Market-maker economics on perp venues: rebates, zero fees and flow toxicity
Post540 words
Decomposes MM PnL into spread capture, adverse selection, inventory risk, funding and fees. Compares maker rebates, zero-fee growth models and Paradex's flow segmentation via RPI, arguing liquidity is stable only if maker EV stays positive in volatility.
Key claims
Lighter standard accounts pay no maker or taker fees
Paradex argues takers can pay zero while makers pay if flow is segmented
Market-maker economics is basically one question with three moving parts:
How much do you earn per fill, how much do you lose to toxic flow, and how much risk do you carry while holding inventory?
Most perp venues tune that equation with fees/rebates.
For a market maker quoting bid/ask and getting filled:
๐๐๐ก ๐๐๐ฟ = ๐๐๐๐๐๐ ๐๐๐๐ก๐ข๐๐ โ ๐ด๐๐ฃ๐๐๐ ๐ ๐ ๐๐๐๐๐ก๐๐๐ โ ๐ผ๐๐ฃ๐๐๐ก๐๐๐ฆ ๐๐๐ ๐ โ ๐น๐ข๐๐๐๐๐/โ๐๐๐๐ ๐๐๐ ๐ก๐ + ๐
๐๐๐๐ก๐๐ โ ๐น๐๐๐
Where:
โค Spread capture (per fill) is roughly = (ask price โ bid price) / 2 ร filled size
โค Fees / rebates are typically applied on notional: notional ร fee rate
(If fee rate < 0, thatโs a rebate: the venue pays you to make liquidity)
โค Adverse selection is what happens when you get filled because someone knows more than you (you sell right before price jumps, or buy right before it dumps). Itโs why โrebate โ free money.โ
โค Inventory risk is PnL variance from holding a net position while you wait to flatten (especially during fast moves).
โค Funding/hedge costs matter in perps: funding flows between longs/shorts to keep perps near spot, and it can become a real carry component for a MM running inventory.
๐ฆ๐ผ ๐๐ต๐ ๐ฑ๐ผ ๐๐ผ๐บ๐ฒ ๐๐ฒ๐ป๐๐ฒ๐ ๐๐๐ฏ๐๐ถ๐ฑ๐ถ๐๐ฒ ๐บ๐ฎ๐ธ๐ฒ๐ฟ๐ (๐ฟ๐ฒ๐ฏ๐ฎ๐๐ฒ๐) ๐ฎ๐ป๐ฑ ๐ฐ๐ต๐ฎ๐ฟ๐ด๐ฒ ๐๐ฎ๐ธ๐ฒ๐ฟ๐?
This is the classic makerโtaker model: pay makers a rebate, charge takers a fee. The goal is to manufacture an order book by compensating makers for inventory risk and adverse selection. For liquidity stability it:
โค Usually produces tighter displayed spreads (more makers show up if the rebate offsets their risks).
โค But it can also attract โrebate-firstโ strategies: queue games, short-lived liquidity, and liquidity that disappears during volatility when adverse selection dominates.
โค In stress, stability depends on whether makers still have positive expected PnL when volatility spikes (rebates rarely cover toxic flow by themselves).
And why do some venues charge makers (or go โzero feeโ) instead?
โ โ๐ญ๐ฒ๐ฟ๐ผ-๐ณ๐ฒ๐ฒโ ๐ฎ๐ ๐ด๐ฟ๐ผ๐๐๐ต ๐น๐ฒ๐๐ฒ๐ฟ (๐ฏ๐๐ ๐ฟ๐ฒ๐๐ฒ๐ป๐๐ฒ ๐บ๐๐๐ ๐ฐ๐ผ๐บ๐ฒ ๐ณ๐ฟ๐ผ๐บ ๐๐ผ๐บ๐ฒ๐๐ต๐ฒ๐ฟ๐ฒ)
Some venues set maker = 0 and taker = 0 (at least for standard accounts), which removes the fee friction entirely.
Example: @Lighter_xyz states standard accounts currently pay no maker or taker fees.
That can boost activity, but it shifts the question to: how do you sustain incentives and operations without explicit fee revenue?
โ ๐๐น๐ถ๐ฝ ๐๐ต๐ฒ ๐ณ๐ฒ๐ฒ ๐บ๐ผ๐ฑ๐ฒ๐น ๐๐ต๐ฒ๐ป ๐๐ผ๐ ๐ฐ๐ฎ๐ป ๐ฐ๐ผ๐ป๐๐ฟ๐ผ๐น ๐๐ผ๐
๐ถ๐ฐ๐ถ๐๐
@Paradex argues that fee assignment should reflect flow toxicity, and discusses models where takers pay zero while makers pay (the โflipโ).
This only works if makers get compensated in a different way - e.g., by giving them access to less toxic flow via segmentation.
Paradexโs RPI (Retail Price Improvement) is an example of flow segmentation: itโs designed to give retail better execution while protecting makers by controlling visibility/interaction rules.
๐๐ผ๐ฟ ๐น๐ถ๐พ๐๐ถ๐ฑ๐ถ๐๐ ๐๐๐ฎ๐ฏ๐ถ๐น๐ถ๐๐ ๐ถ๐ ๐ถ๐บ๐ฝ๐น๐ถ๐ฒ๐:
โค If makers are charged, they will usually widen spreads unless something offsets that cost (better flow, protected lanes, or other incentives).
โค If flow segmentation works, you can get stable liquidity without paying rebates, because makers arenโt constantly getting picked off.
Liquidity stability comes down to whether makers can keep expected PnL > 0 when volatility rises. If the model only works during calm periods or incentive campaigns, liquidity will be fragile.
1 image on X