Things can move really fast around here. There are challenging problems to solve, choices to make, and new information flying at you — often all at once. You’re always welcome to talk to us if you ever need help thinking through a tricky decision. But in case we’re not available, here are five “operating principles” that reflect how we personally navigate decisions and make day-to-day choices. These principles have served us well, not only in running AlphaLab but over the course of our entire professional and personal lives. We hope they help you too.
– John and Michal
Process over outcome
Let’s start by talking about poker. Contrary to popular belief, poker is not gambling. There is a great deal of chance involved in poker but there is also a real advantage (alpha) to playing with a good strategy. But that alpha doesn’t reveal itself in every hand. You can play well and lose a hand, and you can also play poorly and get lucky.
Beginners tend to care too much about winning each and every hand. There’s even vocabulary to describe this: in the poker world, getting upset after a string of losses is called “being on tilt.” Good players take a longer view. They know that if they are consistent with a good strategy, they’ll hit positive expected value over time. Rather than riding a rollercoaster of emotions on each win or loss, they focus on learning the game and improving their strategy. The best professional players have a zen-like indifference towards any particular hand. They’re playing for the long-term.
The only way to be consistently good is to focus on process. There’s a TED talk about this that we recommend by Alex Honnold, who is famous for climbing terrifyingly tall mountains with no ropes to protect him from falls.
In 2008 Alex became the world’s first person to free-solo the 792-meter tall Half-Dome. This was an absolutely incredible record-setting feat. But for Alex, it was an unsatisfying experience. He describes being unsure of himself during the climb, afraid he wouldn’t make it to the top and feeling like he had “gotten away with something.” He says, “I didn’t want to be a lucky climber. I wanted to be a great climber.”
So Alex set as his next goal to free-solo El Capitan, a 1,000-meter cliff that takes other professional climbers four to six days to summit. This time, he prepared meticulously for years, and when he finally did summit El Capitan in 2012, he says it felt “like a victory lap.”
The outcome of our work at AlphaLab also has components of skill, effort, and luck. We can’t control luck, so we try to ignore it and just focus on skill and effort. We try not to beat ourselves up when things go wrong, and we also try to not get too cocky over our wins. Let’s aim to be great, not just lucky.
People over process
Having said that, please don’t confuse focusing on process (good) with becoming a process-ruled company (bad). Paul Graham has a great essay on the cost of process where he says:
We want AlphaLab to be a place where ideas can be executed fast. After all, this company got started as a far-fetched idea when John built and launched the first version of our trading system over a weekend.
To this end, we generally allow people to launch experiments without much or any approval. Sometimes this leads to outages or mistakes, such as when an engineer launched a misconfigured algorithm and cost us a few hundred thousand dollars*. When that sort of thing happens, we conduct a post-mortem to identify what went wrong, what we can do better moving forward. We share that post-mortem publicly, not to shame anyone but so everyone can learn from it. We are aware of the inherent risks in giving people freedom, but we firmly believe that the upside (more good ideas, executed faster) far outweighs any downside.
As our company grows, this sort of freedom becomes harder and harder to maintain. People start to create layers of approvals and policies, often with the best of intentions but without fully recognizing their costs. And like the proverbial boiling frog, it’s hard to see what’s happening until we’re all bogged down by bureaucratic overhead.
We don’t want to become cooked frogs.
To protect our collective freedom and speed of execution, we focus on 1) putting in safeguards against disaster and 2) cultivating trust in people (see below, “err on the side of trust” for more on this concept). We cultivate trust by sharing context, information, and history. Instead of “follow this process” or “get approval before you do that”, we try to say: “here’s what we currently do and why, and here are some other things we’ve tried before that didn’t work so well. If you have ideas on how we can do this better, please go ahead.”
Both safeguards and trust are crucial to avoid bureaucracy, which tends to get created when people don’t trust each other to exercise good judgment. When we do trust each other, and we know there’s protection against disaster, then we can all continue to be nimble and move fast.
* Note: No, he didn’t get fired.*
Err on the side of trust
Some more thoughts on trust: the quant trading industry is traditionally very mysterious. Cloak-and-dagger. Trading infrastructure, algorithms, market access, and performance are closely-guarded secrets that are often not shared even between people at the same company.
Our approach is different. We believe that within an organization, transparency and access to information can be a real competitive advantage.
Look at the Mongol Empire and the Rothschild family: one ruled a continent and the other built the world’s largest banking firm. What did they have in common? Exceptional access to information.
Ogedei Khan (Genghis Khan’s grandson) designed a postal system that was unprecedented in scale and speed. Messages could move across the Mongol empire at the speed of over 300 kilometers per day. (Even a thousand years ago, latency was a big deal.) Similarly, Mayer Amschel Rothschild sent his five sons to set up shop in five different European cities and had them share information with each other. This family network gave them a huge advantage in the markets as they often knew about events before anyone else. For example, Nathan Rothschild in London found out who won the Battle of Waterloo in Belgium a full day before the government’s official messengers even knew, resulting in some very profitable trades.
This is why AlphaLab is unusually transparent. We choose to make (almost all) information freely available to the entire team. This requires a high degree of trust in everyone, which comes with correspondingly high expectations of trustworthiness.
The entire team has access to information about where our technology is deployed, which strategies are doing well, how our technology works, and what other people are working on. While we keep certain sensitive information private, the majority of our Slack channels and working documents are open to anyone in the company to read. We ask everyone to put their to-do lists in a central place (currently we use ClickUp) so that everyone can see what other people are working on. We make a practice of writing and sharing post-mortems: a summary of what went wrong, what we’ve learned, and how we intend to improve going forward.
We believe that arming everyone with as much company knowledge as possible allows all of us to make better decisions, fewer mistakes, and coordinate better with each other. It allows us to move faster, and it also eliminates the power imbalance and politics that tend to come with privileged access to information.
Build, measure, learn, iterate
This one comes from “The Lean Startup” by Eric Ries. Build-measure-learn is a circular series of steps, designed to minimize wasting time and resources on building the wrong things and maximize learning as quickly as possible.
Our fundamental mindset is that everything is an experiment. Whether we’re working on our trading platform architecture, our recruiting processes, or a new DeFi strategy there are always uncertainties, unknowns, and assumptions being made. Build-measure-learn-iterate reminds us to make those assumptions explicit rather than implicit, to formulate our assumptions as hypotheses, and to gather data that can prove or disprove those hypotheses as quickly as possible.
The basic steps here are pretty simple:
- Build: the simplest MVP that will allow you to test your hypothesis
- Measure: what does the data say about your hypothesis?
- Learn: should we persevere or pivot? What should we change, what can we do better?
- Iterate: repeat from step 1
Without the build-measure-learn-iterate mindset, it’s easy for a company to fall into the HiPPO trap (highest paid person’s opinion). In every meeting, there is someone whose job title or seniority gives them a certain status, and when decisions need to be made the group naturally looks to that person. In fact, often we are the HiPPOs! But our opinions, just like everyone else’s, are only hypotheses until the data proves otherwise. We want AlphaLab to be a place where decisions are made based on data, not based on “John or Michal said so.”
Making decisions based on what was successful in the past sort of works if the world is very stable and changes slowly. But the industry we work in is complicated and changes daily, so we must constantly update our products, our processes, and our mental models of the world. Build, measure, learn, iterate.
Load the ship and set out
Here’s a fascinating pattern from the history of human innovation: new inventions do not appear linearly spaced out across time. It’s more like a step function. Once in a while, a new “foundational technology” emerges and then follows a burst of rapid exploration, iterations, and applications. During these windows of opportunity, those who are bold can make fortunes and change the world.
In the 15th century, Portuguese ship-builders invented something called the caravel. The caravel was a small ship, cheap to build, yet incredibly fast, maneuverable, and seaworthy for long journeys. This advance in ship-building technology (along with advances in navigation like the magnetic compass) dramatically lowered the cost of traveling and trading across the globe, launching the Age of Exploration.
These foundational technologies used to appear only once every few generations, but now they are coming more frequently. In the 1970’s, TCP/IP was invented which enabled the development of the internet as we know it. Today, blockchain is the latest such foundational technology. There is so much to build. Much of the infrastructure, applications, and software that we will all take for granted in a few decades are only glimmers in someone’s imagination today.
Right now, right here, we are in an amazing window of opportunity. This is the time to push ourselves, to try new things, to take risks. In 2017, we started trading in cryptoassets before “kimchi premium” was a widely known phrase. In 2018, we were one of Compound’s very first launch partners. In 2019, we started trading DeFi before it was called DeFi. Next year, who knows?
At the end of the day, we’re incurable optimists. And we agree with Nat Friedman (CEO of Github): “Pessimists sound smart. Optimists make money.”
Or if you prefer a much older version of the same sentiment, then we refer you to this Rumi quote that hangs on the wall of our Singapore office: