How mark price actually works on Bitbase when BTC drops 4.82% in a day

The mark price rules come from Bitbase's own explanation of its futures mark price.

Step 1: read the board

• BTC_USDT 24h open 80,026.2, high 80,927.5, low 75,525.1. Last 76,164.0 at 3:56 PM ET, down 4.82% on the 24h.
• At that same second: mark 76,157.3, index 76,153.94. The mark sat 6.7 under the last trade and 3.36 above the index.
• Funding 0.02% per 8h (0.0002), next collection 7:00 PM ET. A positive rate means longs pay.
• Market-wide, about 1.09 billion USD of positions were liquidated over 24 hours, hitting about 211,000 traders, after a sell-off in US tech and AI stocks. ETH, SOL, XRP and HYPE fell 6%-7%.

Step 2: see which price matters

Your unrealized PnL and your liquidation risk on Bitbase are assessed on the mark price, not on the last trade. That is why a wick on last does not, by itself, close a position, and why a bounce on last does not, by itself, save one.

Step 3: the rule

The mark price does not rely on a single trade. It is built from the index price plus a smoothed basis taken from the futures order book. During rapid market moves it may temporarily lag behind the last price. That lag is by design: it keeps short-lived spikes and abnormal trades from setting your PnL and liquidation checks.

Step 4: the formula

basis = (best_bid + best_ask) / 2 - index (sampled every second) moving_average_basis = weighted average of the latest 30 one-second samples, newer samples weighted more mark = index + moving_average_basis (recalculated every second)

Half-life sets how fast old samples fade. Major pairs use 15 seconds; other and long-tail pairs use 30 seconds. Every pair uses the same 30-second window.

Step 5: apply it to Thursday

1. A one-tick print on last under your liquidation line only matters if the book's midpoint stays there long enough to drag the 30-second average.
2. With a 15-second half-life on BTC, a sample from 15 seconds ago carries half the weight of the newest one, so a sustained move shows up in the mark within seconds.
3. The reverse is also true. A bounce on last that lasts a few seconds lifts the mark only partly.
4. At 3:56 PM ET the mark was 3.36 above the index, so over the previous 30 seconds the book traded at a small premium to spot.
5. The lag is not a bug. It is the cost of smoothing. The same 30-second average that ignores a one-second spike also takes a few seconds to catch up with a real move, so in a cascade the mark can trail last on the way down and on the way back up.

Common mistakes

• Setting alerts on last when liquidation runs on mark.
• Expecting the mark to match the chart's lowest wick.
• Assuming the mark and index are the same number. They differ by the smoothed basis.
• Ignoring the next collection when holding a long at 0.02% per 8h.

FAQ

How often does the mark update? Every second.

Does a 15-second half-life mean only 15 seconds of data? No. Every pair uses the latest 30 seconds. Half-life only controls how fast the older samples lose weight inside that window.

Can the mark sit below the index? Yes. When the futures book trades at a sustained discount, the smoothed basis turns negative and the mark falls below the index.

Try it on your own position: on the next fast candle, write down last, mark and index at the same moment and see which one moves first, and which one your liquidation line is actually measured against.