Understanding Centralized Exchange Liquidity

Those who read our newsletters know that Kaiko Research is fond of liquidity analysis. This week, we felt it would be useful to take a step back and explore the liquidity data that we use, how we use it, what it means and why it matters. But first, what is liquidity?

What is liquidity?

In the simplest terms, liquidity in crypto measures how efficiently a token can be bought and sold. According to our token liquidity rankings, BTC is the most liquid token, meaning that someone could, for example, buy or sell $100k worth without significantly affecting its price. Someone trying to sell a $100k of a less liquid token would face slippage, meaning that they receive less in USD terms than what they sold is theoretically worth. For example, if someone tried to sell $100k worth of WLD on Uniswap V3, they would face slippage of 6.3% and only receive $93.6k USDC.

It’s been our contention that liquidity is a better measure of a token’s “true” value than market capitalization; the former measures how much of a token can be converted into fiat/stablecoins while the latter simply multiplies token supply by price.

the data

When we first pull the data from our API using Python, we get a CSV that looks like this: 

Note that I’ve hidden the vast majority of the 70 columns of data that we receive. This data can be broken down into a few main categories:

• Date and Time
• Depth (bid and ask, from 0.1% to 10% from the mid-price)
  • Depth is the sum of limit orders a certain percentage away from the mid-price (more on this later).
• Mid-Price
• Spread 
  • The difference between the best bid and best ask.
• Reference Data (exchange and pair)
• Slippage (bid and ask)
  • This is a percentage, given based on a hypothetical order size. For example, we can get slippage data for a hypothetical $100k buy (executed on the ask side) or sell (bid side)

Granularities and Aggregations

Kaiko’s most granular order book data is taken at the snapshot level. We collect two order book snapshots per minute, for all instruments traded on all exchanges. This data contains every individual bid and ask on an order book.

However, this data is massive and quite difficult for researchers to work with, which is why we designed some more convenient aggregations.

All order book data that we use is aggregated, meaning bids and asks are summed at various various price levels, for example “1% bid depth” vs. “10% bid depth.”

This makes it a lot easier to analyze market depth, but we still run into a granularity problem, in that for any day, market depth for an instrument contains ~2800 data points (derived from our snapshots taken twice per minute). That’s why we also have an additional level of aggregation: we take averages of this data over time intervals.

We can take average market depth at 1 minute, hour or daily intervals. The research team typically anlayzes market depth for an instrument at daily intervals.

Asset Level Liquidity

Instrument-level order book analysis is quite useful when assessing specific market events, however it remains difficult to understand liquidity at an asset level.

That’s why we developed an even easier order book data type: Asset Liquidity Metrics.

This data sums market depth and volume across all traded instruments for an asset. For Bitcoin, this would include liquidity for BTC-USD, BTC-USDT, ETH-BTC, etc on all exchanges in our coverage (powerful!).

To summarize, raw order book data is really tough to work with which is why aggregations are important, especially for researchers.

how we use it

We’ll begin by showing some of the most granular data that we use then progressively broaden the scope to more aggregated, global measures. The first example comes from this week, when a news outlet mistakenly reported that a spot BTC ETF had been approved. The chart shows 0.1% market depth on Kraken, one of the most liquid exchanges for BTC. Put another way, the top half shows the BTC value of limit sells 0.1% above the mid-price, while the bottom half shows the BTC value of limit buys 0.1% below the mid-price. 

The chart reveals some fundamental principles of crypto liquidity. First, order books take time to “refill” when price moves quickly, as was the case between 1:15 and 1:30pm on the chart above. Second, order books are constantly fluctuating as traders and market makers add and remove limit orders. Minute-by-minute measures are especially shaky, which is why we sometimes opt to chart a moving average. Last, order books are rarely perfectly even, and usually show some skew to the bid or the ask side.

Below is a more aggregated measure, this time showing the sum of the liquidity of all BTC pairs on some of the most liquid exchanges.

This is still minute-by-minute, showing that Kraken’s liquidity rebounded faster than on other exchanges, while Coinbase and Binance also performed fairly well.

We can zoom out again, this time expanding our range around the mid-price to 1% on each side rather than 0.1%.

By this metric, Coinbase and Kraken are the clear top performers, followed by Binance in third. 

Our most commonly used depth measures are 0.1% and 1%. As discussed in our token liquidity rankings (which includes both), these measures are relevant for different participants; higher frequency traders are more likely to care about 0.1% while longer term holders will watch 1%. We have moved away from using 2% depth – which those who use CoinGecko and CoinMarketCap will be familiar with – because we had observed that this level is more frequently gamed precisely because it is reported on the aforementioned sites. 

The most aggregated liquidity charts we publish combine the liquidity of a variety of assets on reputable exchanges. On the left is the combined liquidity of BTC and ETH at both the 0.1% and 1% levels, while the other 38 altcoins included in our liquidity rankings are on the right.

These longer term, more aggregated measures provide a 30,000 foot view of liquidity, in this case showing that wider bids and asks for BTC and ETH dwindled last quarter while the 0.1% depth remained flat and that 1% depth for altcoins was volatile but ended the quarter flat, while there was improvement in 0.1% depth.

why it matters

The last bull run was rife with tokens reaching multi-billion dollar market caps and even larger fully diluted values; the most infamous of these was FTT, which FTX and Alameda Research used to prop up their businesses while the token was far less liquid than its market cap suggested. The effects of their collapse are still being felt throughout the industry to this day, with liquidity still more than 50% lower than it was pre-collapse. We are still exploring new ways to maximize the insights from Kaiko’s liquidity data, particularly by incorporating volatility to measure how liquidity changes as market conditions change.