How it works
If you have never provided liquidity, never used leverage, and have no idea what a “5th percentile” is, this page is for you. It explains what the system does, who pays for it, and — the part most products skip — what it cannot tell you.
The problem this exists for
There are three popular ways to earn money in crypto, and each has a way of losing it that almost nobody sees coming until it has already happened.
The yield looks great. You still withdraw less than you put in.
You park two tokens in a pool and earn a cut of every trade. Dashboards advertise the fees — 40%, 200% a year — and never mention that while you earned them, the pool quietly swapped you out of the token that went up and into the one that went down. The name for that is impermanent loss, and it is permanent the moment you withdraw.
You were right about the direction and still lost everything.
Leverage borrows against your money so a 1% move feels like 20%. It cuts both ways: at 20× leverage, a 5% move against you closes your position and keeps your deposit. On most coins, 5% is a normal hour. You do not get the position back when the price comes right.
Getting in was easy. Getting out is the trade.
A small coin's price is quoted for a trade of zero size. Buy $5,000 of something with a shallow pool and you move the price against yourself on the way in and again on the way out — often 15% round trip before your idea has to be right about anything. And the moment you most want to sell is the moment everyone else does too.
The idea, in one paragraph
Somebody has to find out whether a strategy works. Normally that somebody is you, and you find out with your own money. Here, the community submits strategies, a machine tests them against thousands of simulated futures, and the testing is paid for out of ecosystem transaction fees. Discovering that a strategy fails costs the ecosystem about a dollar instead of costing you a position.
How a strategy travels through the system
- 1Someone submits
any connected wallet
Not an opinion — a form of numbers. Which coin, how much, where the stop goes, how long to hold. Numbers are the only thing a machine can actually test.
Costs you: Free
- 2The queue decides order
the community
One vote per wallet, and you can pledge budget to push something up the line. Popularity decides what gets looked at first — and nothing else.
Costs you: One signature
- 3The lab tests it
a simulator, no AI involved
Thousands of possible futures, each one running the strategy start to finish, with real trading fees and slippage charged along the way.
Costs you: ~$0.85 from the fee pool
- 4A gate judges it
fixed published rules
The results are checked against thresholds set in advance. The gate cannot see who submitted it, how many votes it has, or what the thesis says. It reads numbers.
Costs you: Nothing
- 5You see everything
anyone, no login
Certified, rejected, or undecided — the full distribution is published either way, along with the seed that lets anyone re-run it and get the same answer.
Costs you: Nothing
Steps 3 and 4 are where this differs from every “alpha group” you have been added to. The thing that decides whether a strategy is endorsed is arithmetic that cannot see the author, the hype, or the vote count.
What a test actually does
The lab does not try to predict what the price will do. Nobody can, and a system that claims to is selling something.
Instead it takes how much this particular coin has actually been moving lately, and generates thousands of possible futures with roughly that much movement in them. Then it runs the strategy through every single one and writes down what happened.
Out of that you get the two numbers that matter: the typical outcome — what happens in the middle of the pack — and the bad case, the worst 1 result in 20. Most products show you the first number. The second one is the one that decides whether you can live with the strategy.
This is not a backtest. A backtest replays real history. The free market data this runs on has no price history behind it — only a current price and how much it moved recently. Rather than invent a history and call it real, the lab is upfront: it is a stress test, and it answers “how often does this blow up”, not “how much would I have made last year”.
How to read a result
Example strategy
EXAMPLE · by 0x1234…abcd
Made-up numbers, shown only to label the parts. Real cards look exactly like this and are never this tidy.
- MedianThe middle of the pack. Half the simulated futures did better than this, half did worse. Not a promise — a centre of gravity.
- Bad caseThe 5th percentile: 1 future in 20 was worse than this. This is the number to size against. If you cannot sit through it, the strategy is not for you regardless of the median.
- RuinHow often the strategy gave back more than half the money. Above 5% and the gate refuses to certify, no matter how good the median looks.
- RunsSeparate tests, on separate market snapshots. Fewer than 3 and the verdict stays “undecided” — one good test is a coincidence.
What “certified” means — and what it does not
It does mean
- It was tested at least 3 separate times, across 1,500+ simulated futures.
- The middle-of-the-pack outcome made money after all fees and slippage.
- The bad case lost no more than 25%, and fewer than 5% of futures lost more than half the money.
- A machine checked those numbers against thresholds published in advance. No human approved it, and no amount of votes can move it.
It does not mean
- That your next trade wins. Certification describes a spread of outcomes, not one.
- That you cannot lose. The bad case is printed right there — plan for that number, not the good one.
- That it will still be true next month. Markets change; the test is a snapshot.
- That anyone is advising you. This is a measurement, not a recommendation, and nobody here is a licensed advisor.
Certification is deliberately hard to get, and most submissions do not get it. A lab where everything passes is a lab that is not measuring anything — the rejections are the product working, not the product failing.
Who pays for all this
- Step 1
People trade in the ecosystem
Every swap pays a fee, the way it does on any exchange.
- Step 2
A slice of those fees is routed to a testing pool
Not to a team wallet. To a published, append-only ledger you can read on the vault page.
- Step 3
The pool pays for strategy tests
Each run costs about $0.85 in compute and is charged to the ledger with the strategy it paid for attached.
- Step 4
Tested strategies are free for everyone
Including the people who never traded, never voted, and never paid anything.
When the pool runs dry, testing stops until fees top it up again — a budget that keeps spending past empty is not a budget.
Read the pool ledger →One caveat stated plainly: pledging budget to a strategy currently records a pledge and moves it up the queue. No money actually changes hands — there is no payment system wired up yet, and pledges are excluded from the pool's spendable balance for exactly that reason.
What you can do
You want to know which strategies survive contact with a bad week.
- Read certified strategies and the full distribution behind each one
- Vote on which submission the pool tests next
- Follow the public ledger of every decision the agent makes
You have a setup and want it stress-tested before you size into it.
- Submit your strategy as parameters — it costs you nothing
- Run a test on demand instead of waiting for the queue
- Get credited by name if it certifies and the agent adopts it
You hold the token and want a say in what the treasury spends its budget on.
- One vote per wallet on the testing queue
- Pledge budget to push a specific strategy up the queue
- Audit the pool ledger — every dollar in and out is a line item
Connecting a wallet signs a message and nothing else. It never asks for a transaction, a token approval, or a transfer — there is no code in this app that can move your funds. It exists so a vote can be counted once and so credit can be attached to whoever earned it.
Open the lab →Nothing here is financial advice. Nobody here is a licensed advisor.