Observations on market-making from GS to DeFi
For those of you who don’t know me, my background (before Melon) was as a market-maker and trader at Goldman. I spent nearly a decade making markets there in some of the most liquid and illiquid equity markets pre, during and after 2008 and was involved in pricing a handful of billion dollar block trades during my time there.
I was also fortunate enough to witness some of the very first Automated Market-Makers (AMMs) being built about 15 years ago. At the time, many of us feared that AMMs would make our jobs redundant. They certainly made some traders redundant. But guess what, today there is still a fully staffed team of market-makers on the trading desk at Goldman and every other large broker-dealer house on Wall Street for that matter.
There’s been a lot of hype around AMMs in both CeFi and DeFi. However, the hopes haven’t really been matched by the reality. The dream was that AMMs would entirely replace human market-making; the reality is that AMMs can only serve markets which are already very liquid. They do not work efficiently in markets which are illiquid, which means that there are limits to automation.
What does scale well with illiquidity is human market-making. In what follows, I’ll set out some of the nuances and soft-factors that go into Manual Market-Making (MMM) which should help illustrate why AMMs can’t scale efficiently. This is important because there are reasons to believe that this is a particular challenge for DEXs. Don’t get me wrong, there is certainly a role for AMMs to play in terms of taking care of high-touch liquid orders. But ultimately, we’ll never see DEX liquidity really thrive without a much bigger involvement from MMMs to take care of the low-touch, less liquid orders.
What is market-making?
Market-making is the provision of prices to absorb liquidity from buyers and sellers. If someone comes along with a large quantity of an asset that they want to buy or sell — a market-maker’s job is to give them a price where they can clear the whole quantity in one go. At an extreme, block trades (very large volume trades) can take many hours to price and may require input from multiple specialists or experts in the underlying asset.
It is normal practice as a trader to request a quote from several market-makers. At that point, you can decide which party offers you the best terms and trade with them. Sometimes, traders do not disclose whether they are a buyer or seller up front. Instead, they request a ‘two-way price’ and only reveal their ‘side’ (ie. whether they are a buyer or seller) after they receive the bid and offer. This is a common method to avoid concerns around front-running.
In short, being a market-maker is all about determining a price (irrespective of the size) which ultimately results in a trade and handling the risks associated with that price in a sustainable manner.
Automated Market-Making (AMM)
Automatic Market-Making is a subset of market-making. It involves an algorithm which uses inputs to automatically provide a tradable price. Traditional market-making desks came up with the concept of automatic market-makers nearly two decades ago. Typically these algorithms will deal with making-markets on the small and liquid orders by, for example, capping the maximum size of trades allowed.
An important word on Liquidity
DEXs are trying to find a solution to the lack of liquidity. But what exactly do we mean by this? Liquidity is the ability to move a large position in a token without impacting the price too much. For any given asset, liquidity is an inverse function of size. It’s very important to understand that regardless of the asset, there is always a size beyond which trading becomes illiquid.
Let’s take a look at a couple of examples:
Example 1: You might think ETH is liquid. But if a sovereign wealth fund came along and wanted liquidity to buy a block of 11 million ETH/USD (very roughly c. $2.4bn notional and c.10% of the overall supply) how many people would make a price that would actually result in a trade with the sovereign wealth fund? Probably not even Vitalik! At this point, Ether is illiquid until someone makes a price that results in a trade.
Example 2: At the other end of the spectrum, let’s take a look at MLN. Looking in casually, you might think MLN is highly illiquid with its market cap at $4m and CoinmarketCap reporting $50K traded on exchanges in the last 24 hours. However, if I told you that I’ve been doing some market-making privately on the side and that I was regularly printing blocks of 150K MLN and settling them ‘over the counter’, you might have a very different picture! Currently data sources like CoinmarketCap do not capture the whole volume picture in crypto because there are no post-trade reporting requirements.
Hopefully the idea of liquidity makes a little more sense now. We’ll come back to ways of measuring liquidity later in the blog. As is also illustrated from the examples above, information is key to market-making. What information do you need?
What makes a market?
1. Volatility: The volatility of an asset is the statistical measure of the dispersion of returns for it (ie. the standard deviation). In most cases, the higher the volatility, the riskier the security. If you know what the daily implied volatility of an asset is, you can work out how much it is likely to deviate from its current price and with what probability. For example, if TokenX was trading at $10 and had an implied daily* volatility of 5%, you could deduce the following:
These probabilities can be very useful when thinking about what price to make.
2. Liquidity: There are different ways to measure liquidity but when I think about the liquidity of a trade, I think of it in terms of how many days it could take to liquidate the position with minimal price impact. I’ll define minimal impact as being 20% of the volume (although this is likely to vary a lot according to the asset). So for example, if someone is asking for a price for 10m KNC token and only 1m KNC token trades per day on average, you are looking at 10 days worth of volume. Assuming that you can unwind the position with minimal market impact by unwinding the position at 20% of the volume — you are looking at probably 50 days to be able to liquidate this position in the market. That’s a very long time to hold the KNC token and means you should price in a much higher level of risk when making your market.
3. Your market share & ability to influence it: If you have a high market share, you naturally have more information about current/recent buyers and sellers. This can provide you with an edge and enable you to match buyers and sellers with minimal risk = better prices = minimising trading risk (maybe even profitable trading!). This is the ultimate holy grail for market-makers; the ability to match buyers and sellers at a price with zero trading risk (double the volume = double the commission). When you have a high market share and you get known as ‘the market’ in a specific name you will naturally be able to find the other side of the trade a lot faster.
5. News & events: If you are being asked to make a price in an asset, you want to know what might be on the horizon for the period you expect to have to unwind the risk. For example, if you get asked to make a price in a large quantity of a protocol token just as a twitter headline is hitting that there might be a vulnerability/hacker attacking that protocol — you want to take that information into account and adjust the price accordingly. This isn’t always possible because the basic rule of thumb is that traders often know more than market-makers. There are thousands of traders and only a handful of market-makers so the probability that they are going to catch on to rumours or news flow faster than market-makers is high.
6. How easy to hedge? Hedging involves taking a position to reduce the risk of adverse price movements in an asset. Normally, a hedge consists of taking an offsetting position (to the one you will inherit as a market maker)in a highly correlated asset (or basket of assets).
7. Reputation of counterparty — Generally speaking, there are always ‘good actors’ and ‘bad actors’ in every market, including heavily regulated markets. There is even more room for bad actors in DeFi because of the lack of a reliable reputation system. If a buyer shows a pattern of behaving like a bad actor, they are likely to be loss-making to a market-maker. Market-makers will learn to price much less widely and less competitively for bad actors than they would for good actors. It’s as simple as that.
8. Recent price & volume behaviour — Today’s price and the asset’s recent behaviour are other useful inputs for a market-maker. Have there been a lot of buyers/sellers that you can see, does the asset feel overbought/oversold? Any reason for that? If there has been a big move and you understand it, that usually gives you edge. If there has been a big move and you don’t understand it, that should usually be a red flag.
9. Commission rate — The commission you are getting on the trade should be a major factor on how you price. There is no reason why there should be one commission rate for all assets. It is perfectly acceptable that more volatile, less liquid assets which are harder to price should come attached with a higher commission rate (eg. illiquid emerging market equities will typically have much higher commission rates (typically 50bps +) than liquid NYSE stocks (typically 4bps).
10. Your inventory & positioning: If you really build a speciality in an asset pair, you should become quite good at predicting the price moves in that pair. If you have a high conviction that BTC will appreciate against ETH, you can start predicting your market-making flow in advance and adjusting your inventory to be ready for when the buyers come. This will enable you to make more competitive prices with minimal P&L impact. This can be the trigger for attracting market share and having an edge (remember market share = volume = commission = profitability).
The bottom line is — information is key to good market making. Building a niche and expertise in a few specific assets is often much more profitable and valuable than being a generalist market-maker in assets you don’t have much of an edge in.
As you can see from the list above, all the listed inputs give you information about the “riskiness” of the market you are making. This information should enable you to make a market. How well your prices are will determine how good a market-maker you are which is reflected in your P&L (trading P&L + commissions) over time.
Market-making in DeFi — AMM vs HUMAN
DeFi players have been hard at work building infrastructure which establishes trust between buyers and sellers without the need for 3rd parties to take custody of your assets. We’ve seen tremendous progress on the infrastructure side. However, where are the market-makers? DEX’s are not scaling and that is because we haven’t got enough good market-makers in DeFi. The example below illustrates this point.
As of today, DeFi’s AMMs simply can’t function when markets get illiquid
The purple parts of the table above are a quick illustration of market-making in DeFi. The quotes are based on a reference price ETH/USDC 212.5. on May 8th, 2020. I’ve looked up what I would get in exchange for selling 20,000 ETH/USD on the Kraken order book in order to compare it with what I would get for selling 20,000 ETH/USDC from Uniswap, 1inch, DEX AG and Paraswap (These are aggregators which provide liquidity through aggregation of AMMs and order books).
I’ve included Kraken in this table to illustrate a point. Kraken is not actually a market-maker but an order book matching exchange. This is really important because as an order book matcher Kraken’s job is not to provide liquidity but to facilitate the orderly placement and matching of limit orders. Unlike order-book matchers however, market-makers are supposed to be liquidity providers and therefore if they are doing their job properly, their prices should always be much more compelling than what is on offer on all the order books out there!
This little table summarises that unfortunately the market-making role is not currently being fulfilled in any meaningful way on DEXs. AMMs are mostly looking after liquid orders and pricing illiquid orders so badly that no rational person would actually want to trade at those prices unless they had no other choice.
So where are all the Manual Market-Makers (MMMs) in DeFi?
One explanation could be that market-making requires a large balance sheet and getting access to those kinds of balance sheets is hard. A possible solution to that might be expanding on the idea of liquidity pools pioneered by Uniswap and finding similar ways to enable them for MMMs eg. via a Melon fund! 🙂
Another explanation could be an inability to differentiate between good trading counterparties and bad trading counterparties in DeFi (at least in the current AMM model). Understanding what kind of person or entity you are dealing with in DeFi is just as important as it always has been in market-making.
Last, but not least, is the possibility that people are not aware of the soft factors and nuances involved in making markets (and therefore do not put as much emphasis on them) or they recognise the limits of the technology in replicating those soft factors and therefore focus the AMMs on already liquid markets. It’s obviously much more fashionable these days to say you’ve built a fully autonomous and automated market- maker. But it’s not going to solve the liquidity problem any time soon!
So to sum things up, AMMs are dominating a large part of DEX liquidity right now but they only work for liquid trades (unless you’re irrational or careless). They can only work where there is already a certain amount of liquidity, so by definition, they’re not actually market-makers. A market-maker is supposed to create liquidity in illiquid situations and automated market-makers can only provide liquidity in liquid situations.
The only way to make illiquid stuff liquid is by leveraging all the soft factors and inputs that go into manual market-making. Ironically, manual market-making in a decentralised way might be more of a people’s business than it is in CeFi because of the many characteristics that are unique to decentralised systems.
One path to ramping up liquidity in DeFi could involve leveraging the idea of market-making pools to provide MMMs with larger balance sheets. But really the biggest chance of ramping up liquidity in DeFi is ditching the “A” in AMM and paying more attention to the softer skills involved.