May 12, 2021 - 12:28pm EST by
2021 2022
Price: 17.00 EPS 0 0
Shares Out. (in M): 220 P/E 0 0
Market Cap (in $M): 3,740 P/FCF 0 0
Net Debt (in $M): 0 EBIT 0 0
TEV (in $M): 0 TEV/EBIT 0 0
Borrow Cost: General Collateral

Sign up for free guest access to view investment idea with a 45 days delay.

  • Wisdom by Katana
  • bubble
  • please write up more wildly expensive stupid stocks that you're not involved with
  • I'm not saying we're in a bubble but...


NOTE: This is a pairs trade in crypto. The fundamental idea is that SUSHI should be zero; that's why I listed this idea as short. The long leg is for hedging purpose. (You could blow up without the hedge.)



BTC is trading at $56600 per coin. From the buying/selling bitcoin perspective, many of us (me included) don’t know what to do with this painful fact; because on the one hand blockchain enabled decentralization technology is very likely to play an important role in our economic life 10 years later, so shorting coins have blow up risk; on the other hand all crypto currencies are trading at exorbitant prices at this moment and it’s unclear whether any of these coins would be valuable when the decentralization game is played out. Here I propose a crypto pairs trade that:

(1): can help you learn about crypto currencies and serve as a stepping stone into this new world;

(2): has minimal blow up risk;

(3): is easy to execute because both legs have ample liquidity;

(4): is sound based on traditional investment concept;


The trade is to long UNI (Uniswap coin and short SUSHI (Sushi coin in equal USD amount.


Explaining how these two coins work from the ground up would make this write up 10 times longer; so I will focus on key contract terms and ignore technical details. You can read the white papers on their website if you’re interested in more details.


Traditional Market Making

Based on hundreds of years of experience, we know that the most efficient mechanism to trade securities is via two sided order book on centralized exchanges. Market makers make money by collecting a fraction of the spread (and commission paid by liquidity takers), if they can manage inventory well. Without positive spread and trading commission, market makers cannot make money because they suffer from adverse selection; placing non-marketable limit orders on the order book is essentially laying out free options. Therefore:

        Market Maker Profit = (Spread + Commission) - (Adverse Selection Loss)



Just one year ago, the majority of crypto currencies trading happens on centralized exchanges. But now the world is different! A meaningful percentage of total trading volume is on DEX (so called decentralized exchange; see this wiki page ). The key differences between CEX (centralized exchange) and DEX are:

(1): CEX requires custodian service and trades do not settle immediately. CEX (e.g. Binance, Coinbase Pro, etc) is similar to traditional stock exchanges. In order to trade stocks, you need to deposit money and stocks to a broker. You cannot directly trade on stock exchanges, only brokers can. When brokers trade stocks on exchanges, those trades don’t immediately settle (i.e. money <-> stock transfer does not happen immediately); instead, exchanges will keep track of all trades for a given trading day and settle once each day; settlement (i.e. money <-> stock transfer in custodian accounts) usually happen on the next business day. Trading crypto currencies on exchanges like Coinbase Pro follow a similar process. You have to first transfer your coins from your wallet to the exchange wallet in return for a receipt from the exchange. Then you can trade coins on the exchange; notice that such trading does not generate any coin movement on the blockchain because all these coins are in the exchange wallet; they just change numbers on the exchange internal system. (Usually exchanges act as their own custodian so there is no end of day settlement. Also end of day settlement is not practical, because each coin can only generate a designated number of new blocks with limited size within a fixed time interval while there could be arbitrary number of transactions within a day.) When you cash out, the exchange will transfer coins from the exchange wallet to your wallet in return for your claim check on the exchange. Notice that this whole process requires you trusting the exchange. If the exchange goes broke, or being seized by the goverment, or being hacked, then your money is gone and there is generally no way to recover it. Also the exchange could be secretly running its own prop trading algorithms against incoming flows to profit from its informational advantage over regular traders. When you withdraw money/coin from the exchange, the exchange can front-run you by refusing or delaying your withdrawal. Basically, crypto exchanges have thousands of different ways to unfairly profit from its customers and there is currently no regulation to stop them. (Like stock exchanges before 1933……)

(2): DEX does not require custodian service and trades all settle immediately on the blockchain. This is made possible by cleverly designed contract (liquidity pool) on the Ethereum platform. Your first reaction after hearing this might be WTF??? How is this possible; there is no way to efficiently post and cancel millions of limit orders on the blockchain! Also it would be extremely slow because bitcoin blockchain can only process 5 transactions per second by design. There is nothing wrong with this because all great ideas sound stupid/impossible to most people initially. The (current) solution to this problem is called AMM; automatic market making. This is a market making algorithm that does not require the usage of an order book. Keep reading!


Automated Market Making (AMM)

Uniswap is using the constant product AMM: x*y=k so my explanation will be focused on this particular example. The argument here can be trivially generalized to the y = f(x) case.

Imagine you want to make a market to facilitate trading apples against bananas. Suppose the current fair price is 1 apple = 5 banana; usually this “fair” price is generated by taking the MidPrice from an order book based exchange. You hold a “liquidity pool” consisting of: 1000 apple and 5000 banana.

(1): Liquidity Providers. To provide liquidity via this pool; you can deposit 1 apple and 5 banana into the pool; so that after your deposit the pool has 1001 apple and 5005 banana; and in return you get a LP token (liquidity provider token) representing 1/1001 ownership of the liquidity pool. In general you can deposit { apple, banana } basket into the pool using the current pool ratio in return for some LP tokens; for example, if before your deposit there are A apple and B banana; then you can deposit tA apple and tB banana into the pool; and get LP tokens representing 1/(1 + t) ownership of the pool. Later on you can surrender your LP token to get your share of the pool back.

(2): Liquidity Takers. This is where the constant product formula x*y=k comes into play. Suppose there are A apple and B banana in the pool and you want to exchange t apple for some banana. In the absence of transaction fee, the number of banana you get is determined by the following rule:

    (number of apple before) * (number of banana before) = (number of apple after) * (number of banana after)

Suppose you get s banana; then we must have:

    A * B = (A + t) * (B - s)   s = t*B / (A + t)

Notice that if t is infinitesimally small, then the ratio s/t is about B/A. That’s why I said initially if the fair price is 1 apple = 5 banana; then typically we should have (number of banana in pool) / (number of apple in pool) = 5. The number of banana you can get for one unit of apple is B/(A + t) which is a decreasing function of t; this is in accordance with our intuition: trading larger size incurs larger market impact.

(3): Arbitrage. When (number of banana in pool) / (number of apple in pool) != fair price; you can “arbitrage” between this liquidity pool and the reference exchange. Let’s illustrate how this work by a concrete example. Suppose the fair price is 1 apple = 5 banana; and the pool has 1000 apple and 4000 banana. So A = 1000, B = 4000, and the pool implied price for num-banana-per-apple B/A=4 is 20% lower than the fair price; i.e. apple is trading cheap in this pool. Therefore we should buy apple from this pool. Let’s say we buy t apple by paying s banana:

    (A - t) * (B + s) = A * B = 4000000

We should buy all the way until the pool implied price is equal to the fair price:

    (B + s) / (A - t) = 5

Therefore B + s = sqrt((A-t) * (B+s) * (B+s) / (A-t)) = sqrt(4000000*5) = 4472.136; s = 472.136; A - t = (B + s)/5 = 894.4272; t = 105.5728. So you should buy 105.5728 apple from the pool by paying 472.136 banana with per unit price of 472.136/105.5728 = 4.472. On the other hand, you can sell 105.5728 apple on the reference exchange at a price of 5 banana per apple to get 5*105.5728 = 527.864 banana. These two combined gives you 527.864 - 472.136 = 55.728 banana for free!

(4): Liquidity Pool Economics. Random liquidity taking trades against the pool do not increase or decrease the pool value on average; but arbitrage trades as described in (3) will decrease pool value. To incentivize liquidity providing, the pool must charge some transaction fee so that the pool value is growing in expectation. Currently Uniswap is charging 0.3% (i.e. 30bps); if you exchange s banana to t apple you have to pay extra { 0.3% * t apple, 0.3% * s banana } as transaction fee into the pool. Now we come to the key question: what’s the expected annualized return for liquidity providers for a given level of transaction fee? This question is actually very difficult to answer. What I’m going to do is to develop a toy model with simplifying assumptions so that we can understand the situation at least qualitatively. I assume the audience is familiar with basic option pricing theory (Black Scholes, etc). Let’s say the pool initially has A0 apple and B0 banana with A0 * B0 = K. Suppose the price num-banana-per-apple has annualized volatility sigma; and the price at time T is S(T). Then in the no transaction fee case, arbitrage mechanism forces us to have: B(T)/A(T) = S(T) and the constant product requirement gives us A(T) * B(T) = K. So we should have A(T) = sqrt(K/S(T)), B(T) = sqrt(K*S(T)); and we have:

    begin pool value in terms of banana = A(0)*S(0) + B(0) = 2*sqrt(K*S(0))

    end pool value in terms of banana = A(T)*S(T) + B(T) = 2*sqrt(K*S(T))

So we have:

    endValue/beginValue = sqrt(S(T)/S(0))

If we have a percentage transaction fee of alpha then we should have (remember interest rate is zero):

    endValue/beginValue = exp(alpha*T) * sqrt(S(T)/S(0)) = exp(alpha*T) * exp(0.5*sigma*W(T) - 0.25*sigma^2*T)

This thing is a sub-martingale if-and-only-if:

    alpha - 0.25*sigma^2 >= -0.5*(0.5*sigma)^2.   <===> alpha >= 0.125*sigma^2

Therefore, if all trades are arbitrage trades, then we need the transaction fee to be at least 0.125*sigma^2 for the pool to break even! This is an exorbitant fee level because for most crypto currencies we have sigma >> 1; so we need to charge at least 12.5% transaction fee to break even. The fact that Uniswap is very profitable (yield farming return is almost 40% annualized) at alpha = 30bps means the majority of trades on Uniswap are random trades by uninformed traders. This cannot last long because eventually sophisticated market makers will enter DeFi, and miners are already gradually eating this cake. (For example, see this paper: This is not hypothetical. Miners are already doing this.)




Congratulations if you made this far! Because the rest of this write up is the traditional sum-of-parts analysis so it’s pretty trivial. Let’s address the central question; I’m proposing long UNI and short SUSHI; but what exactly are UNI and SUSHI?

(1): UNI. At this point you should already know what Uniswap is doing. Basically it has a lot of liquidity pools; one for each coin pair; and it charges 0.3% transaction fee on these pools. Part of the transaction fee goes to liquidity providers so that LP can have an attractive return. Uniswap coin holders are entitled to the remaining portion of transaction fee. So this UNI coin actually has some value and you can price it using traditional P/E, EV/EBIT, etc. But I’m not going to do that here.

(2): SUSHI. What’s interesting is this Sushi coin; because it’s …… total air!!! and is trading at exorbitant price levels with ample liquidity right now. Try understand what’s going on by researching their website You don’t even know who is behind this Sushi coin; there is no founder; no company; no team; etc; some mysterious people just made hundreds of millions of US dollars in less than a year; …… with a defunct idea. Maybe it’s a practical joke by the venerable Nakamoto-san. OK, enough is enough; let me tell you what this shit is. Sushi swap is also a liquidity provider program. Suppose you’re a liquidity provider. You can provide liquidity by depositing coins to Uniswap’s liquidity pools in return for Uniswap LP tokens (Note, you don’t get any UNI coin; UNI coin is completely separate which you can purchase/mine directly. Essentially liquidity providers are like LP to a hedge fund while UNI coin holders are GP.). You can also do it via Sushi; when you deposit coins to Sushi, Sushi won’t do anything special; it will just deposit your coins to Uniswap on your behalf and pass over the Uniswap LP tokens to you; in addition to this, Sushi will also give you some Sushi coin! Let’s emphasize that Sushi coin does not economically entitle to anything. The selling point is that Sushiswap is strictly better than Uniswap because liquidity providers can get extra sushi coins. This drawed a large crowd and bid Sushi coin price through the roof. Imagine you run a company called tychus-Limited buying TSLA shares on behalf of customers; and for each share bought through tychus-limited the customer can also get one tychus-certificate for free; then you promote this by telling people that it’s strictly better than buying TSLA share directly because they can get typhus-certificates for free. After some period of time tychus-limited has a lot of volume so people think tychus-certificate must be very valuable and bid up its price. Actually those people buying TSLA via tychus-limited have a strong incentive to bid up the price of tychus-certificate, so there will be a self reinforcing boom as long as the TSLA through-volume picks up. Then you can unload your founder’s share of tychus-certificate and retire to Florida. Viola!



Now you know that the proper price for Sushi coin is zero; but nakedly shorting crypto is dangerous because it can wipe you out in no time. On the other hand, a pairs trade of { long UNI, short SUSHI, dollar neutral } should work with strong margin of safety:

(1): For the record; as of this writing UNI price is 322.48 HKD/coin; SUSHI price is 135.05 HKD/coin. The trade is to bet SUSHI/UNI to drop over the next 6-12 months.

(2): Economically we should have SUSHI=zero. But crypto is crypto; if DeFi volume continue to pick up SUSHI could rise together with UNI. That’s why we need to long UNI hedge. This hedge is sufficient because by design Sushi volume is a subset of the Uniswap volume; and the trading psychology of these two coins are mostly determined by their transaction volume.

(3): There are many catalysts in our favor to make this work over the next 6-12 months. (a): Sushi is already past the promotion phase; it’s likely in the “harvesting” phase where the founding team is aggressively cashing out so there will be persistent selling pressure from substantial holders. (b): People will gradually understand what’s going on in Sushi and stop the buying frenzy. Things in crypto move very quickly. (c): The Uniswap AMM cannot maintain the current level of profitability because it’s simply too inefficient compared with order book based strategies; spread on AMM pools is huge compared with centralized exchanges. Traditional market making firms will enter this space (bring on-exchange information to the DeFi world) and eat their meal. This will cause both UNI and SUSHI to drop; but UNI will have some residual value while SUSHI does not. SUSHI likely will crash first in which case our pairs trade will be very profitable. (d): Flash loan enabled arbitrageurs and other DeFi players will also gradually reduce these AMM’s profitability causing a similar dynamic as the previous item.

(4): Traditional value investing, or any type of investing/trading on equities/futures/commodities/fx markets for that matter feels like a group of professional boxers fighting each other for a limited prize pool; and the pool is shrinking because of ETF, higher taxes, etc. The crypto world is different; it’s a world where professional boxers can legally make money by beating up unsuspecting random spectators. It’s the “Red Ocean” vs “Blue Ocean” that we’re all familiar with. Ignore at your own risk.







I do not hold a position with the issuer such as employment, directorship, or consultancy.
I and/or others I advise do not hold a material investment in the issuer's securities.


See (3) in the summary section.

    show   sort by    
      Back to top