FRAGMENTED · DOCS

← TERMINAL

WHAT THIS TOOL MEASURES

Compare the estimated taker fee and slippage for a market order across 15 enabled orderbook venues. Each venue is measured against its own mid price. The terminal compares individual venues, then estimates a cheaper allocation across their available books.

Current scope is orderbooks only. RFQ venues and individual RFQ pairs from mixed adapters are excluded from the selector, comparison results and smart order routing (SOR). Carbon remains available for pairs with a supported hedge book; its RFQ stock pairs are excluded.

One basis point (bp) is 0.01% of notional. A 4.5bp cost on a $250,000 order is $112.50.

ENTERING A TRADE

Tap the trade field, enter a ticker, USD size and direction, then tap GO or the keyboard’s return key. Example buttons run a comparison immediately. On desktop, typing outside a control also fills the command buffer; Enter submits it.

BTC 250K LONG       $250,000 BTC buy
ETH 1M SHORT        $1,000,000 ETH sell
SOL 50000           side defaults to LONG
NVDA 100K LONG      stocks work where an enabled orderbook lists them

SIZE: USD notional, plain number or K/M suffix
SIDE: LONG / L / BUY or SHORT / S / SELL

Open EDIT beside VENUES to change the enabled venues on a phone. Tap a result card to reveal its price, fee, slippage, funding and book details. On desktop, select a table row or use the up/down arrows. The trade link opens the venue; this tool does not place orders.

ENABLED VENUES

Hyperliquid, Lighter, Paradex, Extended, Aster, dYdX, Pacifica, edgeX, Hibachi, ApeX Omni, GRVT, Ethereal, Velocity, Bluefin and Carbon. Asset availability varies by venue. Crypto and TradFi tickers are matched against the available orderbook markets.

WALKING THE ORDERBOOK

A long consumes asks from lowest to highest price; a short consumes bids from highest to lowest. At each level, the simulation takes as much of the remaining USD notional as the level can supply. The resulting quantity determines the volume-weighted average fill price (VWAP).

For level ii, let pip_i be its price,qiq_i its base quantity and RiR_i the remaining USD notional before that level. The amount filled is uiu_i:

ui=min⁡(Ri, piqi)Ri+1=Ri−uiQ=∑iuipiPVWAP=∑iuiQ\begin{aligned} u_i &= \min(R_i,\,p_iq_i) \\ R_{i+1} &= R_i-u_i \\ Q &= \sum_i \frac{u_i}{p_i} \\ P_{\mathrm{VWAP}} &= \frac{\sum_i u_i}{Q} \end{aligned}

Using mid price mm, adverse slippage as a fraction is:

slong=PVWAP−mmsshort=m−PVWAPm\begin{aligned} s_{\mathrm{long}} &= \frac{P_{\mathrm{VWAP}}-m}{m} \\ s_{\mathrm{short}} &= \frac{m-P_{\mathrm{VWAP}}}{m} \end{aligned}

With taker fee ff expressed as a fraction and notionalxx, execution cost is:

sbp=104scbp=104(f+s)C(x)=x(f+s(x))\begin{aligned} s_{\mathrm{bp}} &= 10^4 s \\ c_{\mathrm{bp}} &= 10^4(f+s) \\ C(x) &= x\bigl(f+s(x)\bigr) \end{aligned}

The walk stops at levels more than 10% away from the best price. If the remaining usable ladder cannot fill the order, the row reports INSUF DEPTH. Displayed depth sums the fetched side of the book; the executable amount can be smaller because of this price-band guard.

Carbon uses a Binance hedge-book proxy plus its configured per-market hedger fee, capped by the available notional limit. This is a model of hedge execution rather than a native Carbon limit orderbook or a firm execution guarantee.

FEES AND FUNDING

Costs use base-tier taker fees, overridden by live per-market fees where the adapter provides them. Account-specific discounts, token rebates and maker pricing are excluded. Spread is shown separately for context; slippage versus mid already includes crossing the spread, so it is not added a second time.

Funding is normalised to percent per hour. The annualised figure is hourly funding × 24 × 365, without compounding. Funding, borrowing and holding costs are excluded from execution cost.

SMART ORDER ROUTING — THE OPTIMISATION PROBLEM

The mathematical formulation is a constrained nonlinear resource allocation problem. The decision variable xvx_v is the USD notional assigned to venue vv. Its cost function Cv(xv)C_v(x_v) includes the taker fee and the size-dependent cost of walking that venue’s book. The objective is separable: each venue contributes its own cost.

minimise⁡x∑v∈VCv(xv)subject to∑v∈Vxv=S,0≤xv≤Dv∀v∈V.\begin{aligned} \underset{\mathbf{x}}{\operatorname{minimise}}\quad & \sum_{v\in\mathcal{V}} C_v(x_v) \\ \text{subject to}\quad & \sum_{v\in\mathcal{V}} x_v=S, \\ & 0\le x_v\le D_v \quad \forall v\in\mathcal{V}. \end{aligned}

SS is the requested USD trade size, DvD_v is usable book capacity, and V\mathcal{V} is the set of eligible venues. Fee fvf_v and slippage sv(xv)s_v(x_v) are fractions:

Cv(xv)=xv(fv+sv(xv))C_v(x_v)=x_v\bigl(f_v+s_v(x_v)\bigr)

When the cost curves are convex, this is more specifically a separable convex resource allocation problem. At a differentiable continuous optimum, venues with allocations strictly between zero and capacity have equal marginal costs; venues at their bounds obey the corresponding inequalities (the KKT conditions).

Cv′(xv∗)=λwhen 0<xv∗<DvC_v'(x_v^*)=\lambda \quad\text{when }0<x_v^*<D_v

Here λ\lambda is the common marginal cost at the optimum.

The implementation uses a greedy marginal-cost heuristicwith 200 equal USD slices. It does not call a general nonlinear or quadratic-programming solver. Each candidate slice is evaluated by recomputing the venue’s whole allocated order cost.

δ=S200ΔCv=Cv(xv+δ)−Cv(xv)\begin{aligned} \delta &= \frac{S}{200} \\ \Delta C_v &= C_v(x_v+\delta)-C_v(x_v) \end{aligned}
F(x)={v:xv+δ≤Dv}v∗∈arg⁡min⁡v∈F(x)  ΔCvxv∗←xv∗+δ\begin{aligned} \mathcal{F}(\mathbf{x}) &= \{v:x_v+\delta\le D_v\} \\ v^* &\in \underset{v\in\mathcal{F}(\mathbf{x})}{\arg\min}\;\Delta C_v \\ x_{v^*} &\leftarrow x_{v^*}+\delta \end{aligned}

Start every allocation at zero and repeat this step 200 times. The setF(x)\mathcal{F}(\mathbf{x}) contains venues that can accept another complete slice. If it is empty before the trade is filled, return no split.

For a $100,000 trade, each slice is $500. If the next slice adds $0.20 on venue A and $0.35 on venue B, A receives it; both candidates are then evaluated again for the next slice.

Greedy allocation is optimal on this discrete grid when each venue’s incremental costs are nondecreasing. That condition is not guaranteed for every quote or for the current USD-normalised VWAP cost model, so the displayed split is an estimate, not a certified global optimum. Finite slices can miss a better continuous allocation and can leave small residual capacities unused. Enough combined nominal depth does not always imply a feasible slice allocation.

The panel reports the split’s blended cost, weighted fee and slippage. Savings compares that result with the cheapest eligible venue that can fill the whole trade. If no single venue can fill it, there is no comparable savings figure. A venue unable to fill the whole order may still contribute a smaller allocation.

For background on convex formulations and optimality conditions, see Boyd and Vandenberghe’s Convex Optimization.

DATA QUALITY AND ROW STATES

OK             full order fits the returned usable book
INSUF DEPTH    usable book cannot fill the whole order
STALE PX       mid differs by >1% from the cross-venue median
NOT LISTED     no supported market / book for this ticker
CLOSED         market closed
TIMEOUT        venue exceeded its response deadline
ERR            another venue/API failure

Venue requests run in parallel, each with a nine-second timeout. With at least three usable mid prices, a venue more than 1% from the median is marked stale and excluded from SOR. Failed quotes are also excluded. The last comparison refreshes approximately every 15 seconds while the page remains mounted, subject to browser timer throttling.

Ticker aliases match each venue’s naming conventions. For example, NVDA can be xyz:NVDA, NVDA_24_5-USD or NVDAUSDT; XAU can map to GOLD or XAUT. A matching name does not guarantee an enabled venue offers an executable orderbook for the asset.

LIMITATIONS

Snapshots are fetched at different times and books can move before execution. Costs are measured against each venue’s own mid, not a shared execution benchmark. The suggested split ignores transfer costs, gas, collateral placement, latency, minimum order sizes and account-specific limits. It is an analytical estimate, not an executable quote or financial advice.

← BACK TO TERMINAL