Disclaimer: All views expressed are our own. It’s purposely written in an entertaining way so don’t take it too seriously
One month after starting on the desk, I asked my mentor what is the most important concept in trading for a junior person to learn. I was expecting something mathematical – maybe expected value or variance or risk, but he gave me a rather peculiar answer. He told me that in both trading and in life, to start thinking about adverse selection… because it’s everywhere.
Adverse selection? What the heck is that? As a math student, I had heard about selection bias before, but what was adverse selection? Is it even related? Economists would define adverse selection as any situation where asymmetric information is exploited. For example, if I’m looking to buy car rental insurance because I know I’ll be driving in some rough terrain, I wouldn’t tell that to the car rental agency. But they’re smart, so this is probably why it’s more expensive to rent cars near national parks – you never know what to expect with bison (true story).
There’s an extension to the classic definition that I like to think about, and it’s weirdly relevant to trading. I like to think of it as, in this world of imperfect and unknown information, given that I observed something happen, can I make better assumptions about those unknowns that could have influenced the outcome?
Going back to the car example, if I was the car rental agency underwriting the price of optional car insurance, is my fair value the average damage to a car across all renters? Or should it be higher than the average because if someone is choosing to purchase insurance, they likely intend to take the car for a wild spin? For the math people out there, it’s just like Bayesian probability.
You have some prior belief about the world, and given some event occurred, should you update your posterior distribution? Thinking about the world as a chain of conditional probabilities, if the last event in the chain occurred, does that increase the likelihood of the events earlier in the chain?
Just as my mentor forewarned, this concept of adverse selection started creeping up everywhere, and as I continued down my career, it became so obvious in a trading context. If I see a really good trade based on my pricing but no one else does, does that increase the likelihood my pricing is wrong? It got so extreme that every time I noticed some strange phenomenon, my first instinct was to ask myself if there was some adverse selection going on. Let me give you some concrete trading examples to drill down the concept.
There’s a lot of reasons why your phenomenal backtest might not perform as well during live trading. Assumptions were wrong, path dependency, overfitting, the list goes on. One underrated element is adverse selection. This is HFT. Even a strategy with a 51% success rate traded a million times will make money. That’s just the law of large numbers. But in practice, during the times that your signal is correct, everyone else is thinking the same thing and maybe using the same signal, so there’s competition in getting the trade.
You might end up executing at a worse price or not at all. Conversely, during the moments where your signal is wrong, you might be the only one trading so you’ve overfilled! Simply speaking, your fill percentage is negatively correlated with signal correctness. Even if you’ve accounted for fill percentage in your backtest, it’s really hard to account for adverse selection so you’ll need to dampen your P&L estimate even more.
Some of the most intuitive concepts in trading can be tied back to adverse selection. Counterparty is one of them. There’s a bunch of different types of market participants. Some major ones include: retail customers, market makers, and banks executing on behalf of funds. Intuitively, you can probably guess that retail customers are the most benign, meaning if you trade against them, price is probably not going to move against you. Market makers can be very toxic (opposite of benign) in shorter time frames while banks can be very toxic in longer time frames.
Given this toxicity difference, as a market maker, you want to interact as much as possible with retail customers and as little as possible with other market makers and banks. This toxicity really boils down to asymmetric information, where if they traded against you, they likely know something you don’t. You probably want to hedge that position ASAP or even flip and trade in the same direction as them.
Similarly, if you aggressively trade into a market maker or a bank, you might second guess yourself and ask “is my valuation correct?” This is especially true if you hit into a market maker that you know is very fast and is likely to have pulled their quotes on the same signal that resulted in you hitting them. Maybe your signal is wrong and you should stay out of the market for a bit while your valuation stabilizes.
As you can see, counterparty information is very important in the lens of adverse selection. Some markets like equity options give you counterparty information, while other markets like equities don’t tell you directly but there’s a lot of contextual order book clues (exercise left to the reader). Then there’s very toxic markets like futures where they don’t tell you the counterparty but do they really need to? Think about who actually trades futures – I can promise you it aint mom and pops on the other side.
Ever wondered why it’s so important to be fast? Some firms have made fortunes on speed. There’s the obvious arbitrage trades that everyone is going for, but HFT is more than just arbitrage. Generally, if you’re the fastest, it means you get first dibs on doing a trade. A Robinhood customer just posted a limit order way through your theo? Hit that. AAPL stock just moved but the in-the-money (ITM) options quotes weren’t pulled? Hit that. There’s some very straightforward trades that by being the fastest, you pick up the low hanging fruit.
There’s a bit more nuance here. If you’re the slowest, and one of these “obvious trades” still exists, what does that tell you? It means all your competitors deliberately chose to not do this trade, so clearly there’s something they see that you don’t. Maybe it’s not an “obvious trade” after all, so if you do the trade, chances are it’s going to move against you. That’s textbook adverse selection.
The same applies for quotes. Every market maker has a valuation that they’re quoting around, and this valuation moves a lot during the day. As an old price level gets chipped away on the bid and a new level is about to form on the offer, there’s a race to be there first. Why? Because being first in a price-time exchange gives you priority when someone comes to trade against you.
If you’re slow and everyone already quoted that level ahead of you, your quote will only execute when everyone else pulls (they no longer like that level), or a large order cleared the entire level. These large orders are analogous to informed flow, while those small orders that the guys ahead of you get filled on are analogous to retail flow. Needless to say, you know which one is more toxic.
There’s this hilarious meme that I saw that I can’t find, but because we care so much about your education, I’ll recreate it for you.
If you have a brilliant trade, the way to execute it isn’t by posting limit orders to “save on the spread”. Just full send it. The reason should be obvious to you by now but if you’re posting limit orders, the only time you’ll trade is when the price moves against you. The only reason to be posting limit orders on the BBO (best bid and offer) is if you’re a professional market maker with the infrastructure to back it up.
This is the same reason you’ll never get a favorable trade where some benign retail customer trades into you, because you’ll never have priority over a market maker unless they pull their quotes (which means you’re probably about to get traded through). If you’re just trading at home, don’t sweat about the spread.
The only times I ever use limit orders when I personally trade is when I want to forget about the trade so I just post something deep at a price level I like. I know that when it fills, it’s a bad trade for me in the short term, but my time horizon is longer than that.
Ok this one isn’t trading related but I felt compelled to mention it. As your fellow math nerd, I never used tinder in college, but when my friend first mentioned it to me, I was like “wow, did you really just say adverse selection and tinder in the same sentence?” He was trying to date for real, but kept complaining that these girls on dating apps sucked, and he’d rather find love more organically, whatever that meant. When I asked him why he felt that, he said think about it, the most desirable girls are already in relationships, or have no problem meeting people in real life, so they wouldn’t be on dating apps. Crude I know, but we were young.
Please don’t cancel us! I ended up finding an amazing partner through dating apps so I obviously don’t believe this. Tim Urban has a good argument for why online dating is great if you’re interested.
Why did I go to great lengths discussing this concept? Because now after 5 years, I agree with my mentor that adverse selection is one of the most fascinating and ubiquitous concepts in trading. It’s also really fun to think about. This is classic game theory, and wasn’t that a big reason I wanted to be a trader in the first place? To play at this high stakes strategy game? I hope I gave you something to think about.
The practical way to apply this if you’re a student is to think about adverse selection as it occurs in your daily life. We could go on and on with examples but I’m sure now that I brought it up, you’ll encounter it everywhere. If you have an interesting example, feel free to email us! Next, when you do your trading interviews and you get a challenging case study or problem solving question, ask yourself if adverse selection applies. It might just be more often than you think and you’ll impress the hell out of your interviewer. Good luck!
