#k线 Community Insights & Market Sentiment

Shift the daily boundary by 12 hours, and LINK’s up days drop from 17 to 13

Not a single candlestick was added or removed. The same 10 assets, unchanged. The spans are all 30 days. Just move the daily cut-off from 00:00 UTC to 12:00 UTC, and LINK’s up days drop from 17 to 13, while SOL’s fall from 16 to 12. BNB goes the other way, rising from 16 days to 18.

All three panels in the chart show this comparison. Gray is Binance’s native daily candles, cut at 00:00 UTC. Gold is the same batch of 1-hour candles, regrouped into 24-hour periods with the boundary moved to 12:00 UTC. Both use 30 complete buckets and span 30 days; their start times differ by 12 hours. The data comes from Binance USD-M perpetuals’ /fapi/v1/klines endpoint: 800 1-hour candles for each of 10 assets, fetched on October 24 at 14:22 (UTC+8).

The panel on the left makes the point most clearly. The number of up days looks like a property of the market. In fact, it’s a property of the buckets. What time a day starts determines whether a big green candle counts as today or tomorrow, and whether it gets split in two across two buckets. LINK lost 4 days, SOL lost 4 days, and the market itself didn’t change at all. Only the way it was divided changed.

The middle panel is even more striking. XRP’s maximum daily gain is 2.45% with the 00:00 cut-off and 8.50% with the 12:00 cut-off—a 3.47-fold difference. DOT goes the other way, dropping from 13.46% to 7.57%. Same trades, same price sequence. The maximum daily gain can really differ by more than three times, just because that green candle fits completely into one bucket under the new boundary—or gets split in two.

The panel on the right is the counterexample. Average daily range—that is, the day’s high minus its low, divided by the open—barely moves. BTC: 2.48% versus 2.58%. ETH: 3.47% versus 3.55%. LINK: 5.88% versus 5.82%. Even AVAX, with the biggest difference among the 10 assets, is off by just 0.37 percentage points. Range is calculated from the extremes within the interval, so which day those extremes fall on doesn’t matter. Up days and maximum daily gain are calculated from each bucket’s open and close, so move the bucket and they all change. Put the two sets of figures side by side, and it’s immediately clear which are sensitive and which aren’t.

Cumulative returns are affected too. DOT falls 1.40% over 30 days with the 00:00 cut-off, but rises 1.82% with the 12:00 cut-off—the direction flips completely. AVAX goes from -11.75% to -17.23%, a further drop of 5.48 percentage points. Nine of the 10 assets show larger losses. DOT, though, turns positive.

A lot of people think daily candles emerge naturally from the market, and the exchange just records them. What Binance actually does is package them up. It bundles the 24 hours from 00:00 to 23:59 UTC into one candle. That boundary is entirely human-defined. The same goes for using 12:00 UTC. Both ways of dividing the day are valid; nobody can say the other one is wrong. When you see 17 up days, that assumes the day starts at midnight.

There’s one thing I can’t decide for you here. The two windows start 12 hours apart. Strictly speaking, they don’t use the same set of candles. I made that trade-off to get 30 complete buckets. Want to lock the interval and only move the boundary? The 12:00 cut-off then gives you two extra half-day buckets, one at each end. The day counts won’t match. That would make the comparison even less fair. So this chart can show that the boundary affects counts and extremes. It absolutely can’t tell you how many days LINK actually went up. Also, these 10 assets moved up and down together over this period, so 300 daily observations don’t count as 300 independent samples.

You can verify this with a stricter version. Fix the interval from September 24 at 00:00 to October 24 at 00:00 (UTC), keeping all 30 days, and only then change the boundary. XRP’s maximum daily gain still jumps from 2.45% to 8.50%. DOT’s maximum daily loss drops from -6.95% to -10.18%. The interval doesn’t move at all. The numbers still change.

So what is stable? Range. It only looks at the high and low within the interval, so it doesn’t depend on how the buckets are cut. Cross-asset comparisons are stable too, as long as all assets use the same cut-off. What’s truly unstable are metrics built up from day-by-day comparisons: up days, consecutive up or down days, maximum daily gain or loss, and daily win rate. These are precisely the things most vulnerable to a change in boundary. Switch exchanges, and they change. Switch time zones, and they change again.

There’s one thing I plan to track next. Calculate the number of up days for the same set of assets using both the 00:00 and 12:00 cut-offs, and see whether the difference has a consistent sign. If an asset consistently differs by more than 3 days between the two cut-offs, that suggests its big moves are mostly concentrated between 00:00 and 12:00 UTC. That’s more useful than looking at how many points it gained on any one day.

$LINK $XRP $DOT
#K线 #数据口径 #Trading day

Source:
Binance USD-M perpetual candlesticks https://fapi.binance.com/fapi/v1/klines
Binance spot daily candles (for reference cut-off) https://api.binance.com/api/v3/klines

k线

Trending Topics