When Liquidity Matters

Comparison of crypto exchanges during the April Fool’s Day Bitcoin Rally

Photo credit: bitcoinist; cryptobasicpodcast

Bitcoin Rally on the Night of April 1st PDT

In the short period between 21:30 and 22:22 PDT, bitcoin, the most widely traded digital currency, jumped 21 percent reaching $5,079, a level not seen since November 2018. Since then, the entire crypto community is in an uproar over following the Bitcoin rally.

Understandably, the price surge has led to a renewed interest in Bitcoin, resulting in skyrocketing market demand. As the price climbed higher, an increasing amount of people frantically tried to log on to exchanges and apps, trying to ride the trend, hedge risk, or sell for profit. This resulted in a snowball effect of buying as shorts were closed and buy stop orders were triggered, causing its price to rise faster and higher. Usually, big buy or sell in Bitcoin can often lead to outsized moves, and trend-following individual investors and evacuation of market makers can also exacerbate volatility.

Thus, it is very important for exchanges to be able to provide enough liquidity to investors. This liquidity crisis provides us with a great opportunity to evaluate the real quality of each exchange.

How Exchanges Reacted Under Crisis

Following the Bitcoin (BTC) rally, it is time for an in-depth look on how exchanges reacted. According to the hourly k-line of BTC on Binance, we could readily see that at 21:00 PDT on April 1st, Binance recorded a total of 12,384 BTC traded in a single hour, and soon reached 21,248 as of 22:00 PDT. As a comparison, the volume of the previous hour was merely 1,510. Similarly, Coinbase-Pro recorded 6,889 BTC, Bitstamp handled 3,798 and Kraken traded around 4,121 BTC at the same time of 21:00 PDT, all increased more than an order of magnitude.

Hourly k-line of Bitcoin/Tether (BTC/USDT) on Binance

In fact, as shown in the Hourly trading volume graph of major BTC exchanges, nearly all of them spiked 10 times after 21:00 PST.

While, qualitatively, the volume increase looks similar in all exchanges, we are interested in detail how much volume increased across different exchanges and how much order book has thinned out (i.e., the price impact in different exchanges).

Change of hourly trading volume on recorded exchanges before and after the Bitcoin jump. On the majority of exchanges, volume increased by more than 10 times.

Trading Volume is the total amount of transactions traders conducted in a specific market trading window. It sometimes reflects the potential volatility of the market. High trading volume indicates that there are two groups of people holding very different opinions on the future trend of the market. The trading volume data be can be collected from various sources.

Price impact refers to the correlation between an incoming order (to buy or to sell) and the subsequent price change. It can be measured by several different statistical models such as the famous Kyle’s lambda or Amihud measure. In this article specifically, we loosely use this term to refer to the immediate price change caused by a 10-BTC market order (which reflects the thickness of the orderbook). We first measure the half-spread (half the price difference between the bid and that ask), then we remove the top 10-BTC from the orderbook on both sides, and measure the half-spread again. Their difference can be used as a reasonable measure of the thickness of the orderbook, which reflects the market liquidity, and gives the investors greater flexibility with minimal loss of value. In other word, the lower the price impact, the more robust the market.

By looking at these two aspects, we are able to get some perspective on which exchange is more robust, and which exchanges are only good in “normal” periods, and a crisis debunks their flashy appearance.

In the price-impact-over-time plot above, we find that Gdax (Coinbase-pro), Binance, Huobi and Liquid maintained relatively low price impact, before and after the incident. Moreover, Binance stands out as the exchange whose price impact increased the least during the incidence, which shows it robustness in crisis. We will see the same conclusion later with more quantitative analysis.

Quantify and Compare Crypto Exchange Reactions

Sophon Tech Inc. records L2 data (order-by-order activities) from several major exchanges all over the world, including China, US, Japan and Europe. Quantitative analysis are performed on these orderbook data before and after this April Fool’s Day incidence to compare those exchanges in different aspects.

Observation 1: Orderbooks thinned out by different degrees

Some exchanges have thick orderbooks in normal time, but they quickly thin out in crisis. We have observed this qualitatively in the price impact graph above. However, we can better observe this via the peak-price-impact vs 24h-median-price-impact plot below, which illustrates the behavior of an exchange during normal trading and peak trading period. The ideal exchange with the highest liquidity and strongest resilience to crisis would locate at the bottom left corner of the plot.

The ratio between the peak-price-impact and the 24h-median-price-impact is written on each data point.

For normal trading, many exchanges have very low price impact, such as Binance, Gdax, Houbi, Liquid, and Okex, which all locate at the left side of the graph and the average price impact (measured by 10-BTC method) is less than 10 bps in general.

During the crisis, however, the price impact jumps up, and the quicker market makers escape from it, the higher it becomes. For example, for Gdax, Houbi, they are among the exchanges that have low average price impact but the peak price impact increased significantly. In these exchanges, the orderbook thinned out quickly and it is very likely that many market makers temporarily left the exchange without hesitation.

On the other hand, the lower changes in price impact do not necessarily indicate high resilience to risk, especially if the exchanges have very high price impact to begin with. As an example, Zb has the lowest peak-price-impact to 24h-median-price-impact ratio, but it is still one of the most illiquid exchanges, with a staggering 60+bps price impact. Therefore, these exchanges are not suitable venues for trading even in normal periods and should not be compared in the same context as the premium exchanges.

On the opposite, Binance, which has low 24h-median-price-impact and relatively low increase during the crisis, is a perfect example of a robust exchange. The resilience to the sudden price change of Binance stands out, which compelled more market makers to stick around in on the exchange.

Observation 2: Fake trades could not hide

Fake trading volume in the cryptocurrency community is a persistent problem. There are many different ways to measure this. Interestingly, with this recent price surge of BTC, we have yet another way to make this fake-trading pandemic apparent. When crisis happens, real volume increases significantly, but not at all for the fake volumes. So the ratio of volume increase would be greatly diluted by the fakes trades on those culprit exchanges. Therefore, we can be a measure of how “fake” the volumes are for each exchange by measuring peak vs 24h-median volume ratio.

The ratio between peak-volume and 24h-median-volume are written on each data point.

In the plot above, noticeably, Liquid, Binance and Gdax all have very similar volume increase ratio of 23.5, 22.4 and 21.0, respectively. On the other hand, Okex, Huobi, Hitbtc have only 7~9 times volume increase with respect to their 24h-median-volume.

With our previous argument, and by assuming Gdax, Binance and Liquid volumes are authentic, we can infer that more than half of Okex, Huobi and Hitbtc volumes are fake.

As a matter of fact, Gate.io, and Kucoin have even higher volume hike ratio numbers. However, since their normal volumes are very small, we treat these high ratios as irrelevant outliers.

Aftermath of the Price Jump

After the BTC jump, the orderbook thinned out and hence the price impact increased significantly. More intriguingly, it had continued to increase on most exchanges that we record in the following few days, as shown in the 24h-Moving Median Price impact plot. This BTC price jump incidence has a long-lasting repercussion.

After the BTC jump incidence (the dotted line), the 24h moving median of hourly average price impact increased on most exchanges that we recorded.
The price impact after the BTC jump generally increased by 1.5 to 3 times.

After this event, the orderbook continues to thin out in the following three days. This can be shown by the steady increase of the 10-BTC price impact across all exchanges, be it from China, Japan, US or Europe.

This indicates that market makers are still very cautious even several days after. If this trend continues, it implies large price volatility, and it is potentially ominous of another round of big price move.


Through the BTC April Fool’s day rally, we come to several observations about the real quality of exchanges in the current crypto market that we may not able to observe so clearly in normal days.

Some exchanges have much more resilient liquidity than its counterpart under crisis (e.g. Binance vs Gdax). In addition, the different ratio of trading-volume hike reveals fake volume from several exchanges (Okex, Huobi, Hitbtc). Lastly, this BTC price surge incidence has long-lasting repercussion towards market makers, who are very reluctant to place orders back near the top of the orderbooks.

About the Author: SophonTech Inc. is a technology-driven company that provides solutions to cryptocurrency software and quantitative research.

Contributor and Editorial Credit: H. Zheng, D. Zhao

Legal Disclaimer: SophonTech Inc. is not an investment advisor, and makes no representation regarding the advisability of investing in any security, fund, token, derivatives, physical assets or any other investment vehicles. All SophonTech Inc. materials have been created solely for informational purposes based upon public information from sources generally believed to be reliable. The data and analysis demonstrated do not represent the results of actual trading/investing.