# EVEDEX vs Lighter: A Published Fee Against 300 Milliseconds

Updated: 24 September 2026 / Tested: 22 September 2026 / Venue facts read 18 September 2026

EVEDEX vs Lighter now separates by half a point, 29.1 against
28.6 of 40. One venue wrote the four strategies it runs; the other charges nothing
behind 300 ms and documents the order controls, the rate limit and a way out the operator cannot
block. Leverage shortens the distance to liquidation on both.

Weights: What the venue automates for you 24 · Programmatic access 14 · Measured liquidity 22 · Round-trip cost 16 · Security and record 12 · Custody and settlement 8 · Market coverage 3 · Access 1. One formula for the whole site, fixed on 23 September 2026: https://ellsworthvane.com/rating-method

## EVEDEX vs Lighter, criterion by criterion

| Criterion | EVEDEX | Lighter |
|---|---|---|
| Score of 40 | 29.1 | 28.6 |
| What the venue automates itself | Four strategies of its own on the BTC, ETH, SOL and XRP perpetuals, parameterised and switched on by the trader, with a capacity cap in USDT against each | A pooled vault a whitelisted operator trades; the strategy is not published or set by the depositor |
| What the order interface documents | A rate limit as a number, 30 heavy requests per 60 seconds per account. Of post-only, reduce-only and a time-weighted order, the prose documents none | A weighted rate limit per tier, plus post-only, reduce-only and a time-weighted order type |
| Fee at the entry tier | 0.015% maker and 0.045% taker, fixed, lowered only by cashback | Nothing on the default Standard account; 0.004% maker and 0.028% taker on Premium |
| Highest multiple on BTC | 200x to $50,000 of notional, stepping down to 100x at $200,000 | 50x, at 2% initial margin, with no notional steps published |
| What reaches the chain | Position data written to Arbitrum once enough changes have accumulated | Every batch of exchange operations proven on Ethereum |
| Exit if the operator stops answering | No forced-withdrawal path in the documentation we read | A priority queue on Ethereum, then a desert mode in which users leave with proofs |
| Audit reports on file | Two CertiK audits of EVEDEX smart contracts, the latest delivered 28 July 2025 | Nine reports from zkSecurity, Nethermind and Zellic, August 2025 to June 2026 |

Venue documentation read 18 September 2026 in each venue's own documentation and terms.

## Marks by criterion

Each criterion is marked out of ten before the weights above are applied.

| Criterion | EVEDEX | Lighter |
|---|---|---|
| What the venue automates for you | 10.0 | 3.0 |
| Programmatic access | 6.0 | 10.0 |
| Measured liquidity | 6.6 | 6.5 |
| Round-trip cost | 6.5 | 10.0 |
| Security and record | 7.0 | 7.0 |
| Custody and settlement | 5.0 | 10.0 |
| Market coverage | 7.8 | 9.2 |
| Access | 8.0 | 8.0 |

## EVEDEX — for a strategy that cannot spend a third of a second

Score 29.1 of 40.

EVEDEX is a hybrid perpetual futures exchange: matching off-chain inside a central order book,
settlement on-chain on Arbitrum, position data reaching the chain in batches (the EVEDEX
documentation we read on 18 September 2026). The rate is 0.015% maker / 0.045% taker, cashback returning
up to 35% of fees already paid, and its BTC book held a median $19.0 million within 10
basis points of mid.

- Works: Four strategies the venue wrote itself, parameterised rather than coded, with a published capacity cap in USDT against each; Median BTC depth of $19.0 million, the deeper of the two books, and shares, commodities, currencies and two pre-IPO markets on one cross-margin balance.
- Falls short: Reduce-only is documented nowhere, and post-only sits in the interface specification as a bare field with no values, no default and no stated behaviour if the order would cross; 52 perpetual markets against 214, and the four strategies reach four of them; No forced-withdrawal path in the documentation we read, and no bug bounty on the CertiK Skynet listing, read 18 September 2026.
- Not for: a strategy that needs an exit it can run while the operator is silent.

## [Lighter](https://lighter.xyz) — best for a strategy that can pay for its fills in time

Score 28.6 of 40.

The default Standard account charges 0% maker / 0% taker with 300 ms added to a taker order;
Premium buys the delay back at 0.004% maker and 0.028% taker
([Lighter docs](https://docs.lighter.xyz/trading/trading-fees.md), checked 18 September 2026). An operator-run sequencer orders the
flow and every batch is proven on Ethereum, where a priority queue and a desert mode let a trader
leave without it. Its book held a median $17.4 million at the tighter spread of
0.012 bps.

- Works: Nothing to pay on either side of the trade at the default tier; Post-only, reduce-only and a time-weighted order type all documented as named constants, beside a rate limit published per tier; A documented exit that does not need the sequencer, audited by Nethermind in September 2025.
- Falls short: No strategy, grid or backtester of its own: the venue-run automation is a pool a whitelisted operator trades, with nothing published about how; 300 ms added to every taker order at the free tier, against 140 ms on the paid one; A 4.5-hour outage on 10 and 11 October 2025, and a $100,000 order would have paid 0.67 bps against the book we read.
- Not for: a strategy whose edge is gone inside the next quarter-second.

## What we read off the books

Both BTC perpetual books were polled at ten-minute spacing, the last reading dated 22 September 2026; medians of all successful snapshots.

| Venue | Round trip on $10,000 | Median spread | Depth, 10 bps | Impact, $100,000 order | Snapshots |
|---|---|---|---|---|---|
| EVEDEX | $9.00 | 4.12 bps | $19.0 million | 2.06 bps | 117 |
| Lighter | $0.00 | 0.01 bps | $17.4 million | 0.67 bps | 57 |

## Read-offs

1. One criterion carries the gap and four carry almost all of it back. Automation is worth 24 of the hundred points, ten marks against three, or 6.72 of the 40; cost, programmatic access, settlement and coverage hand back 6.25, and half a point survives.
2. Neither of these two tops the site board. Aster does, at 30.0 of 40, with this pair second and third behind it.
3. On the books we read, the cheaper venue was the thinner one: a median $17.4 million of resting depth against $19.0 million within 10 basis points of mid. Two crossings of $10,000 cost $9.00 and $0.00, or $5.85 against $5.60 once cashback and the paid tier are counted.

## One venue wrote the strategy, the other rented out a pool

This is the criterion added on 23 September 2026, and it is worth 24 of the hundred points. EVEDEX
runs four strategies of its own, one each on the BTC, ETH, SOL and XRP perpetuals, all of the same
mean-reversion family: the trader chooses a strategy and a coin, sets the parameters and activates
it, and the algorithm trades unattended from there. Every card publishes a trade-frequency label
and a cap in USDT on the capital it will take, with the amount already allocated against it, and
the seven-day backtest we recorded is on the XRP card (EVEDEX documentation and product page,
read 23 September 2026). The cards print return figures too. None is reproduced here, for any venue.

Lighter publishes no trading bot, no grid and no backtester. What it has is Public Pools, where
participants combine funds under a designated operator who trades for a fee
([Lighter docs](https://docs.lighter.xyz/trading/public-pools), checked 18 September 2026). Ten marks against three, the widest
gap between these two columns. What the other seven dials do with that gap is below.

Everything else runs the other way. Cost returns 2.24 of the 40 to Lighter, programmatic access
another 2.24, and settlement — five marks against ten, on a criterion now worth 8 of the hundred
— a further 1.6. That is how a seven-mark lead on one dial ends up as half a point on the board.

## What 300 milliseconds costs, and what it buys

The published rate is the easy half. Two crossings of $10,000 cost $9.00 at the
EVEDEX taker rate and nothing on the default Standard account at Lighter
([Lighter docs](https://docs.lighter.xyz/trading/trading-fees.md), checked 18 September 2026). Cashback narrows it: a ceiling of 35%
of a trader's own fees puts the effective taker side at 0.02925%, or $5.85, against $5.60 on the
Lighter tier without the delay. Nine dollars of daylight becomes twenty-five cents.

The free tier charges in time instead: 300 ms on a taker order and 200 ms on a maker order,
against 140 ms on Premium, where the EVEDEX documentation we read on 18 September 2026 names no
latency tier at all. A strategy rebalancing hourly will take that trade; one quoting inside the
spread cannot.

## A published rate limit, and two controls that are not described

The other new criterion asks what a strategy placing its own orders is told, and it is worth 14 of
the hundred: four marks for a rate limit published as a number, two for each of post-only,
reduce-only and a time-weighted order the order interface documents.

EVEDEX publishes the number: 30 heavy requests per 60 seconds per account, heavy meaning an order
created, cancelled or replaced, a position closed, the same three for a take-profit or stop-loss,
and a withdrawal request. Beside it sit up to 500 active ordinary orders, up to 500 active
take-profit and stop-loss orders and a minimum order size of 5 USD. Four marks.

The order controls are the gap. Reduce-only appears in no page we read — not the documentation,
not the Help Center, not the interface specification the venue publishes — so a position is
reduced by closing it in whole or in part, or by a take-profit or stop-loss. Post-only is the
awkward case: the field is in that specification, on three order endpoints, declared as a bare
string with no accepted values, no default and nothing said about an order that would cross. It
cannot be called a documented feature and it cannot be called absent. Present and undescribed is
not something a strategy can build on, and we record it as it reads rather than ask anyone to
settle it.

Lighter documents all three as named constants: post-only as a time-in-force setting, reduce-only
as an execution option, a time-weighted order as a type of its own
([Lighter API docs](https://apidocs.lighter.xyz/docs/trading), checked 18 September 2026). Ten marks against six, worth
2.24 of the 40 — more than the half point between these totals, and the one thing here most likely
to move the order if it moved.

## How the two columns were filled

One panel, eight dials, 2 venues on it here; weights set on 23 September 2026, before any
mark existed ([how we rate](/rating-method)). Venue facts are each venue's own wording, read
18 September 2026; the two automation dials were read on 23 September 2026. Depth and spread are
polled: each BTC book sampled at ten-minute spacing, the last reading dated 22 September 2026, nothing sent.

Seven of the eight dials are documents rather than readings: an audit records a review, an exit past
the operator is a claim on a page, a shipped strategy is marked on what the venue publishes. Entry-tier
rates ignore volume tiers, rebates and cashback. The measured dial: one instrument, one machine, one
day. Open interest and volume are the venues' own, via CoinMarketCap.

## Quotes the numbers came from

> "An off-chain order book (Matcher) system manages buy and sell orders outside the blockchain. This approach follows the same principle as centralized exchanges, enabling faster order processing and matching." — EVEDEX documentation, off-chain order book, 18 September 2026.

> "Taker latency: 300ms" — Lighter documentation, Trading fees, 18 September 2026. https://docs.lighter.xyz/trading/trading-fees

> "Users can submit critical exit operations—such as withdrawals, public pool exits, or reduce-only IOC orders—on-chain, ensuring that the Sequencer must process them within a predefined timeframe." — Lighter documentation, Technical architecture, 18 September 2026. https://docs.lighter.xyz/about-lighter/technical-architecture-lighter-core

## FAQ

### Is Lighter really free to trade?

On the Standard account, yes: 0% maker / 0% taker across perpetual and spot markets, in the
documentation read on 18 September 2026. The charge is latency, 300 ms added to a taker order,
and Premium removes 160 ms of it for 0.004% maker and 0.028% taker.

### What is the difference between a Standard and a Premium account on Lighter?

Price and speed. Standard is the default and pays nothing, with 300 ms added to a taker order and
200 ms to a maker order. Premium pays 0.004% maker and 0.028% taker for 140 ms. A third tier,
Plus, charges half a basis point both ways for higher throughput.

### Which is cheaper to trade, EVEDEX or Lighter?

Lighter, on the published rate: nothing against $9.00 for two crossings of
$10,000. Cashback of up to 35% of a trader's own fees takes the second to $5.85 and the paid tier
costs $5.60. Impact ran the other way on a million, 2.74 bps against
1.47 bps.

### How many markets does Lighter list?

214 active perpetual markets on its own public endpoint on
18 September 2026, plus ten spot markets. The list runs past crypto into shares, currencies,
commodities, indices, pre-IPO names and a ten-year government bond, which is seven classes of
asset on one balance.

### What happens to an open position if an exchange sequencer stops?

The position stays open and the margin stays where it is. What differs is the way out. Lighter
documents a priority queue on Ethereum that forces the operator to process a withdrawal, and a
desert mode if it does not. EVEDEX documents no such path in the pages we read.

### Is EVEDEX a decentralized exchange?

It is a hybrid perpetual futures exchange. A central order book matches off-chain; settlement is
on-chain on Arbitrum, which receives position data only after enough changes have built up. That
is the documented design, read 18 September 2026, and it is neither a pool-priced venue nor a
chain of its own.

### Does EVEDEX document post-only and reduce-only orders?

Reduce-only, no: it appears in no page we read. Post-only sits in the interface specification the
venue publishes, on three order endpoints, as a bare string with no accepted values, no default
and nothing stated about an order that would cross. Present and undescribed is not something a
strategy can rely on.

### What is desert mode on a zero-knowledge exchange?

An emergency state written into the contracts. If the operator stops processing exits queued on
Ethereum, the contract freezes and traders withdraw by proving their own balances, with no part
for the operator. Lighter documents it, and Nethermind audited that system in September 2025.

### How much cashback can a trader get on EVEDEX?

Up to 35% of the trader's own trading fees at the ceiling, returned on fees already paid rather
than deducted in advance. At that ceiling the taker side works out at 0.02925% and the maker side
at 0.00975%, which is $5.85 rather than $9.00 on two crossings of $10,000.

### Which has more open interest, EVEDEX or Lighter?

EVEDEX, on the CoinMarketCap ranking captured on 22 September 2026: $704.9 million of open
interest against $585.4 million. Volume ran the other way, $700.5 million against
$2.0 billion over the same 24 hours. Both are venue self-reports, and neither is
scored here.

## The desk behind this comparison

One published formula, every figure dated and sourced. Corrections: editorial@ellsworthvane.com.

Signal Grid research, 24 September 2026

These panels are paid placement.
