> For the complete documentation index, see [llms.txt](https://docs.fermi.trade/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.fermi.trade/19-risk-warning.md).

# 19 risk warning

Trading perpetual futures is high-risk. Read this page carefully.

## You can lose more than you think

Perp trading uses leverage. A 5 % adverse move on a 10× position wipes out 50 % of your collateral in that account. Your entire collateral can be wiped out in seconds during fast markets.

You may also lose **more than the position you opened**: if the market gaps through your stop-loss before any matcher pass runs, you will be liquidated at the post-gap price, not at your stop. The liquidation discount, platform fee, and any insurance/socialization events apply on top of the price loss.

## Operational risks specific to Fermi

These are above and beyond market risk:

### Oracle pause / staleness

If the Pyth (or Switchboard) feed for a market goes stale beyond `max_staleness_slots`, **every read against that oracle reverts**. Symptoms:

* You can't place new orders on that market until the feed comes back.
* You can't be liquidated on that market either, but you also can't *close* yourself.
* Funding for that market freezes (no funding accrual until the next successful `update_funding`).

Pyth pull updates can be triggered by anyone. If your bot stalls because of staleness, push a fresh price update yourself.

### Relayer outage

If the relayer is down, your intents can't be sequenced via the fast path. You can still:

* Submit through the **direct fallback pool** (see [21 - Direct Fallback Pool](broken://pages/31cdbebbf779c24c05fcc4cf4188188d5f306bc7)) — your intent is staged on chain and the on-chain executor will admit it after the speed-bump.
* Use the on-chain `perp_cancel_*` instructions directly with Anchor — these don't go through the queue at all (they're admin-free instructions that operate on the orderbook directly).

### Solana congestion

Solana periodically experiences periods of high congestion. During these, transactions may not land for tens of seconds. Implications:

* Settle PnL bonuses are highly competitive and may take retries.
* Liquidation racing is competitive; the first to land wins. If you're a liquidator bot, use priority fees.
* A user's `cancel` may not land in time. Always assume the order could fill before your cancel does.

### Stable-price drag

When the oracle moves quickly, your `init_health` lags because the stable price hasn't caught up. This is **protective for the system** but means your effective borrowing power is reduced during volatile moves. Plan accordingly: don't compute "max position" off the oracle alone.

## Self-trade and replay protection

Self-trade prevention modes do exactly what they say. If you set `AbortTransaction` and your bot accidentally collides with its own order, the entire intent fails — and a slot in the v5 queue is burned. Use `CancelProvide` if you want a never-fail replace flow.

The on-chain replay cache rejects identical intents within `~10` slots / `512` entries. If you replay a signed intent quickly, the second submission gets `ExecutionQueueV5IntentReplay`. Sign a fresh intent (with new nonce / `client_order_id`) for each retry.

## What the protocol does **not** protect you from

* Bugs in your own off-chain code (wrong size, wrong direction, wrong account).
* Wallet compromise or signing-key theft.
* RPC outages / bad blockhashes.
* Loss of access to your authority keypair.
* Off-protocol scams (social engineering, fake "Fermi" sites, etc.).

## What the protocol **does** protect you from

* A malicious v1 sequencer/relayer can't silently reorder your transactions relative to the emitted [POSq order](broken://pages/ac31e9ddbaaf6ba0dc00cd9026f42e08117ea99d) — encrypted VDF ticks, the commitment trail, and on-chain AMQ queue commits make reordering detectable.
* A malicious relayer can't substitute account lists — the canonical user-intent message (`canonical_user_intent_message_v2`, `instructions/execution_queue_v5.rs:671-682`) binds the account list into the user's signature.
* A replayed signature can't fire your order twice — the on-chain replay cache catches it.
* The relayer can't take custody of your funds — every state-modifying instruction requires *your* (or your delegate's) signature, verified inside the on-chain program.
* The matcher can't fill you at a worse price than `max_quote_lots` caps — set both base and quote caps for hard slippage limits.

## Recovery paths

{% stepper %}
{% step %}

## Check the trace endpoint

`/trace/sequence/{market}/{seq}` reports every stage of an intent (or absence of it). Most "lost orders" turn out to be intents that never reached the POSq/relayer fast path.
{% endstep %}

{% step %}

## Use direct submission

Bypasses the fast path entirely for cancellation / close-only flows.
{% endstep %}

{% step %}

## Use the Anchor IDL directly

All `perp_cancel_*`, `token_withdraw`, etc. are direct on-chain instructions you can sign and submit yourself with no execution queue at all.
{% endstep %}

{% step %}

## Reach out to support

See [29 - Support](broken://pages/affc0a808b406efad6224bf9735688ee3b9ade77).
{% endstep %}
{% endstepper %}

## Jurisdictional note

Fermi is open-source software deployed to a public blockchain. Use of perpetual futures may be restricted in your jurisdiction. Operators of front-end UIs are responsible for restricting access where required by law; the on-chain program is permissionless.

If you're unsure whether you're allowed to trade perps where you live, **don't trade until you've checked**. The protocol cannot return funds based on jurisdiction-after-the-fact.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.fermi.trade/19-risk-warning.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.