The Gap

What the book promised vs what you actually filled.

Each cell shows a venue's realized fill cost in bps, with the gap in parentheses. Realized is what a clip of that size actually filled at vs the mid; advertised is what walking the order book for that size implied; the gap is realized minus advertised, so a positive gap means the venue filled worse than its book promised. The gap is floored at zero: a venue can't truly fill better than the book that existed before the trade, so any sub-zero value (book-age noise) shows as 0. This is execution quality only. Fees and funding are added in "Cost to run a position" below. Cheapest realized in bold. ^* = sparse (fewer than 5 clips in the window): shown for transparency, but not ranked or crowned cheapest until it has more data. fl = Pacifica book feed-limited (realized shown, not gap).

cell format: realized (gap vs advertised)

Hover any cell or chart point for its clip count (n).

Cost to run a position

Your all-in cost, sized to your own volume and hold.

Your all-in cost: slippage + taker fee (both legs) + funding over the hold. Computed live in your browser from your inputs.

Funding & carry

Who pays you to hold, and how dependable the rate is.

Funding is the recurring payment between longs and shorts for holding a perp. A positive APR means longs pay shorts; negative means shorts pay longs (you're paid to be long). The rate resets hourly and drifts, so a trailing average alone can mislead: the stability columns show how often the rate changed sign (flip rate) and how directional it has been, so you can tell a dependable carry from a whippy one.

Track record

How execution holds up over time, not just today.

How the gap and cost evolve over time, which a static snapshot can't capture. History compounds.

By time of day (Asia / EU / US, 2-hour windows)

Behavior under stress

Does fill quality hold when the market breaks?

When the market gets liquidated, does fill quality hold? Cascades from the market-wide liquidation feed; execution during vs the prior-24h baseline.

Methodology & confidence

What's measured, how, and where the seams are.

A single trade usually fills in many small pieces; before measuring its cost we group those pieces back into the one order the trader actually sent. How precisely we can do that depends on what each venue publishes, shown in the table below. We also slice the data into 2-hour windows: a window is "clean" if no feed disconnect or data gap was detected during it, and windows with a gap are dropped from every number above, so a brief outage is never mistaken for a real reading. The last column shows each venue's share of clean windows, i.e. how reliably we captured its data.