by Teddy Wayne
I am uncertain if she hears me, because then my father is on the telephone. I ask him if he has heard the news about Iran yet. He has not, and I explain the situation and tell him that the news said a terrorist group in Iran has claimed responsibility. “You should not believe everything you hear on the news in the U.S.,” he says.
“Why do you say that? Do you think they are lying about the attack?”
“No,” he says. “But they call them a terrorist group. You do not know what this group stands for. They do not define themselves as terrorists. To them, the French government is a terrorist group.”
“Yes, but the French government is not bombing civilians,” I say.
“No, they have simply colonized other countries for centuries and oppress Algerians in their own country.”
“Where are you getting these ideas from?” I ask.
“Just because I labor in a store does not mean I do not read, Karim.”
“I did not say you do not read,” I say. “I asked where you are getting these ideas.”
“From newspapers that are not about money and computers and are not published in the U.S.” Then he adds, “You should read one sometime.”
The sounds of people celebrating and cars honking in the street because of the Mets victory rise all the way up to my apartment.
“I have to go to sleep for work tomorrow,” I say.
We disconnect, and I consume my final dumpling, but its skin is now cold and has little flavor and I do not feel like microwaving it. The cars continue honking outside, and I open my window and lean my head out and shout for them to be quiet in Arabic, but of course it achieves nothing.
Alpha Phi = a social group for university females
claim responsibility = take responsibility for an event others view as a negative but that you are boastful of
date-rapist = a man who forces a female he knows into sexual activity
number-crunch = make intensive calculations
shoot an email = send an email, especially about business
to boot = in addition
vapid = non-stimulating
JOURNAL DATE RECORDED: OCTOBER 19
On Monday at the office I am even more quiet than average, which is nearly mute because on average I converse exclusively when someone first consults with me or if I have an urgent query.
During lunch, Dan reads The New York Times on the computer while he eats the Indian chicken tikka masala he orders daily and Jefferson scans baseball statistics.
“You hear about this French embassy bombing in Iran?” Dan asks. “Times says a splinter terrorist cell took responsibility and vows more attacks. This shit’s not even front-page news, that’s how common it is. Why don’t they just incinerate their whole uncivilized backwater country and jack up gas prices even more?” He looks quickly at me. “No offense, Karim.”
“I am not from Iran,” I say.
“I know,” he says. “I didn’t mean anything by it.” Then he asks Jefferson about the fantasy baseball production of a player named Yoshii. Jefferson owns all the Japanese players.
In a few minutes I receive an email:
Sender: Rebecca A. Goldman
Recipient: Karim Issar
Date: Mon, 18 Oct 1999 12:26:18
Subject: Dan is a…
…jackass. (Not front-page news, either.)
After I research the word “jackass,” I smile at her. She reciprocates, and I feel enhanced, as we have had restricted conversation since our coffee meeting.
And then I have another mental image of the stars at night.
I research today’s crude oil futures ASAP. They have risen 77 cents. That is expected because of the news.
I use the search engine on The New York Times and input the phrase “Middle East.” It lists all the articles from the last 14 days about the Middle East. Of course, it is not always about terrorist attacks or war, e.g., the articles typically discuss government leaders meeting or business negotiations or other events that are nonviolent. I note which days the phrase appears most frequently, and how many times it appears. Then I correlate those days to the crude oil futures prices of that day or the next day.
Although I am not making intensive calculations, I think I see a correlation between how frequently The New York Times discusses the Middle East and the fluctuations of oil futures.
I input the names of specific countries, e.g., Saudi Arabia, Iran, United Arab Emirates, Kuwait, Iraq, and Qatar, which only produces approximately 0.5 million barrels per day, but that is a great amount for its small size.
The correlations appear stronger.
Schrub has a subscription to a service that scans all the major U.S. newspapers. I log in and input the country names again for the last 14 days.
The correlations appear very strong.
“Karim, want to do me a major-league favor?” Jefferson asks.
Whenever he asks if I want to do him a major-league favor it means he wants me to repair a glitch that he has caused himself.
I close up the windows about oil prices. I should not be laboring independently on company time anyway.
The remainder of the day I cannot contain my stimulation. Before Dan and Jefferson leave, they converse about a nightclub they are going to that is providing free tequila to promote its launch. This time I am careless.
Then Rebecca leaves, and I am free to labor on my project.
I can now utilize spreadsheets and other programs to determine if the correlations are accurate, and broaden my newspaper search to 60 days for additional coverage.
1. First, I decide that not every expression of a country’s name is equally integral, e.g., “Iran AND bomb” has more influence than “Qatar AND diplomatic talks.”
2. So I begin by employing a boosting algorithm that weights specific words, which I perform by reverse-correlation, so that I see what days the oil prices moved most sharply and then determine what keywords ignited their movement. “Terrorism” and “terrorist” are heavily weighted, of course, and so are “war” and “attack” and “gunfire” and similar terms. Words like “unrest” and “protest” and “demonstration” are in another class, and words like “treaty” and “diplomatic talks” are also in a different class. Also integral is that some words are important exclusively in pairs or in longer phrases, e.g., “white” and “house” mean little independently, but “White House” is critical. Words have elastic meanings from their context.
A. When a word or phrase proves that it has high predictive abilities, the algorithm boosts its weight.
B. The names of the countries and cities that produce more oil and are more volatile also have different weights.
C. More recent articles weigh more.
D. Although logic predicts certain actions, such as a terrorist attack, always raise the price, this is not true, as it depends on a constellation of variables, and in a few instances an attack actually lowers prices.
3. But because the algorithm is automated and it analyzes every word in an article, it also selects many words that I think no one else would pay attention to, such as “bitter” and “weary” and “resigned,” as in this sentence: “The Prime Minister, after a round of bitter questioning, appeared weary and resigned.” I think these kinds of words can in fact be more important because:
A. By the time a bombing has occurred, e.g., everyone knows about it and they can predict what will happen to oil prices and they act accordingly.
B. But fewer people read about a politician appearing weary and resigned after receiving bitter questions.
C. A few people do read it, however, and they begin acting in a predictable way; then a few more people follow their lead, and more and more, until it becomes as if everyone did read it, even though they did not.
4. I can aid the automated algorithm by examining articles manually, and as someone whose native language is not English, I must pay closer attention to the words to produce log
ic from them, and sometimes I observe things others do not about English.
5. Therefore, if I can collect enough data like this, I can gain a major advantage over others who are merely using obvious data that are front-page news.
Because this central idea is truly an analog to scanning unobserved partitions of a Pollock painting, I am able to piggyback it onto my previous program. I also link it to the newspaper search engine. It is taxing (although it takes less time because of the previous program), but it is the class of labor I enjoy.
The nighttime janitor cleans as I program, and when I give her my trash bin, it is the solitary time I look away from my monitor.
In fact, only when I am nearly finished and my cursor is on the word “casualties” do I evaluate the big picture of what I am creating. When violence occurs, especially in the Middle East, my program will attempt to leverage it for financial gain. But this violence will happen with or without my program. Therefore, by making money, the program produces at least some positives from a very negative situation. It turns the violence into a zero-sum game, because the money and violence cancel each other out, instead of producing exclusively a negative game.
I decide to complete the program.
I finally stop near 3:30 a.m., and I notice I have been alone for several hours and I have not eaten dinner. I am not even that hungry, but I purchase an apple from a vending machine. Typically when I am programming like this at home, Zahira forces me to consume food because I forget.
My program is finally functional, although I do not know how accurate it is until it sifts historical data. I direct it to use data from the last six months and to make oil futures predictions on each day as if it were truly that day.
It will have to number-crunch overnight, so I leave my computer on but turn my monitor off and open a spreadsheet so others will not identify the true labor of the computer, and then I go home and wait until the morning to discover if my program is successful.
backwater = an unimportant or unsophisticated location
incinerate = burn down
jack up = inflate prices
jackass = stupid person; Dan
major-league favor = significant favor
JOURNAL DATE RECORDED: OCTOBER 20
I do not sleep the entire night. At 6:00 a.m. I get out of bed and decide to go to a mosque, as that is a profitable destination when my brain is overstimulated, and my program at work will not be ready until probably 8:30 a.m. The other mosques near my office and apartment are adequate, but it is time for me to visit the chief mosque in New York.
I take the subway to the Islamic Cultural Center on the Upper East Side. It is as attractive as I have read, with a dome and Turkish architecture that resembles the Hagia Sophia with sharper lines.
The prayer hall has tall glass windows and pure white walls and hanging cords creating a circle with lights attached to the end, like an inverted birthday cake with candles. There are no columns inside, and the dome is simple but elegant, and the patterns on the carpet have a complex repeating pattern. This is what I want my programming to have: mathematical precision that is also beautiful. As a parallel, Jackson Pollock’s paintings are beautiful even though they are not precise, but by being so imprecise they are also in a way precise.
Half of the men are black, and in one corner men in robes read the Koran together. I consider partnering with them, but I want to be alone with my prayers now.
Once I begin praying I do forget about my program. It is as if saying words to Allah mutes all the calculations and ideas that are making noise inside my head and transports me to the spiritual world that is non-numbered, and performing the actions I have performed thousands of times reminds me of my body, which is also non-numbered, which is why I also enjoy racquetball.
After an hour I am recharged. On the subway to work I observe all the businessmen reading The Wall Street Journal who are trying to find ways to decipher the stock market. Maybe I have done so with oil futures.
Jefferson said Schrub’s programs yield 3–4% above market returns for yearly returns. So I hope to gain 5% above market returns on oil futures based on historical data, with minimal risk. This will mean a minimal average daily return above market, approximately 0.02%, but it is like the way a child becomes taller: You do not observe daily growth.
My podmates are not in yet. My hands vibrate slightly as I approach my desk and power on my monitor.
I close the spreadsheet window above my program. There are many numbers on the monitor, and I still need to do some calculations to receive the final results.
For overnight predictions, which means the user trades immediately in the morning and trades again at the end of the workday, my program correctly predicts the converted price of oil futures on that day within a 12% error, e.g., if oil rises $1, then 68% of the time my program predicts that prices will rise between 88 cents and $1.12. On the historical data, this means its average daily profit on oil futures is 1.1%.
There must be an error, so I reenter the calculations.
It is again 1.1%.
I try to sit very still although I am vibrating even more. I will not say anything about this yet before I know it truly works. I cannot risk humiliating myself again.
Dan enters the pod, so I reopen the spreadsheet window in case he sees the program, although he will not understand what it is. He turns on his monitor. His computer has been downloading music overnight without paying for it. He does that frequently, which is not only illegal normally but is even more illegal to do at work. The current song is titled “Mashup—Livin’ La Beasta Burden (Livin’ La Vida Loca vs. Beast of Burden).”
An error range of 12% is impressive, but I must refine the program to gain even higher average returns and minimize risk. I cannot resist, and I start recoding a section.
“What are you slaving away at?” Rebecca asks.
I am so focused on my work that I did not hear her enter and I left the program observable. Rebecca might understand it more than Dan and Jefferson do.
I consider revealing my project to her. But there is no way to do it quietly without Dan and Jefferson hearing, and they would understand the idea when translated to English.
In addition I am afraid she will again think I am interested exclusively in money.
“It is only some number-crunching,” I say, and close the windows.
Later in the day Rebecca strikes her hand on her keyboard. “Fucking machine,” she says quietly.
“Are you having a technical issue?” I ask.
“Yes, I’m having a technical issue. How’d you guess, Karim?” she says. Then she adds, “Ignore me. It’s not your fault. I’m just having a hard time dealing right now.”
She explains that there is a virus in a spreadsheet she has been working on for several hours which prevents her from accessing it.
“I have some experience with viruses,” I say.
First I quarantine the document in our pod’s recycle bin, which is stored on a separate drive, so that it cannot impact any other important documents. It is a class of virus I am familiar with, so I approximately know how to proceed.
But as I fix it, I notice a document in the recycle bin: “market prediction.doc.” It must be Jefferson’s refined version of my first program proposal. I open it.
The document looks similar to what I gave Jefferson although with slightly enhanced language, but the end does not include my name, as I originally wrote. In fact, it does not include anyone’s name.
I define two possible theories: (1) At Schrub New York it is considered unprofessional to include your name at the bottom of a proposal, and Jefferson told the higher-up (whose name I read is George Ray) that I was the programmer, or (2) Jefferson claimed responsibility for my program and pretended it was his.
I decide it is the first, as ultimately Jefferson could not claim responsibility because he would have to come to me for the program, unless he was skilled enough to decipher and recreate it from my proposal, but I do not th
ink he possesses sufficient skills.
After several minutes I heal the virus and return it to Rebecca. Healing a virus is a delightful feeling, especially when you do it for someone else, because they previously thought their file was corrupt and lost but now it is healthy and accessible.
“It’s funny how you only seem to lose data that you really need, and not, like, idiotic joke emails your mom forwards you,” she says. “I owe you big-time.”
“You do not owe me anything,” I say. “We are coworkers, and coworkers are parallel to family members in that you do not incur debts.”
She looks at me with a strange expression. Then she says, “Okay. You’re a lifesaver, though. Thanks.” She contacts my shoulder as she says this, and then she retracts her hand as if she touched a hot stove. This is the first time she has contacted me.
I want to tell her that I do not 100% agree with all rules of Islam, and that some of them are in fact impossible to fulfill while in a modern workplace, e.g., technically Rebecca and I are not permitted to be alone, and the only conversation we are allowed to have must be humorless (which is not difficult for me, because I am always humorless, but Rebecca enjoys producing jokes).
But I do not know how to explain this without making us both more uncomfortable, so I merely say, “You are welcome.”
At home I refine my program until it attains 8% error range. On historical data, it averages daily profits of approximately 1.3%.
This does not sound like much, but over 20 business days 1.3% daily profits means that investing $1,000 in a futures contract on the first day, then using that new money to invest in another futures contract the next day, will yield by the end of the month $1,295. Of course, you are not guaranteed to make money each day, but this is a potential outcome: 29.5% monthly profits.