cap; the inherited $50M ceiling came out.
Scout size: this share of the pool, in dollars0.5% of the poolset by hand2026-10-06Set by hand on 2026-10-06: half the conviction share, since a scout tier means a thin pool or a concentration between 30 and 45%. The record tests the share.
A token gets its trading check the first day it trades at least this much; it is not re-judged on a schedule after that1,000USD traded in 24hset by hand2026-10-06Set by hand on 2026-10-06. A cadence line, not a verdict line: it says when the one-time trading check runs, never what it answers. 878 tokens traded this much on 2026-10-06. Re-judged only when a watch trigger fires (sells stop clearing, pool down 70%, a large holder leaves).
A passed token is watched (volume, momentum, X, holders) from this market cap; below it we don't keep reading what changed since50,000USD market capset by hand2026-10-06Set by hand on 2026-10-06: a token too small to trade is not worth the reads. A cadence line, not a verdict line.
From this market cap a watched token is read every 6 hours; below it once a day250,000USD market capset by hand2026-10-06Set by hand on 2026-10-06. Cadence only.
From this market cap a watched token is read every hour1,000,000USD market capset by hand2026-10-06Set by hand on 2026-10-06. Cadence only.

The outcome views are re-cut every day. A value in this table changes by decision, with the date of the change on its line.

What a row is, and what it isn't.

A pass is a token that cleared our structural checks at the hour our scanner saw it. We log its price then, and read it again 24 hours, 7 days and 30 days later.

How we count

Return is the change in price from the pass to the read. We use the median, not the mean, because one 60x hides 40 zeros. "Worth more" is any gain. "Halved" is -50% or worse. Every table carries an all-passes row, because in this market most tokens go down: a group is only telling you something when it differs from that row.

Two records, two limits

The pass ledger holds the holder, pool and spread facts at the moment of a pass, and only for tokens that passed. The every-token record has no such start and no such filter: every token from its first pool to today, pass or fail, but only the facts a price series carries.

Both grow by the day, and every view is recomputed every night from the whole of each.

Two checks

The code check is read the hour a pool appears, from the chain alone: verified source, a launcher we score, no live owner privilege unless renounced, no transfer-tax code, no honeypot flag. The trading check is the full check made later: the code again plus sells going through, the pool not drained, how spread out the holders are, and whether the token is not up more than 30% that day with real trades going through. The every-token section shows what clean code on day 0 did next; the passed sections show what the trading check did.

The checks

The rules list every line we use with where it came from. The outcome views test each one every day: on every pass, and for the price-read lines on every token, every day. A line earns its place when the side it keeps did better than the side it drops; where it did not, the table says so and the line is either changed or named as what it is. No third-party score decides anything: the checks read the verified code, the launcher, the owner’s powers and the tax code from the chain.

Distance from the 30-day low

We group that distance into 4 bands: at base (within 30%), just off base (30 to 100% above), well off base (100 to 500%), already ran (more than 500%). The band is on every passed token; the band lines on the rules table are measured from the passed and every-token views.

How spread out the holders are, and what we'd put in

What we'd put in comes from how much of the supply the 10 largest wallets hold, with pools and contracts removed. The 30% line comes from a 2025 study of 34,988 meme coins across 4 chains (arXiv 2507.01963, section 5.1.3), where the 10 largest holders owning over 30% of the supply was the most common anomaly; the 45% line is where a handful of wallets can set the price alone. The spread decides how much we'd put in, never whether a token passes. Concentration is a rug vector whatever the returns say.

Time to outcome

The horizon table reads every pass at each distance, and the day-one table splits the week by where the price stood after the first day. Both are recomputed from the whole ledger every night.

Launchers

Most tokens on this chain are stamped out by a launcher, a factory contract that deploys the same code for everyone. The landscape counts every launch by launcher from the factories’ own events, and the launcher table under At the pass shows what passes from each did. “Clean” code says nothing about the people behind a token.

Probability check

A model's guess at the odds a token halves this week comes from Jev, a model that answers a typed question with a probability. It is asked two things about every pass: “will this halve within 7 days?” and “will it be up in 7 days?”. The halving answer is shown in two ranges, under 50% and 50% and over, and the table shows how often each range did halve. The “up in 7 days” answer carries no words. Neither probability admits, blocks, sizes or ranks a token.

Holder growth

From 6 Oct 2026 our daily read’s holder count joins both records. On the passes: the count at the pass against the count 7 days on, in 4 bands, against the 7-day return. On every token, every day: the same growth over the prior 7 days and over the next 7 days, with the pool read that day. The daily read started 5 Oct, so these tables fill as the rows age; an n of 0 is honest, not broken. Holder growth is context beside the checks and never moves one.

Not measured yet: what the deployer did

Creator fees claimed, tokens bought and burned, money moved off the chain, and the 7-day return after each pattern. The tool that reads this ships next; the table needs a few weeks of its output before there's anything honest to show.