This is an extract from June Monthly Digest
This section is written by Samuel Trabucco - Quant Trader
Read through each of these (real!) situations from the crypto trading world, then spend 60 seconds making a game plan for how to deal with it. That’s how long I had to evaluate them when they actually happened (maybe – for some of them it was much less!):
- BCH is having a hard fork soon, and it’s proving to be contentious. It’s really unclear how much value will be retained in the coin that will be called “BCH” after the fork, but it’s becoming clear that the answer will be “less than all of it.” OKEx has futures listed that expire after the date of the fork, and so you put on a large position selling those futures vs. buying spot (they’re trading at the maximum discount to spot OKEx even allows). OKEx then tweets that they’ll be expiring them early – in a few minutes, even, so that the futures will expire to a much higher price than you thought. What do you do?
- BNB begins crashing, and then a few minutes later it goes public that Binance – a major exchange with a pretty great record and reputation, and where you trade quite a bit (and hold a lot of capital) – has been hacked. It’s clear some people knew about this beforehand, but the world has just found out within the past 10 seconds. What do you do?
- Amidst a global pandemic and financial meltdown, BTC has been in freefall for hours and it’s showing no signs of stopping. Massive liquidations are getting triggered on all major products and platforms (all of which you trade a lot of and have big positions in), and trading is insane. Premia are massive in magnitude (for derivatives and spot between exchanges), there are great trades to do everywhere, and your own positions are at risk of getting liquidated (and the blockchain is slow, so transferring capital is hard). You’ve got a finite amount of capital, and the right thing to do with it changes by the minute. Bitcoin falls below $4000 from BitMEX liquidations, and then BitMEX goes down. What do you do?
Situations like these have the potential to make otherwise brilliant, rational decision makers freeze. These are difficult decisions with giant stakes that need to be made under extreme time pressure – it’s easy to imagine panic setting in, rendering coming to any decision hard – let alone something close to the right decision. What are the core reasons people struggle with this?
There are many effects at play. A big effect is that many people – especially really smart, math-oriented people who are the sort who end up as quantitative traders – are trained at deep thinking and coming up with the perfect solution to a problem. Given a lot of time and a clear problem to work with, they’re going to nail it – and this is the sort of thinking and skill set that Alameda depends on for building and maintaining its models. But faced with a really complicated decision to make in 5 seconds, they might struggle because they’re not used to coming up with an OK solution in little time – they’re used to coming up with a great solution in lots of time.
Another effect is the natural human tendency toward a “fight or flight” response. Faced with a complex, new situation that you’ve had very little time to think about, even if someone is naturally inclined toward fighting adversity, they might just actually not see the steps they need to take in order to do so. Especially when one of the only features of the system that you do understand is how big the stakes are (in Alameda’s case, the best decision is often worth millions more than the worst one), fighting doesn’t always feel like a realistic option. So, many people default to flight (while trading, this often means taking minimal action, or looking to others for direction).
A third effect is a bias usually referred to as loss aversion. People are often wired to make choices which minimize their potential loss, even if the decision is not the most positive EV (expected value) one they could make. Maybe winning big isn’t enough to justify maybe losing big, so they just act passively and avoid either. This can actually make a lot of sense if the possible loss is gigantic vs. net worth (or future earnings potential) – where the line is can be a subtle risk management problem. But many people avoid loss even if it’s small, and they know the risk is +EV, and that’s a bias that a great real-time trader needs to quash.
So, how do you get better at combating these effects? The dissatisfying answer is “practice.” But that can’t quite be right – it’s pretty hard to practice reacting to situations where millions of dollars are at stake, because before you’re good at them you don’t tend to end up in them. But you can do things to erode the underlying issues so that you’ll be better-suited to reacting to these situations when they do come up – I can point to experiences I’ve had and conscious training I’ve done that’s helped me with all of them.
I studied math in college, and for a period I was pretty confident I’d go into academia. I’ve spent a lot of time solving hard problems that take hours and days, and so my natural inclination is to work toward a perfect solution in a lot of time. Amusingly, my experiences from middle and high school math competitions help a lot here – I practiced a lot of MathCounts, for instance, and the whole goal there is solving a ton of problems as fast as possible. Practicing quick math problems is a direct way to get better at thinking quickly rather than deeply, trying to usually get the right answer but never let yourself be too slow.
My whole life, I’ve loved strategic games, especially ones where you have to make many difficult decisions over the course of a single game. It’s varied what games I’ve played a lot of – I’ve spent many hundreds of hours playing each of chess, Scrabble, poker, Hearthstone, Dominion, and Magic the Gathering – but the thing that draws me in is how much I enjoy the tough decisions they force you to make under uncertainty. You get thousands or reps in making hard (probability-based, in all these cases) decisions, and the skills you learn from it translate pretty directly into the skills you need when trading – and this also helps refine the fast thinking that middle school math problems got me on the right track for.
During college and in the first years I was trading, I spent a lot of time playing poker in particular. At first, I was terrified of losing money, even when it didn’t represent a high fraction of my net worth (even when it didn’t even represent a high fraction of the money I’d bankrolled myself with!). There were many points where, for instance, the correct play was to go all in and I didn’t because there was a chance I’d lose all my money. But refusing to make the correct play in a spot like that can easily be the difference between making money on average and losing it, in the long run. Eventually I convinced myself of that, tried it a few times, and lost “big” a few times – but won big more. And losing “big” a few times gave me the confidence to actually bet big – and now I’ve won and lost quite a bit bigger than I’d have imagined, many times over – 7 figure hours no longer faze me.
How do I apply these skills to trading? Let’s dive into how I (and other Alameda traders) reacted to one of the situations above – the March 12 COVID crash.
BitMEX orderbook on March 12th 11am UTC
This was really several days in a row of hard, fast decisions to make (despite being exhausted and “emotionally unstable” during this period, it was honestly a huge rush for me). One microcosm of the whole thing, though, was when the cascading liquidations on BitMEX’s BTC perpetuals looked like they might just take BTC to 0: What were the relevant factors in deciding what to do?
- Buying a given derivative product has three important factors: how good is its premium, how much capital can we spare for its margin, and how close to liquidation will that leave us? Normal metrics for liquidation danger just fell apart completely – Bitcoin was moving 10% in 5 minutes routinely.
- For products where we didn’t have capital but we REALLY wanted to buy – how badly? What was the expected amount of time a transfer would take, and how much EV would losing the capital for that long cost us? Again, considering that capital to be tied up for an extra-long transfer had really unique implications like “this is unavailable for margin - we’re 20% more likely to get a giant position liquidated and lose a ton” which had never happened before.
- Should we be long or short deltas at any given time? The factors were evolving really fast here, and they were both really different from usual, but also, in many ways, much more intuitive and “easier” to nail if you thought quickly about implications various market events have – for instance, staring at the BitMEX order book and guessing when liquidations were over (for a while) had a pretty high success rate for predicting when to get long, and similar things were happening all over the landscape.
All of these things were changing on a dime, and our whole team was making many decisions each every single minute which had large implications about trading and PNL. There was a ton of shouting and urgency – it was a really cool (if super stressful) environment to be in. I mostly lost my voice during it!
BitMEX going down was the peak of the fun. There were a TON of unresolved liquidations on BitMEX at the time – would those still happen? Was it on purpose? How would the market react? What was going to happen to the funds on BitMEX? If I had a position there and BTC kept going down, would I just get liquidated? How long would this last? There was a lot going on, and a lot of it was super uncertain, but some decisions needed to be made immediately – such as how the market will react.
Figuring out that the market was likely to go up because of this (customers thinking the liquidations were over now so there’d be a recovery) meant you could actually predict a momentum-like effect – BTC goes up, which actually does quell the need for those pending liquidations, which means it will keep going up. Further, this changed the calculations around margin and liquidations – if we expect BTC to go up, we can be riskier in, for instance, putting on short positions in the richest futures. And understanding that fast made millions of dollars in difference for when you could get deltas off.
We were making many other decisions during this time, too – how should we be considering SPX in our models? How do we prioritize our market making obligations when liquidity has evaporated on every exchange? What is the exchanges default? Is there news that’s driving price moves we’re not seeing? And on and on. And the answers to all these questions were constantly changing! It was an impossibly complicated situation, and making the best decisions was just never going to happen – making good ones could though, and we made that happen.
No matter how good you get at making quick decisions, though, the nature of them will always leave something to be desired. When you have five seconds to make a decision and a lifetime to regret it, even a decision in the top 10bp of decisions anyone would make won’t hold up. The ability to assess that you’re good enough that if you got to redo the same situation 100 times you’d make a ton – and therefore you should trust yourself to do it the one time it’ll actually happen – makes the loss (and the opportunity for regret and scrutiny) easily worthwhile. Getting there is the hard part, but its reward is sweet.