Why do perpetual contracts, which never stop trading, consistently exhibit anomalies precisely every 15 minutes?

2026.09.07
ShareShare to X (Twitter)Share to WeChatShare to WeChat
Share to WeiboShare by Link
Share by Image
The crypto market has given birth to its own dedicated “opening bell.”

Translated by: Luffy, Foresight News

At 14:59:59 UTC, much like other electronic trading markets, prices on Bitcoin perpetual contracts and beyond were ticking constantly, with orders from global traders flowing in continuously. But the moment the clock struck 15:00:00, market activity surged sharply. Trading volume spiked, capital turned over rapidly, and price volatility widened noticeably over the next 10 seconds, without any breaking news driving the behavior.

This trading pulse arrives precisely at the 15th, 30th, and 45th minutes of every hour. Smaller-scale versions of this pattern also occur at five-minute marks and the start of each minute, while the burst of trading activity is strongest at the top of the hour. Although the crypto market operates as a perpetual exchange, trading software fragments it into countless micro-trading intervals.

South Korean policy researcher Chan Kim and Peter Reinhard Hansen of the University of North Carolina documented this phenomenon in an academic paper. The study sample consisted of tick-level trade data for six major perpetual contracts on Binance from January 1, 2021, to October 31, 2024. Underlying assets included Bitcoin, Ethereum, XRP, Solana, Dogecoin, and Cardano, spanning 1,400 full, uninterrupted trading days.

The instrument under study was the perpetual contract. Traders can use perpetual contracts to bet on asset price movements, amplifying their positions with leverage. Traditional futures have fixed expiration dates, whereas perpetual contracts can be held indefinitely as long as margin requirements are met. Long and short parties regularly swap funding payments to ensure contract prices track the spot index. When the perpetual price trades above the spot index, long traders pay funding fees to shorts; when below, shorts pay longs.

Perpetual contracts account for a massive share of global crypto trading, giving these short-duration price pulses a broader market impact. Prices on perpetuals guide cross-exchange arbitrage, hedging, and market-making activities, meaning that volatility originating in the futures market quickly transmits to the spot markets for Bitcoin and other assets.

The Crypto Market Gets Its Own “Opening Bell”

If one hour is plotted on a radial chart, the 15-minute trading pulse appears clearly. The researchers’ chart shows peaks at 0, 15, 30, and 45 minutes, with trading volume and price volatility distributed in a star shape. The vast majority of these surges are concentrated within the first 10 seconds of each cycle.

Aggregating data across the six contracts, trade counts in this 10-second window increased by 26% compared to normal periods, USD-denominated volume rose by 32%, and absolute price volatility expanded by 26%. Absolute return metrics measure bidirectional price movement; both upside and downside spikes become significantly more pronounced at the start of the 15-minute interval.

Polar coordinate charts illustrate the patterns of absolute returns and trading volumes for BTC, ETH, XRP, SOL, DOGE, and ADA perpetual contracts at different minute nodes within an hour.

This pattern holds true regardless of the underlying asset’s market capitalization. During the sample period, Bitcoin averaged 1.54 million daily trades with a contract volume of $14.58 billion, while Cardano averaged roughly 290,000 daily trades with a volume of only $544 million. Yet both assets traded in highly synchronized rhythm.

This cross-asset consistency is the study’s most significant finding: the trading pattern is driven by market-wide infrastructure and universal trading mechanisms, rather than being a unique characteristic of any single token.

Most trading platforms organize continuous price feeds into standardized candlestick charts—1-minute, 5-minute, 15-minute, etc. A single 15-minute candle aggregates the open, high, low, and close prices for that period. This visualization tool not only facilitates manual market analysis but also provides quantitative algorithms with standardized data units.

As each candle closes, various technical indicators are recalculated, and automated trading strategies refresh their signals based on the newly completed candle. Algorithms designed to break down large orders submit remaining limit/market orders at these time markers; market makers anticipate capital flows and adjust their quoted spreads; faster quantitative systems position themselves in advance.

Once enough programs synchronize to the same timeframes, what began as merely a data display format becomes an integral part of the market itself. At the end of an otherwise quiet 15-minute interval, trading converges into a centralized burst reminiscent of a traditional exchange opening bell. That real-world opening surge occurs because traders queue up after extended market closures. The crypto market replicates this frenzy through uniform candlestick cycles and default software settings, replaying it every 15 minutes around the clock without interruption.

Traces Left by Machine Trading

Binance’s trade history displays the asset, quantity, and price, but cannot distinguish whether an order originated from a human trader, a market-making firm, a liquidation engine, or another automated system. Kim and Hansen looked for indirect clues through order size.

Human traders tend to prefer round numbers and clean price levels—such as 0.1 BTC or roughly $10,000 per order—rather than calculating precise, fragmented decimals. Conversely, quantitative algorithms typically determine position sizes based on volatility, available liquidity, current exposure, or large-order slicing targets. To human eyes, the resulting trade amounts often appear completely arbitrary.

Researchers tracked the frequency of order values ending in zero. Within the first few seconds of a pulse event, the proportion of round-number orders dropped significantly. The statistical sample only filtered for sufficiently large trades to avoid biases introduced by minimum lot sizes, preventing small orders from being mistakenly classified as non-human.

The more critical the time node, the sharper the decline in round-number orders. Normal minute starts saw a slight dip, five-minute marks saw a wider drop, 15-minute marks fell further, and the divergence peaked at the top of the hour.

Taking Bitcoin orders meeting the double-zero criterion as an example: during regular minute starts, the proportion of round-number orders dep of the hour, this difference reached 0.20, making the hourly effect five times stronger than in normal periods

The chart shows that trading activity for all six crypto assets peaks within the first 10 seconds of each minute, with higher average volume at 15-minute intervals.

Standard de metric does not directly equate to the proportion of machine-driven trades. It simply proves that during sudden spikes in activity, the market stops habitually submitting round-number orders, a behavioral shift that corroborates increased automation participation

Order size alone still cannot pinpoint theed liquidations, and funding rate arbitrage trades also produce non-round figures. Therefore, the paper uses order characteristics solely as indirect evidence of quantitative trading activity

The authors conducted multiple control experiments to rule out other cyclical events causing these pulses. During the sample period, Binance settled funding rates at 00:00, 08:00, and 16:00 UTC. After excluding these three windows, the 15-minute pulse effect remained statistically significant. Even when filtering out all hourly observations, the 15-, 30-, and 45-minute patterns persisted. An independent analysis of Bybit exchange data yielded highly similar market dynamics.

These controls demonstrate that this is a widespread form of electronic, coordinated trading. While traders could technically customize any timeframe, exchange APIs, chart defaults, and standard technical indicators guided a massive influx of algorithmic logic to converge on identical temporal boundaries. The top-of-the-hour and quarter-hour marks, which attract the highest attention, also see peak capital concentration.

Predictable Price Signals Fall Short of Trading Fees

Having confirmed the periodicity of these pulses, the research team further tested whether pre-interval price data could predict directional movement during the opening 10 seconds.

A rolling prediction model ingested prior 15-minute return data alongside classic volume-price indicators, leveraging contemporaneously available market information for out-of-sample forecasting.

Backtesting across the six contracts showed the model achieved a 56.6% accuracy in predicting direction. The average out-of-sample R-squared was 3.4%, indicating the model explains only a tiny fraction of the volatility within that 10-second window. The area under the curve score sat at 0.60 (where 0.5 represents random guessing and 1.0 denotes perfect prediction).

In a highly noisy 10-second market, these limited metrics prove the pattern does hold reproducible signal value. However, translating this into a simple, stable profit strategy is impractical due to negligible predicted moves. Strictly following the model’s signals at every 15-minute mark would yield an average gross return of just 0.51 basis points (0.0051%) per trade before fees. On a $10,000 principal, that translates to roughly $0.51 in gross profit.

Throughout the observation period, Binance charged a base taker fee of 5 basis points and a maker fee of 2. A $10,000 taker order cost roughly $5 to enter, plus another fee to exit. The model’s average gross profit fell well short of even one-tenth of a single entry fee.

With margins this razor-thin, the dataset’s core value lies in highlighting the vast chasm between statistical predictability and the actual profitability accessible to retail traders. Market patterns may repeatedly pass rigorous statistical tests, but short-term volatility gains fail to cover basic transaction costs. This disparity explains why highly automated markets feature identifiable patterns yet resist easy arbitrage.

Market makers and large institutions can still leverage these findings to refine execution. Firms providing two-way liquiditywindow and reduce quote sizes when anticipating directional inflows. Institutions executing large sliced orders can schedule trades away from congested timeframes to minimize slippage caused by their own footprints

The first 10 seconds of the 15-minute window also contain signals for longer-term trends. If aggressive buying volume exceeds selling volume at the quarter-hour mark, this order imbalance tends to drive prices higher over the subsequent 4 to 12 hours. Conversely, seller dominance exerts downward pressure on medium-to-longer-term trends.

Order imbalance measures the net difference between aggressive buys and sells relative to total volume within the window, indicating which side—long or short—is exerting greater momentum.

At the 4-hour horizon, medium-to-long trends largely absorb the capital flow signals from preceding 15-minute nodes. Extending to 8-hour or 12-hour frames, traditional volume-price metrics gain stronger explanatory power. This market logic confirms that quantitative systems use the 15-minute cadence as a unified signal framework to digest long-held, cross-market trading data.

These long-term conclusions require careful interpretation. The 4-hour, 8-hour, and 12-hour return windows overlap, meaning large swing moves recur across multiple statistical buckets. While the authors mitigated bias using block bootstrap sampling tailored for non-independent data, aggregated trade logs still cannot differentiate whether orders reflected private information, reacted to shared public catalysts, or simply stemmed from market makers absorbing massive unilateral flows.

Stripping away complex statistical models, the underlying logic is straightforward. The crypto market abolished the closing bell to enable 24/7 continuous trading; yet APIs, candlestick intervals, and automated strategies have effectively reconstructed millions of micro-opening moments throughout the day.

Every 15 minutes, thousands of independently running trading programs converge on the exact same timestamp. In mere seconds, this market designed for seamless continuity resembles a massive crowd rushing through a single door.

Join TechFlow official community to stay tuned

Telegram:https://t.me/TechFlowDaily

X (Twitter):https://x.com/TechFlowPost

X (Twitter) EN:https://x.com/BlockFlow_News

Source: www.techflowpost.com