Bitbase 2.0 mark price under the hood: what a 15-second half-life does to a 10-second BTC wick (advanced)

Advanced tier. Live Bitbase 2.0 numbers, 4:00 PM ET, Nov 3, 2026: BTC_USDT: last 77,836.2, mark 77,836.3, index 77,847.30, best bid 77,836.2, best ask 77,836.3, funding 0.02% per 8h with the next collection at 7:00 PM ET. XRP_USDT: last 1.2799, mark 1.2797, index 1.27912, best bid 1.2798, best ask 1.2799.

What the mark is made of

Bitbase defines it as: mark = current index + moving average basis. Basis = (contract best bid + contract best ask) / 2 - index at the same moment. The basis is sampled every second, the last 30 one-second samples are averaged with newer ones weighted more heavily, and the mark is recalculated every second. Major pairs default to a 15 second half-life; other and long-tail pairs default to 30 seconds. Half-life is not the window length. Every pair uses the same 30 seconds; half-life only sets how fast older samples lose weight.

Reading the 4:00 PM numbers

BTC instant basis = 77,836.25 - 77,847.30 = -11.05. BTC mark minus index = 77,836.3 - 77,847.30 = -11.00. The averaged basis and the instant basis almost match, so the BTC book had been holding a steady discount of roughly 0.014% to index.

XRP instant basis = 1.27985 - 1.27912 = +0.00073. XRP mark minus index = 1.2797 - 1.27912 = +0.00058. Here the book sat at a premium, and mark (1.2797) was below last (1.2799). Same minute, opposite sign: BTC at a discount, XRP at a premium.

The weights, worked out

Each sample's weight = 0.5 ^ (age in seconds / half-life). Summing 30 samples:

15 second half-life: newest sample about 6.0% of total weight, newest 5 about 27.5%, newest 10 about 49.3%, newest 15 about 66.7%. 30 second half-life: newest sample about 4.6%, newest 5 about 21.8%, newest 10 about 41.3%, newest 15 about 58.6%.

Worked example: a 10 second dislocation (illustrative)

Suppose the BTC book midpoint drops 300 below index for exactly 10 seconds, then snaps back. Index does not move. Last prints 300 lower.

With a 15 second half-life, at the end of the 10 seconds the averaged basis has absorbed 49.3% of the drop: mark sits about 148 below where it started, not 300. Ten seconds after the snap-back those samples are older: they carry 31.1% of the weight, so mark is still about 93 low. Twenty seconds after: 19.6%, about 59 low. After 30 seconds the spike leaves the window and mark is back to index plus the normal basis.

With a 30 second half-life the same spike pulls mark about 124 low, then about 98, then about 78, then it is gone. Less immediate impact, longer tail.

Two lessons. First, a short wick hits mark at roughly half strength or less, which is why last can touch a level your liquidation check never sees. Second, Bitbase says plainly that in rapid moves mark can temporarily lag last. A real move shows up in mark within seconds, so the protection is against brief spikes, not against trends.

What this means for your positions

Liquidation on Bitbase is triggered by mark, not last. When it fires, the process runs through order cancellation, netting of hedged positions, partial liquidation down to risk-limit tier 1, then the insurance fund and auto-deleveraging if needed. The estimated liquidation price, the liquidation trigger price (the mark recorded when the process starts), the bankruptcy price and the final execution average are four different numbers.

Your unrealized PnL can be shown on mark or last. Pick mark if you want the display to match the risk engine.

Your TP/SL has its own trigger type. If a stop uses last and mark reaches the liquidation condition first, liquidation takes priority.

Common mistakes

• Expecting mark to equal last. Nothing in the formula uses the last trade directly.
• Assuming half-life means only 15 seconds of data. It is always 30 samples.
• Treating the estimated liquidation price as a guaranteed trigger, especially in cross margin.
• Stacking a last-price stop right on top of the liquidation line.

FAQ

Q: Can mark be below both last and index? A: Yes, with a sustained book discount while last stays above mark.

Q: Which pairs use 15 seconds? A: Bitbase says majors default to 15 and others to 30, based on depth, spread, volatility and data stability.

Bottom line: the Bitbase mark is index plus a 30 second memory of the book, so judge risk on that memory, not on the latest print.