Once again, Bitcoin prices experienced a short period of volatility this week based on a false post by the official SEC account on X on January 9, 2024 at 21:11 UTC. Approximately 15 minutes after the fraudulent announcement, Gary Gensler gave an official statement from his personal account explaining that the SEC account was compromised.
This blog post offers a case study showcasing the Pyth Network’s handling of this sudden price spike and subsequent crash.
Another One
In the plot below, you can see the Pyth aggregate plotted along with its confidence interval shown shaded in purple, between 21:11 and 21:26 UTC on January 9, 2024.
During the peak of the spike, BTC hovered around $48K on most exchanges. Pyth confidence intervals widened accordingly to reflect uncertainty between different BTC markets.
We overlaid the Pyth aggregate price series for BTC/USD to the BTC/USDT market in OKX, a considerably liquid pair and therefore a good proxy for a live Bitcoin price. You can see in the plot below that the Pyth aggregate price (in purple) closely follows the OKX price (black). Even during times of sudden volatility, Pyth Price Feeds can track this activity across multiple real-world markets at high resolutions.
Let’s take a closer look at the composition of the aggregate Pyth BTC/USD price.
The graph below illustrates the aggregate price, accompanied by its confidence interval. It also displays the lowest and highest prices of the active publisher (or data provider) prices, along with the 25th and 75th percentile of active data provider prices.
The proximity of the 25th and 75th percentile ranges suggests that the majority of data providers reliably tracked the price surge on primary centralized exchanges. While there were a few anomalies in both directions (e.g. the max of the active publishers), the Pyth aggregation method successfully eliminated these outliers.
The graph below displays the standard deviation among active publishers, the interquartile range (the gap between the 75th and 25th percentiles) of these publishers, and the breadth of the width of the aggregate confidence interval.
The Pyth aggregation algorithm's ability to filter out anomalies is evident in this visualization: the standard deviation among publishers notably exceeds the interquartile range, primarily because of a handful of outliers. In particular, the publisher that reported the maximum price between 21:20 and 21:45 may have had an issue with their data sourcing during that time period. However because the BTC/USD feed generally features around 30 active publishers and due to the outlier-robust nature of aggregation, the aggregate price cleanly filtered out this anomaly. The Pyth confidence interval closely aligns with the interquartile range, instead of aligning with the outlier range.
One interesting observation: During the one-hour window in question, the CEX price on one of the major exchanges was captured by one confidence interval around the Pyth aggregate for 73% of the time. The CEX price was within 2 confidence intervals 93% of the time.
This outcome reflects good calibration: data providers’ confidence intervals widened when there was confusion between traders and liquidity venues, and tightened when there was consensus between them. This behavior is expected from the Pyth confidence interval features and offers powerful, actionable insights to downstream protocols using the Pyth BTC/USD price feed.
Zooming In: The Fake SEC Tweet
Let’s take a look at one slot in particular at 21:12:30 GMT that day.
The aggregate price here was $47,673.93. The minimum publisher published a price of $46,720.02 and was one of only 2 publishers to publish a price below $47,000. The maximum publisher had a price of $47,826.18. The aggregate confidence interval width was $105.77 considering that 23 active publishers were in a relatively tight $350 range and only 5 publishers were outside of that range.
Zooming into the main period of the initial spike, 21:12-21:15 GMT, we see visually that the confidence intervals expanded amidst the greater volatility and uncertainty.
Indeed, during that 3-minute stretch, the average aggregate confidence interval width was over $77, or approximately 0.15%.
Over the remainder of that hour, the average aggregate confidence interval width was about $45, or approximately 0.1%.
Zooming In: The Gensler Redaction
Upon publication of Gensler’s post debunking the fraudulent ETF approval announcement by the SEC’s official account, the price of Bitcoin predictably responded across many centralized exchanges. The dip went as low as $44,750 for some venues in that one-hour window.
Pyth publishers immediately mirrored the real-time price decline, causing a swift drop in the reported aggregate. The Pyth confidence interval widened substantially to capture the price dislocations across different liquidity venues.
As prices across these platforms gravitated back to approximately $45,600, the quotes from the publishers also aligned, leading to a corresponding decrease in aggregate confidence.
Pre-news confidence (around $46, or approximately 0.1%) was lower than post-spike confidence (around $71, or approximately 0.15%).
Increased uncertainty surrounded the BTC price because of ongoing concerns about the news and its subsequent correction, leading to heightened price volatility after the spike and wider variances in publisher quotes.
As uncertainties diminished over time, the aggregate confidence interval also narrowed. This case study demonstrates that the Pyth Confidence intervals once again functioned effectively as intended, maintaining resilience amidst market unpredictability.
We can’t wait to hear what you think! You can join the Pyth Discord and Telegram, and follow us on X and LinkedIn.
Stay Updated with Pyth
Stay informed about Pyth network's development and upcoming events!
