To Serve Traders

[TL/DR: Proof has just released a pretrade analysis tool. You can skip the below and find the tool here, but be warned — this post is about the dangers of skipping the details and using something without understanding its nuances ;) The whitepaper detailing the research efforts behind the tool can be found here.]

Is predicting the future a gift or a curse? Across time this question has inspired classic works, from the Greek myths of Oedipus and Cassandra to the Back to the Future series. My father, however, rarely grapples with its complexities. He loves to spoil twist endings of movies. Whatever I was watching, without fail, he would march confidently into the room, watch for a few minutes, and announce who the killer was. “And I haven’t even seen it before!” he would declare proudly. And then he would leave. The worst part was, he was always right.

There was always one exception to my father’s compulsion to predict. The Twilight Zone episodes were sacred. We would watch them back to back in TV marathons every New Years, and he would hold his tongue while each episode progressed from a disorienting premise to a surprise (or at least insightful) conclusion. He would content himself with a satisfied “pretty cool, right?” at the end. I would nod. Yes, it was pretty cool. I was obsessed with The Twilight Zone, a dimension where the usual narrative rules were bent and twisted to new effect. A place where judgement could be distorted, and the right balance between hubris and humility could get harder to maintain. [For the kids out there who may not be familiar, The Twilight Zone was your parents’ version of Black Mirror, though it sometimes had happy endings.]

This dimension seems to me to be the proper home for the task of pretrade analysis, an inherently arrogant and yet necessary attempt to predict the future.

“What will happen if I make this trade?” Our trader character wonders aloud. He is standing in his Italian loafers in the center of a room packed with monitors, all swimming with numbers and charts. “If I start buying a lot of MSFT stock, how much will the price go up by the time I’m done?”

“I can tell you,” a woman’s voice responds. Startled, he turns around to find her standing right in front of him. She is wearing a stylish suit and flashes him a knowing smile. In her hands, she is cradling a mysterious device.

“How did you…”

“I have something for you,” she continues, holding out the device. “It’s a pretrade analysis tool from Proof. You simply tell it how much you plan to trade, in which direction and in which symbol, and over what time period (one day or one hour). It shows you the predicted distribution of price changes that are likely to occur as a result.”

The trader takes the device from her outstretched hands. “What’s the catch?”

“Oh there are many catches,” she says, still smiling eerily.

Just then the trader’s boss appears in the doorway. “Traderson!” he barks. “You better get off your ass and get moving on those portfolios!” Just as quickly he is gone. It’s not clear if he noticed the woman.

“Ok,” Trey Traderson says, looking at the pretrade analysis tool. “I’ll give it a shot. I gotta get back to work.”

He waives her away, but she is already gone.

Rod Serling appears. “Meet Trey Traderson,” he invites us. “A man living in the twilight of a world of suits, deals, golf, and cigars. A man who takes pride in his job and wants to do it well. A man who has been offered a tool that can help him do just that. But it’s a tool that will take him straight into… The Twilight Zone.”

After the commercial break, Trey looks at his fourteenth monitor for the trade he needs to initiate next. Global Airlines, sell 1.5 million shares. He hesitates for a moment. He stares at the mysterious pretrade analysis tool. “Global Airlines, sell, 1.5 million shares” he mutters as he keys it into the tool.

“1.5 million shares of Global Airlines is 5% of its ADV,” the tool says in a confident, robotic voice. “Global Airlines is an active, volatile symbol. Selling 5% of ADV in an active, volatile symbol in one day results in a median price impact of -0.1%. This is based on a sample size of 14,000.”

“Seems reasonable,” Trey Traderson says. He goes over to his fifteenth monitor to enter the trade. It is 9:31 am, so it will have nearly the whole trading day to execute. He notes the opening price of Global Airlines, which was $32.05.

Over the next few hours, he takes a few phone calls. He reads the news. He watches the markets on his monitors. He leaves to meet his friend Dean Dealington for lunch.

“You ever heard of this Pretrade Analysis tool from Proof?” Trey asks Dean between bites of his steak sandwich.

“Nah,” says Dean. “Is it any good? Most pretrade tools are pretty crap. We just use them to cover our asses when something goes south. ‘Well, did you check the pretrade tool, Dean?’ Yes, Sandy. I fucking checked it.”

Trey nods. “Yeah. This one’s probably the same I guess. We’ll see.”

It’s the end of the trading day and Trey is at his desk. The closing price of Global Airlines flashes across his sixteenth monitor: $32.01, down by about 0.12% from the open. He looks at SPX on his third monitor. Flat. “Huh,” he mutters.

One month later, Trey Traderson is riding high on a wave of good fortune. He lounges back in his chair, contentedly surveying his many monitors. His boss appears in the doorway.

“Good work lately, Traderson!” he barks. “Keep it up!”

Traderson interlocks his fingers and stretches his arms out in front of him. He looks at his fourteenth monitor for his next assigned trade. “Whipple Inc., buy, 50,000 shares,” he reads. He keys it into the tool.

“50,000 shares of Whipple Inc. is 10% of its ADV,” the tools responds in its clear, robotic voice. “Whipple Inc. is an active, stable symbol. Buying 10% of ADV in an active, stable symbol in one day results in a median price impact of +0.3%. This is based on a sample size of 2,700.”

“Hmm…” Trey muses. “That’s a little high. What I were to buy half of it today and half of it tomorrow?”

As if it heard him, the tool answers: “Buying 5% of ADV in an active, stable symbol in one day results in a median price impact of +0.0%. This is based on a sample size of 6,600.”

“Well I’d be a fool not to take that,” Trey says. He enters the buy order for 25,000 shares and notes the opening price of Whipple Inc.: $109.14.

He looks again at his fourteenth monitor. “Cayuga Productions, buy, 100,000 shares,” he reads. He keys it into the tool.

“100,000 shares of Cayuga Productions is 31% of its ADV,” the tool says. “Cayuga Productions is an active, highly volatile symbol. Buying 31% of ADV in an active, volatile symbol in one day results in a median price impact of +0.9%. This is based on a sample size of 156.”

“Sounds like a small price to pay for a big move!” Trey Traderson says. He enters the trade. He notes the opening price of Cayuga Productions: $126.40. Feeling confident, he leaves the office and goes out to enjoy a day on the golf course. He returns to his desk at 4:01 pm. He pours himself a glass of bourbon and sits down. He looks at his seventeenth monitor to check the closing price of Whipple Inc.: $109.14. He takes a full, satisfied sip. He looks at his eighteenth monitor to check the closing price of Cayuga Productions: $200.00!

Trey spits his bourbon across the desk. “WHAT THE??” he yells. “How can this be??” Frantically, he checks SPX. Flat for the day. “Fuck,” he says.

His boss appears in the doorway. “Traderson!” he barks. “My office!”

Through the frosted glass of the boss’ office door, we hear muffled shouting. We watch as the shadow of Trey’s boss shakes a finger menacingly at Trey’s shadow, which shrinks back in response.

Trey returns to his desk. His eyes settle angrily on the Proof pretrade analysis device. “You cursed thing!” he yells, tossing it into the trash. “I’ll never listen to you again!” Exhausted, he sinks into his chair.

Another month later, Trey is pacing around his office nervously. He looks at his monitors, overwhelmed. He types into his keyboard, his keystrokes haphazard and lacking confidence. His boss appears in the doorway. “Traderson,” he says gently. “My office.”

Inside the frosted glass this time, we watch the boss’s face contort into an uncomfortable expression of actual remorse. “I think I’ve been too hard on you, Traderson,” he says.

Trey stares back at him, stunned.

“This is a stressful job,” he continues. “An impossible job, nearly. No one can predict the future with certainty. We all have to do the best that we can with the information we have. And let’s face it, the information we have from our own trades is just a tiny drop in a big ocean of market turmoil. Your trades go well, I reward you. Your trades go poorly, I punish you. But what’s the point? Maybe none of it is really your doing anyway. Maybe it’s basically luck of the draw.”

He pauses to take a puff of his cigar. “We’re living on borrowed time anyway, Traderson,” he says. “People like us? Soon enough we’ll be replaced by machines.”

Trey thinks of the Proof pretrade analysis device, buried in a landfill somewhere by now. He feels a pang of regret about his improper disposal of electronics. Suddenly he realizes something important. “The sample size!” he shouts. His boss looks at him in confusion.

“We’re not obsolete yet,” Trey explains. “The machines don’t know everything. But they know what they don’t know!”

He races out of the room and back to his desk. He nearly bumps into the woman, wearing the same stylish suit. She holds out another Proof pretrade analysis device. “Looking for this?” she asks.

Trey takes the device and sets it on his desk. He pulls up an extra chair and gestures for her to sit down.

“How does it work?” he says.

She smiles. “Of course! You know, I’ve been waiting a long time for someone to ask me that.”

She sits back her chair, settling in for a long explanation. Her eyes catch onto the bottle of bourbon on the shelf above the desk. “Could I have a glass of that?” she asks.

“Certainly,” Trey says. He pours her a glass. She takes a sip.

“Good stuff,” she says. “Now, you’re wondering how the tool could possibly guess what’s going to happen to a stock price as you trade, when you haven’t started trading yet, right?” she asked. “How can it predict the future?”

Trey nods.

“The thing is,” she continues. “You may not have traded yet. But many people like you, on many similar days, have traded in similar ways before. The tool looks through recent history, and tries to find examples of trading activities that match up with yours in terms of size, side, and time horizon, for similar symbols. And it looks at what happened to prices for all of those examples.”

Trey squirms a bit in his chair. “You’re tracking us?” he asks. “You’re somehow tracking traders and keeping records of what we each do? Nobody’s supposed to be able to…” He imagines a room somewhere with even more monitors, filled to the brim with monitors, where each one is showing a trader like himself, pacing around and entering trades like a lab rat in a science experiment.

The woman laughs. “Relax,” she says. “We’re not tracking anyone. We’re only looking at TAQ market data that has the prices, sizes, and times of trades. No identity information.”

Trey relaxes a little. “But then how do you find examples of trades that are similar to the ones I do?”

“It’s not an exact science,” the woman admits. “We try to infer examples by matching up pairs of trading days that look mostly the same except that one has added activity that moves the market in a particular direction. Our methods right now leave a lot of room for improvement. We’re working on it.” She flashes him a humble smile. “Like I tried to explain initially, there are a lot of caveats.”

A heavy whitepaper appears in her hand. She passes it to him. “Here,” she says. “You can read all about it in this.”

Trey drops the dense paper on his desk. “What’s the quick version? It has to do with sample size, right?”

The woman shakes her head disapprovingly. “Everyone these days,” she says sadly, “wants the quick version.” She gestures at the many monitors. “Machines were supposed to free our time,” she laments, “not steal it.”

She takes another sip of bourbon. “Fine. The short version. When we go looking for examples of bigger and bigger trades, we find fewer and fewer of them. The ones we do find may be heavily biased. For instance, perhaps traders who set out to buy 20% of the ADV in some symbol often see the price move up and back off, so the times they actually complete the trade are when conditions are atypical, and hence our data is misleading. Also, the smaller the sample size, the more the noise of the market drowns out any meaningful pattern. You know, the tool shows you how noisy the data is: you can actually look at the distribution of the price changes we see for the examples we’ve gathered.”

Trey looks down at the floor. “I don’t really look at the graphs,” he admits. “I just go by the median price impact numbers.”

She shrugs. “You gotta go by something,” she says. “If you want, you can see what we go by if you push the ‘what would Proof do?’ button. But our goals and constraints might be quite different from yours, so you should take that with a grain of salt.”

Trey nods. “I remember hearing a low sample size for my terrible trade,” he says. “I didn’t take it seriously. I guess I should have.”

She smiles at him. “It’s never too late to learn something, Trey Traderson,” she says. “We make mistakes all of the time. And we keep trying to learn from them. I might learn, for instance, that our method of finding examples of trades that look like yours doesn’t work very well. And if I do learn that, I’ll try to fix it.”

She gestures to the pretrade analysis device. “It’s not magic,” she says. “It really can’t predict the future. But it can help us learn a little from the past.”

She finishes her bourbon and puts down the glass. “Thanks for the bourbon,” she says. Trey nods.

His boss appears in the doorway. “Traderson!” he barks. “What are you still doing here? You should go home for the evening.”

The woman is gone. Trey looks down at the whitepaper on his desk. He picks it up.

“I’m reading,” he says.

“Trey Traderson,” Rod Serling narrates. “A relic of a passing era who just might find his place in the future, if he can stand to learn a few new tricks. A man who wanted better answers, but instead found better questions — here in the Twilight Zone.”