Most risk pages are boilerplate written to be skipped. This one is written to be read. It explains, in plain language, what this platform is, how meme-coin markets actually work, every major way people lose money in them — including the failure modes of research methods like ours — and exactly what we do with your data.
Effective date: 2026-07-02 · Last updated: 2026-07-19
Meme-coin trading can result in the loss of ALL funds involved — quickly, permanently, and with no recourse. Never use money you cannot afford to lose entirely.
Alphalumina measures the meme-coin market on Solana — primarily tokens launched on pump.fun and similar launchpads. If you use this platform, you are looking at data about one of the highest-risk, most manipulated, least regulated markets that has ever existed. That is not marketing language. It is the literal, measurable truth, and our own numbers prove it:
| Measured fact | Value | Sample |
|---|---|---|
| New tokens that die shortly after launch | ~85% | ongoing launch stream |
| Tokens that trended and STILL round-tripped (pumped, then gave it all back) | 56% | n=739, kline-reconstructed |
| Tokens that reached $100K market cap and round-tripped anyway | 63% | same cohort |
| Median peak multiple after trending | 1.68x | same cohort |
| Median value remaining after the peak | ~41% of entry | same cohort |
| Tagged "KOL / smart money" wallets net-negative on realized trades | the majority — see /kols | on-chain, realized flows |
The single most important sentence on this site: reaching a milestone is not the same as being tradeable. Most tokens that "succeed" by any headline metric — trending, graduating, hitting $100K — still round-trip. If you hold through the peak, the median outcome in our data is losing more than half your entry value. The people who make money in this market are overwhelmingly the ones who sell into strength, not the ones who believe.
If reading that table makes you want to close this tab — that is a rational response. We built this platform because these numbers are systematically hidden by tools that profit from your excitement. We make nothing from your trades. We do not want you to trade at all unless you fully understand everything below.
↑ back to topYou cannot judge risk in a market you do not understand mechanically. Here is the machinery, stripped of jargon.
On pump.fun, anyone can create a token in seconds for a few dollars. No team, no product, no vetting. The token starts on a bonding curve — a smart contract that acts as an automatic market maker: price rises as people buy from the curve, falls as they sell back. Early buyers get the lowest prices by construction, which is exactly why launches attract automated snipers (bots that buy in the same block the token is created — faster than any human can).
When enough money has bought into the curve (roughly a $69K market cap), the token "graduates": its liquidity migrates to a public decentralized exchange. Graduation is often sold as an achievement. Our data says it is not an outcome — in our current lifecycle cohort, not one observed graduation has gone on to reach $100K yet, and across the broader reconstructed cohort most graduates still round-tripped. Graduation changes the venue. It does not mean the token "made it."
Market cap is a headline; liquidity is the truth. A token can display a $500K market cap with only $15K of real liquidity in the pool — the paper value mostly cannot be realized. If you sell into thin liquidity, the price collapses beneath you as you exit. Every major loss pattern in this market is a version of "the market cap was a mirage; the liquidity was never there." When liquidity is withdrawn entirely, that is a rug pull — the token becomes worthless instantly, with no recourse, no support line, no regulator to complain to.
Meme-coin trading is adversarial. When you buy, someone sells to you. Know who they typically are:
Launch → a few minutes of sniper and bot activity → either immediate death (the vast majority) or an attention spike → a pump measured in minutes-to-hours, not days → distribution (early holders selling to late arrivals) → round-trip or liquidity death. A small number of tokens have second lives ("revivals") — genuinely new episodes of attention after the first death — and an even smaller number become durable runners. Our platform exists to measure which is which after the fact, honestly, because predicting it in advance is largely unsolved — by us and by everyone else, whatever their marketing says.
↑ back to topAlphalumina is an early-stage research platform under active development. Features change, measurements get revised, models get retired when they fail validation, and parts of the pipeline break and get fixed. Numbers young enough to carry small sample sizes are labeled provisional — a provisional number is a hypothesis, not a fact. Treat the whole platform accordingly.
↑ back to topNothing on this site is financial, investment, legal, tax, or trading advice. Alphalumina is research and measurement software. No content here — no brief, radar entry, score, grade, receipt, scorecard, or chart — is a recommendation, solicitation, or offer to buy or sell any asset. We are not licensed advisors, brokers, or dealers in any jurisdiction.
Assume any meme-coin position can go to zero, quickly and permanently. These assets have no cash flows, no underlying business, and in most cases no reason to exist beyond attention. Prices are driven by reflexive crowd behavior and can drop 90%+ in minutes.
This market contains rug pulls, honeypots (tokens coded so you can buy but not sell), wash-traded volume, bundled fake holders, paid shills, coordinated pump-and-dumps, bot swarms, and impersonation scams. Assume manipulation is the default condition, not the exception. No tool — ours included — detects all of it.
Smart contracts can have bugs or backdoors. Networks congest at exactly the moments everyone wants to exit. Wallets can be drained by malicious token approvals or phishing. RPC providers fail or lag. Self-custody means mistakes are irreversible: no password reset, no chargeback, no insurance.
The legal status of meme-coins is unsettled and varies by jurisdiction. Rules can change, sometimes retroactively. Trading may be restricted or illegal where you live — knowing this is your responsibility, and the Service is not directed at any jurisdiction where accessing it would be unlawful. Trades are generally taxable events; high-frequency meme-coin trading can create complex tax obligations even in losing years. Consult a qualified tax professional.
This market is engineered — by its structure and by the people profiting from it — to produce FOMO, revenge trading, and addiction-like engagement loops. The casino does not close. If trading is affecting your finances, sleep, or relationships, stop. If you are in a difficult financial situation, this market is the worst possible place to try to fix it: the measured odds above apply most brutally to people who cannot afford the losses.
↑ back to topEven when a token "goes up," turning that into realized profit is a separate problem. This distinction — an eventual runner versus a realistically tradeable opportunity — is central to how we grade everything, because it is where most honest-looking track records lie.
A token can be down 90–100% from a signal before later becoming a runner. A caller who said "it did 10x eventually" is not lying — but a real position entered at their call would have been wiped out or shaken out long before the run. In our forward testing, tokens in exactly this pattern went through complete drawdowns before their eventual runs. We never count an eventual runner as a successful call when the original entry would have been destroyed first — and you should apply the same test to every track record you see anywhere.
On thin liquidity, the act of buying moves the price against you, and the act of selling moves it against you again. A "10% move" can be fully consumed by the round-trip cost of getting in and out. Displayed prices are for marginal, tiny trades — not for your size.
Decentralized trades route through pools and aggregators. Routes fail, quotes go stale between signing and execution, partial fills happen, and failed transactions still cost fees. During volatile moments — precisely when you most want to act — fill quality degrades sharply and front-running (MEV) extracts value from your transactions.
Liquidity can be withdrawn by its providers at any moment without warning. You may be unable to exit at anything near the displayed price, or at all. Deciding your exit before you enter is the only defense the data supports: holding through peaks is the measured default losing move.
↑ back to topAlphalumina's core promise is honest measurement. Part of honest measurement is telling you exactly how research like ours goes wrong. Every failure mode below is real — several are things our own instruments did before we caught and published them.
Every score, grade, lane assignment, and transition odd on this platform is a statistical statement about the past behavior of a category — never a prediction about a specific token's future. A token with "good" measurements can die instantly; a rejected token can run.
Backtests replay history with today's knowledge and hindsight-clean data. They systematically overstate what would have been achievable: real-time data is messier, fills are worse, and decisions made under uncertainty differ from decisions graded after the fact. We treat backtests as hypothesis generators only; nothing goes into our validated column without forward, out-of-sample evidence.
Datasets quietly lose their corpses. Our own early cohort statistics looked far too good because our scanner stopped watching tokens shortly after they trended — we saw the pumps and missed the deaths. The published 56% round-trip figure exists because we rebuilt the cohort from full price history and found the bodies. Any statistic that seems too good — anywhere — should make you ask: where are the dead ones?
How you pick what to measure changes the answer. When we first harvested influencer calls by searching tokens that had already moved, every account looked like a winner — because we had only collected calls on winners. We rebuilt collection around full account histories, losers included, and hit rates fell from a fake 85–100% to an honest ~37%. Every leaderboard you see anywhere was built with some selection rule; most are never disclosed.
Using information that was not actually available at decision time — grading a "call" with data from an hour later, or scoring an entry with the day's eventual peak. Our grading is timestamped to what was knowable at the moment of each call, and our out-of-sample tests split strictly by time (classify early, measure later). When you evaluate anyone's track record, ask what they knew when.
Test enough signal combinations against the same history and some will fit it perfectly — by chance. Those patterns evaporate on new data. It is why most published "strategies" fail live, and why our pipeline separates research, shadow testing, and validated status: a pattern only graduates when it keeps working on data it has never seen.
Parts of our pipeline — classification, text analysis, anomaly detection, written summaries — use AI models. AI systems make mistakes: they misclassify, hallucinate plausible-sounding but wrong statements, inherit biases from training data, and degrade when the market shifts away from what they learned. AI output on this platform is subject to the same rule as everything else — it must survive measurement — but between measurement cycles, errors can and do appear. Never act on any single AI-generated statement without independent verification.
Paper-trading results are simulations. They model fill costs deliberately, but cannot fully capture slippage under stress, failed transactions, MEV, congestion, or the psychology of real money. Simulated results systematically flatter reality. All paper results on this platform are labeled simulated; treat them as an upper bound, not an expectation.
All grades and receipts are historical measurements, produced after the fact. Past outcomes — including honestly graded ones — do not predict future results. Our own persistence testing found that even track records that look real often fail to persist: wallets with good histories performed no better on future trades than wallets with bad ones.
↑ back to topOur measurements are built on external sources: on-chain data via RPC providers, market data from charting and screener services (e.g., GMGN and similar), social data from X and related feeds, news feeds, and wallet-data providers. Each can be delayed, rate-limited, incomplete, revised, or simply wrong — and each can change or shut off access without notice. We filter, cross-check, and flag anomalies, but garbage occasionally gets through; several of our published corrections exist because an upstream source lied to us convincingly.
Data is collected on cycles, typically minutes apart. Sub-minute spikes and deaths fall between snapshots. A number on this site was true when collected; the market may have moved since. Where data is missing we aim to say "missing" rather than silently show zero — if you see a number that seems impossible, treat it as suspect and tell us.
Our source scorecards measure what is publicly visible. KOL and dev records cover publicly linked wallets only — many trade from undisclosed side wallets invisible to us and to everyone. Social metrics can be botted; follower counts and engagement are manipulable and are treated as metadata, never as reliability. Whale labels identify wallets, not intentions — a "smart money" buy can be the setup for a dump. Every scorecard on this platform is incomplete by construction, and we say so on the page rather than pretend otherwise.
The Service can go down — maintenance, upstream failures, infrastructure faults, or plain bugs. Pages regenerate on schedules and can temporarily show stale editions. We monitor our own pipeline health and publish data-quality warnings, but you should never build a time-critical process on the assumption that this (or any) platform will be up at the moment you need it.
Nothing here removes your obligation to verify before acting. Check contract addresses against multiple sources. Confirm liquidity on-chain. Cross-reference any measurement that would change your decision. The correct amount of trust in any single data source — including us — is: enough to investigate further, never enough to act alone.
↑ back to topBy accessing alphalumina.com or any Alphalumina service ("the Service"), you agree to these terms. If you do not agree, do not use the Service.
Alphalumina provides research, analytics, and measurement tooling for Solana-ecosystem tokens. The Service is informational. It does not execute trades, custody funds, provide individualized advice, or act as a broker, dealer, exchange, or advisor of any kind.
You must be of legal age in your jurisdiction and legally permitted to access this kind of information where you live. You are solely responsible for every decision you make, for compliance with your local laws, and for your own taxes. Using the Service creates no fiduciary, advisory, or professional relationship between you and Alphalumina.
You may not: misrepresent Service output as investment advice (yours or ours); resell, scrape, or redistribute the Service or its data without written permission; use the Service in connection with market manipulation of any kind; present receipts, grades, or scorecards out of context in a way that implies return promises; or attempt to disrupt, probe, or overload the Service.
The Service's content, methodology descriptions, and published measurements are ours. On-chain data itself is public. You may quote our published findings with attribution and a link; you may not republish them stripped of their caveats and sample sizes — the caveats are part of the finding.
The Service links to third-party tools (block explorers, charting sites). We do not control them, do not endorse them, and are not responsible for what you encounter there.
The Service is provided "as is" and "as available," without warranties of any kind, express or implied — including accuracy, completeness, timeliness, merchantability, fitness for a particular purpose, and non-infringement. Data can be delayed, incomplete, or wrong; upstream sources fail; measurements get revised when better data arrives (we log revisions rather than hide them).
To the maximum extent permitted by law, Alphalumina and its operators are not liable for any direct, indirect, incidental, consequential, or special damages — including trading losses, lost profits, or lost data — arising from use of, or inability to use, the Service, even if advised of the possibility. Your sole and exclusive remedy for dissatisfaction with the Service is to stop using it. Where liability cannot be excluded, it is limited to the amount you paid us for the Service in the twelve months before the claim (currently: nothing).
You agree to indemnify and hold harmless Alphalumina and its operators from claims arising out of your use of the Service, your violation of these terms, or your violation of any law or third-party right.
We may modify, suspend, or discontinue any part of the Service at any time, and may update these terms as the platform evolves. Material changes will be reflected by the "last updated" date at the top of this page; continued use after changes constitutes acceptance. If any provision of these terms is found unenforceable, the remaining provisions stay in effect.
↑ back to topThe wallets, transactions, and token data we analyze are public blockchain information, readable by anyone. When we publish measured records for dev or KOL wallets, we publish the public identity those wallets already carry (e.g., the handle major tools tag them with) and measured on-chain facts — never private information, because we have none.
We do not sell, rent, or share your email or any personal data with third parties for their marketing. We do not build advertising profiles. We do not buy data about you.
Platform accounts are early-access and invite-only. You can export your own created data (watchlists, saved views, alert rules, notification history) from Settings at any time, and request account deletion there — deletion disables sign-in immediately and queues removal of your personal records. Security audit logs (sign-in events) are retained for a limited period for abuse prevention. The platform is experimental software; see the risk sections above.
Waitlist emails are kept until you ask us to delete them or the waitlist is retired. Want your email removed? Contact us (section 10) and we will delete it. Depending on your jurisdiction (e.g., GDPR, CCPA) you may have additional rights — access, correction, deletion, portability. Contact us and we will honor them; with the little data we hold, this is easy.
If we later add accounts, analytics, or any new data processing, this policy will be updated before the change ships, and the "last updated" date at the top of this page will move.
↑ back to topQuestions about these terms, a privacy request, a data error you spotted, or anything else:
If you believe a published measurement is wrong, please include the page, the number, and the time you saw it — corrections are part of the product, and verified errors get fixed and logged publicly.
One last time, because it matters: this platform exists to measure a market where most participants lose. Use it to understand what is real. Do not use it as a reason to believe you will be the exception — the numbers on this page are the numbers.