Questions people ask before paying
Every answer here is true of what ships today, and the ones carrying a number take it from the same module the product does. Where an answer would be a hope rather than a fact, the question is not on this page.
The basics
What is Fortell?
A funded-trader protocol for Solana prediction markets. You buy a 14-day evaluation, trade a challenge ledger against live venue prices under risk rails, and are graded on five criteria at once. On a pass, Fortell provisions a vault account for you and the shared pool allocates real backer capital into it. You are named as a trade-only delegate on that account: you can trigger a venue swap that the vault signs, and you can never move cash out of it. The full reference is here.
Is this live on Solana mainnet?
No. The reference publishes a per-cluster deployment table for all four programs, checked by calling getAccountInfo on each program id rather than by remembering. The app defaults to a mainnet fork. Read the table, with the date it was checked.
What do I pay, and what does it buy?
$75 for Standard or $150 for Pro. That is an entry fee for the evaluation, and it is the only payment the product asks for. It buys 14 days of trading against live prices and a graded decision at the end of it. A pass provisions a $500 seat on Standard, $1,500 on Pro. The tier contract shows the arithmetic behind each number.
Do I have to deposit or stake my own capital?
No. There is no deposit, no stake and no collateral at any point. A funded seat holds backer capital, and the first-loss buffer that sits between a seat's losses and backer principal is funded by Fortell, not by you: $125 on a Standard seat and $300 on Pro, which is the seat size times its maximum drawdown. How the buffer works, and how to read its current size off the chain.
Does Fortell hold my funds?
No off-chain service holds a user key. Your wallet signs; the seat is a Solana account and its own program-derived address signs the venue call. A claim pays a destination pinned to the seat rather than to whoever signed, so it does not matter who submits it. The division of authority is written out here.
Passing the challenge
What are my chances of passing?
In Fortell's own simulation, a trader with no edge at all who plays the rules perfectly and stops the moment all four criteria are met passes Standard 11.7% of the time, and Pro 3.8%. That is 25,000 simulated 14-day challenges behind each of those two figures, measured 2026-08-08 against the rails that ship, on 1d buckets. Those conditions are load-bearing: it is a zero-edge ceiling under optional stopping, a trader who keeps trading after qualifying does worse, and the rules it ran against and the bucket it was measured on are part of the figure rather than footnotes on it. The number moves whenever the rules move, so the same percentage quoted against a different ruleset, or against a different bucket, is not a weaker claim, it is a different one. That is why this one is measured again when a rail changes rather than carried forward. One more condition, and it is the one we got wrong ourselves: both percentages are the best play a search over strategies found, not a figure read off a formula. Across nine repeat runs of that search the answer moved as much as 1.4 points. So read them as bands rather than as points, and take that width as the resolution of the instrument. A movement bigger than that is something real; a movement smaller than that is our instrument, and it is evidence for nothing in either direction, including that a rule which moved it by nothing did nothing. The rate from real decided challenges is on the transparency page, where it says “not yet” until challenges have actually been decided. How grading works.
What am I graded on?
Five criteria at once, all five of which have to hold. On Standard: at least 30 scored calls, a forecast score of at least 55, drawdown never past 25%, profit of at least +4%, and no more than 50% of that profit from a single day. Pro asks for 50 calls, a score of 60, 20% drawdown, +5% profit and the same 50% limit on any one day. That last one is a bar to clear rather than a rule to breach: being over it does not end your run and does not fail you, and it falls as other days earn. The decision is made off chain, by an indexer, and the reference says so plainly rather than implying the chain verifies a grade. Evaluation and grading.
What can I actually trade?
BTC Up or Down rounds on Jupiter Forecast, through the BisonFi prop AMM that Forecast routes to. One venue is funded-tradable today. A second adapter exists in the vault program, and every instruction on it that could put capital in or take it out is refused on a pool-funded seat by an explicit check rather than by convention, because that path carries custody guards but no risk rails. The venue, and what it is.
What stops me from blowing up a funded seat?
Five rails in the vault program: a per-round size cap, a total open-exposure cap, a minimum hold, a drawdown floor against a high-water mark, and a daily loss floor. They are checked against the real token-balance change of the venue call rather than against the venue's instruction bytes, so a breaching buy is refused rather than caught afterwards. Three of them are per-seat fields that provisioning has to set, and zero means that rail is off, so whether a given seat is railed is something to read rather than assume. The reference shows you how to read a seat's rails yourself.
Getting paid
When can I take profit out?
Any time your seat has no open position. Everything above your seat size is claimable then. There is no payout window, no minimum hold on the payout and no approval step. Rounds on the live venue run in minutes, so flat is never far away. The payout model.
How is a payout split?
70% to you, 20% to backers, 10% to Fortell, in the same instruction. Your seat returns to its starting size and its drawdown floor resets with it. Each seat stores its own split on its vault account, and the transparency page reads that field per seat, so if a published split ever disagreed with the stored one, the stored one is the one that pays.
Can Fortell refuse or delay my claim?
The claim instruction has no admin gate, no approval account and no timer that Fortell holds. Once a seat has sat flat past the grace period it is permissionless outright, so a stranger can cause the payout to happen and still cannot change where it goes. How the claim is constrained.
Is my unrealised profit mine?
Not until you claim it. Profit you have not claimed is equity in the seat and moves with the seat, so it is at risk exactly like a real trading account. A seat that runs up $200 and then gives back $125 has $75 of claimable profit, not $200, and that $75 is before the split. Claiming is how profit becomes yours.
Backing a pool
What does a backer earn?
20% of every trader claim, at the moment of the claim rather than when a seat closes. Backers deposit into the shared pool and hold Token-2022 shares against it. The shared pool, and what protects it.
What protects backer principal?
A first-loss buffer of Fortell's own cash, sized at seat size times maximum drawdown, held inside the pool and enforced by a check that refuses any settlement it cannot fully cover. The allocation instruction refuses to fund a seat whose buffer it cannot reserve. The buffer's current size is a live read rather than a constant, and the reference is emphatic about that because an earlier internal document hard-coded it and was wrong a few hours later. Read it yourself, with the exact call.
Can I see what is actually in the pool?
Yes, decoded in your browser from the chain rather than served from a database. The transparency page shows net asset value, what is working in seats, the first-loss buffer, every live seat's address and each seat's stored split. It prints “not yet” instead of zero when a number has no sample behind it.
Trust
The two questions below have unflattering answers. They are here because leaving them out would be worse than answering them.
Has this been audited?
Not by a third party. There is an internal adversarial audit of the on-chain programs and a formal verification effort with machine-checked properties, and the verification page publishes its own caveats rather than a green tick: a passing property test is not the same thing as a true property, and this repository has two documented cases where it was not, including one that compiled to x >= x. What was proved, and what the proofs do not cover.
Who is behind Fortell?
Nobody is named on this site yet. That is a real gap and it is recorded as one internally rather than being left unnoticed. What can be checked without knowing who built it: the programs are on chain and their addresses are published, every risk rail is a field you can read off a seat account, and the reference has a page of exactly those calls.
Are the published docs the whole set?
The technical reference is. A few files alongside it are internal operations material rather than a product reference, so they are not on the site. The rule worth knowing is the one that governs everything else: a statement about how the protocol works is published however unflattering it is, and that is what nearly the whole book is. The reference is rendered from those files themselves, so it cannot quietly drift from what the repository says.
Something went wrong. How do I reach a person?
The contact page has the current answer, including what you can settle yourself from the chain without waiting for anybody.