PITBULLS: An interview with Michael Steiner (Part 1)

On July 12, 2023, Kevin D of Quantitative Traitor interviewed Michael Steiner, former derivatives trader and lead teacher for new traders at Susquehanna International Group (SIG) and creator of the trading game Pitbulls. What follows is a lightly edited transcript of the interview.

Kevin D: You started out trading derivatives for SIG in 1996 on the AMEX. What was a day in the life of Michael Steiner like back then?

Michael Steiner: The nice thing about working on a stock exchange back then was that you were subject to market hours. You had to be there for the opening bell, and you had to stay until the closing bell. I started as an assistant trader at SIG, eventually becoming a trader. My day would start out by getting to work early, checking trades for my immediate supervisors, who were different people who’ve since gone on to amazing things. Tim Reynolds, one of my first bosses, went on to Jane Street. Also Mike Jenkins, another Jane Street founder. Tim Murtha is another one of my bosses that became a colleague, and he worked at Bluefin, amongst other places.

Whatever trader I was supporting was my boss, and he was the most important person in my life. You had traders and assistant traders. Assistant traders anticipated and stayed ahead of anything the traders needed, wanted or had to do. We were called clerks also because we were another set of hands keeping track of everything. If the trader left, a very good assistant trader could still make decisions. When brokers came in with orders, an assistant trader had some authority. However, if there was a big order, usually a trader from another post came over to make sure everything was on the up-and-up.

Assistant traders would take on more responsibility, and during that time, assistant traders like myself would go to an evening class with an experienced trader and mock trade. The mock trading simulated the stock exchange floor in a pit trading environment, where you’d have to do instantaneous transactions, respond to bids and offers — akin to the movies where there’s a big crowd of people screaming and yelling prices.

KD: You mentioned pit trading. How does trading today compare to pit trading back then?

MS: Pit trading exists in very limited form today, but it’s still important to understand. All things equal, pit trading favored the largest, loudest people. You also had to be first and correct. Imagine a large competition where you’re constantly doing Bayesian probability calculations in addition to nonstop negotiations. As more information comes in, you’re reevaluating the prices you’ve computed, and you’re ready to change your mind on the fair value for something and then act on it right away, based on listening to six or more things going on at once. There’s a lot of compartmentalization and prioritization.

The rehearsals were really fun. In your mind, you imagined what you’d say or do if someone said a certain thing. You can’t cover every single contingency out there because there’s an infinite number of things that people can say, but there are little tricks to come up with shorthand ways to make decisions.

For example, consider a customer who has come in before to sell calls after a stock goes up. Typically that would be a call writer: Someone buys a share of stock, and they want to write an out-of-the-money-call — sell an out-of-the-money call for extra premium. If the stock goes up past the strike, they’re happy because they’ve made money, but if it goes up and doesn’t go past the strike, they’re also happy because they’ve kept the premium in the call that they’ve sold.

Our customer comes in, and I think he’s more likely to be a seller than to be a buyer when he quotes the market. I’ll anticipate that order flow by trying to keep the market lower because if he’s got a lot to sell, I don’t want to be the first person to start buying — there’s some adverse selection there. I might look at other calls in the marketplace that I can sell in anticipation of the seller. A natural hedge against buying calls is selling calls.

KD: When you started out at SIG, what was going on in the markets?

MS: My spot was quite limited. I was on one exchange in one city in one particular type of product, and that particular type of product was options, which is puts and calls. Even more specifically, when I first started I traded index options, which were options on indices made up of baskets of stocks. I traded options on airline, Internet and broker dealer indices. That was hot at the time — I did that for a while, and it was exciting.

I also traded equity options shortly thereafter. I eventually became the lead market maker in Cisco and IBM on the American Stock Exchange. I traded in crowds for other stocks like Brocade, DoubleClick, Pfizer. Pfizer is probably where I developed the great majority of my trading skills.

KD: What was the hardest thing to learn starting out?

MS: By far and away, I’d get in the way of my own learning, and I’d be shy and worried about making mistakes. I should worry about making mistakes when there's money on the line, but when learning — mock trading — that’s the time to open your mouth, scream, and buy and sell stuff. Once you get through the mechanics, you can really open up and start to feel more comfortable.

Consider a typical dialogue in an introductory Spanish textbook. Diego asks “¿cómo te llamas?” and then Mariela responds “me llamo Mariela”. Even if you don't know exactly what the words mean and how they fit together, simply making the noises is crucial to getting past the intimidation of a new language. I brought that into my own teaching — getting people to say the words and not be intimidated by using them incorrectly. Someone will always help you out.

Being intimidated by the language was one thing that tripped me up early. The other thing that tripped me up was seemingly zero short-term memory as to the last trades that happened. I watched the traders around me, and they could recall every trade they had made for the past five days — what was happening in the market at the time, who said what, the joke someone told in the middle of a trade. They had this accessible memory that seemed to be completely unattainable by me in the beginning. Once I relaxed, every little bit of data — the options, prices and all the players — became my friend.

It might have been Tim Reynolds who asked me, what was the trade we did two seconds ago? I said I don't know, and he responded, what did you have for breakfast? I described the breakfast to him, and he asked, was it wrapped in aluminum foil? I said yes, it was even folded over at the top. He said you remember all of this, but you can't remember the broker’s name nor the fact that we just paid two and three quarters for 50 March 75 calls from Bear Stearns? I said I'll try harder, boss — I can do it.

The truth was it wasn't about trying harder, it was about relaxing, letting it happen and nonstop practice. What am I thinking about? I’m not thinking about my sandwich this morning, I’m looking at the prices on the screen. Why is this market up there two and three quarters bid at three and an eighth? Why isn’t it something else? Watch that market for a while and and it becomes your friend — you actually know where all the markets are without even turning your head, which is really cool.

KD: You stayed at SIG for almost a decade, leaving in 2004. What were some of the things that changed over time during your tenure, and what stayed the same?

MS: I want to step back from the question of my eight years at SIG to cover one thing. My first year I was trained as an assistant trader at the finishing school, the headquarters at Bala Cynwyd. All assistant traders who were going to be trading options had to go through finishing school before they were sent back to the exchange. You would sign a three year non-compete agreement, and you became a trader. We didn’t get business cards.

I was a student and assistant trader my first year, and I was a teacher for the following seven. I traded during the day on either the Mercantile Exchange or the American Stock Exchange, and then in the evenings, I would come back and run mock trading sessions with my students and colleagues.

The question was on how the industry changed over those eight years. I was primarily an equity options trader. By trader, I mean a market maker in a crowd, or a specialist, which is a lead market maker — you’re behind a post with a crowd in front. I wouldn’t say I was managing the crowd, but I was managing the prices that the crowd and I fought over to send out to the world through my quote machine. That was the main thing I did.

During that time, we switched from fractions to decimals. While it was the most obvious thing to do, it was a disrupter. Also, more exchanges listed the same products over time. We had smaller tick size, more competitors and more interest in the products. It was very exciting, but as any market maker will tell you, the smaller the tick size, the tighter the market and the less edge there is — you need to develop other ways to find edge. I thought about relative volatility plays, and I looked at options and thought about the underlying distribution implied with so-and-so prices. Is it a good trade or not based on various metrics? I continued to grow as a trader and stayed on top of things.

However, the industry was slowing down a lot in the few years before I left. I was still doing really well, but my life priorities started to change with children. My family business started taking off, and I shifted my focus to supporting that, taking care of the children and enjoying that part of my life. But I always wanted to get back to teaching, tutoring and being in front of students. The best place for me is to be in front of a bunch of people who want to learn. That’s why I left SIG.

But to answer your question, I wouldn’t say equity options dried up necessarily, but the business wasn’t as lucrative anymore, at least from my specialist post on the floor. More of the business and profitability was in upstairs trades with much larger volume.

KD: Can you describe the process of automating various aspects of trading that went on during your tenure?

MS: It depends on how you define automation. There’s automatic trading — those kinds of decisions can be handled by some algorithm that you can write.

When I started, there was very little automation. There was a machine that you could set with certain volatilities — that’s one parameter that goes into the pricing of options. You would set volatilities, and your machine would automatically send out the quotes for your products based on those volatilities. That was it — in the late 1990s and early 2000s, it was a very pretty simple situation. Orders could come in electronically or through a broker. The quotes were sent out from a local computer where you would set your volatilities right there on the stock exchange, and you’d try to trade around those values.

There were other companies getting more and more automated. For example, Timber Hill — now Interactive Brokers — competed with SIG. They had a burg in a central location, and they sent out prices to their traders in different crowds. Their process was somewhat automated. The trader in the crowd had some discretion, but if the box told him to do something, his boss would be asking him why he hadn’t done it yet — hopefully he had a good answer. Mad respect for Timber’s model because they did really well.

Over time people added on more layers of automation. I’m sure I would recognize the basics of the models and how things work today. Confirmation would still have to come in the same way, and there are still face-to-face negotiations for certain aspects of a trade that happen on an exchange floor.

There’s an American Stock Exchange part of the New York Stock Exchange that I visited recently. Brokers still come in and ask for quotes. I hadn’t talked to some of these guys in 20+ years, and we still had the same jokes, conversations and vibe. They still made the same mistakes. It’s quite funny.

Anyways, automation has taken over a lot. Who knows when the stock exchange buildings will become museums. But they’re more than that for me — they’re stadiums.

KD: You were the lead teacher for new traders at SIG. What was that experience like?

MS: I wanted to teach to make myself a better trader. Also, as I moved up through the company and became more tied to the company’s profit, I wanted to have a hand in training and certifying the newcomers. They might not have been risking my capital, but their risk did impact my bottom line in a small way.

Building systems in general — in my case, building an educational system for assistant traders to become proficient enough to go to the finishing school — is a great way to understand how things work. Having just gone through the process myself, I knew what I needed, and talking to the assistant traders, I knew what they needed.

KD: Can you talk about distilling your experience trading during the day for your students? What skills were most worth imparting? 

MS: During the day I was trading, and there’s a lot of downtime — if the market isn’t moving, there’s not much going on. During that time, not only would I rehearse what I was going to say for the next trade, but I’d think about how to codify better the process of becoming a trader — going from an assistant trader who just knows probability, statistics and some of the language to a full-on portfolio-managing trader with hundreds of lines of position. I was pretty excited to trade during the day, but going back to the main office after work and running mock trading sessions was just as exciting.

What was a typical mock trading session like? We had tickets that you wrote your trades on. Before we started, I wrote down the stock price on the whiteboard, and then I wrote down a few months’ worth of options prices — bids and offers for puts and calls with the strike prices. As I wrote down these prices. I tried to set up some story that the participants would see happening in the prices, so that when I came in with an order, hopefully they’d see the arbitrage I was trying to setup or at least imply with the prices. Anyways, I rung the opening bell, we started mock trading and went for maybe 30 minutes to an hour. We moved the stock and options prices around and let the 10-15 assistant traders react. We gave them feedback if they made a mistake, and we asked them to explain their good decisions. We flied through as many trades as possible to give them experience, and observed who was the best — it was competitive.

The person who made the most money wasn’t always the best trader. It was the person who made the most money while taking on the least amount of risk. They better be able to explain back to me every decision they made without hesitation. If they said they bought it for two and a half, and I know they paid two and five eighths, I would stop and say, I think you lost an eighth somewhere, so let’s start over.

Also, there’s this idea in education called drill-and-kill that many are critical of. Say you have a set of multiplication flashcards that you have to answer as fast as possible. 7 x 4, 6 x 11, 14 x 2 — bootstrap up to two and three digit numbers and get super fast at multiplying them. You’re drilling in the concepts, and you’re killing people’s desire in exhausting them with tasks completely out of context. Why do I have to know my multiplication tables? The teacher says trust me, you’ll need this. When the student says this is boring, the teacher responds, just do it.

However, I think of it as drill-and-thrill. There are skills that you must get good at in order to be a trader. I told my students they had to get good at simple calculations with fractions, multiplication, division etc. There’s 40 other people in the pit who can do mental math as fast as Steiner can — you can’t be in the pit if you’re slower than me. My students would ask me for worksheets, also Excel spreadsheets generating nonstop “find the arbitrage in 15 seconds” scenarios. You have 45 seconds to do 100 calculations and can’t get any of them wrong because there's money involved.

If you come up with the wrong answer and yell it out, the market will let you know, and you’ll lose money. But sometimes legitimate mistakes are forgiven in an open outcry pit because people don’t want to win by someone else’s mistakes. Say you make a market and everybody yells at you, trying to trade with you. Someone nice may intervene to ask if you really intended to bid $75 when it’s offered at $10, and you reply you didn’t — you were talking about something else.

You have to get fast, and the best way to get fast is by drilling.

KD: With teaching new traders, what worked particularly well? What didn’t work so well?

MS: I loved the scripts I wrote — I wrote them so new people would know the language, yell this stuff out and get more comfortable. If someone had a problem, it usually stemmed from shyness and being afraid of making a mistake. I got people to read the scripts, open up and and become the self-advocate that’s necessary in an open outcry pit and trading in general.

For example, let’s suppose I’m willing to pay $12 for something, but I’m shy. I’ll half-heartedly yell I’ll pay $12, not getting enough attention. Someone next to me screams he’ll pay $11 because he didn’t hear me. Someone else responds with sold, and they sell to my neighbor at $11. I should be upset — not only because I’m willing to pay more, but because there was another person on the other end of that transaction that sold at $11 when they could’ve at $12. The seller lost a dollar, I missed the trade, and my neighbor who bought at $11 isn’t even at fault because he couldn’t hear me. The moral of the story is I need to be louder and stand up for myself. When my neighbor trades $11, I need to get right over there and say no, I was the established bid. I was willing to pay $12, and you shouldn’t be allowed to buy this for $11. I can talk to the seller and say excuse me, I’ve got a better price for you — don't do that trade. Now, there’s a lot of other stuff going on in the middle of it, so I might not have the time for that conversation. But if I’m doing my job, I never need to have that conversation.

The math never held any traders back, it was completely transparent for almost everyone. People’s personalities didn’t get in the way of them becoming good traders, it’s more a matter of comfort. I had trouble in the beginning remembering prices because it was hard to remember everything that happened. You’d have to remember someone just sold 50, they sold 50 again and then they sold another 50. If they’re coming in now and asking for the market, you can be pretty sure they’ll be selling 50, unless the market has changed, in which case you think they’ll switch to buying. The only way to remember everything is to relax enough to concentrate at the right level in order for the information to flow to you.

KD: Tell me about your game Pitbulls. How did the game come about?

MS: One of the volunteer teachers at SIG was King Yao, and he was a trader who was at SIG before me. He was running a mock trading session, and we traded on the information in a basketball game recorded on VHS tape. King put it into the VHS and said he was pretty sure nobody knew the winner of the game — we’re going to watch this basketball game and trade a straight over/under on the total number of points. I loved it. Was that player going to make a basket? What’s the score after a quarter? Will it go into overtime, in which case the over/under should go up? Maybe that was the game here — maybe King picked this game because it was an overtime match.

We were trying to buy and sell from each other. As we were doing that, it fascinated me that we were watching a game and trading on the information live. In the back of my head, I registered the need for a simulation to trade on — a race or contest that’s predictably random. I knew I needed something that was a stochastic process — the underlying principles were probabilistic — and unique, but repeatable. I was essentially looking for a random walk generator.

I started building my materials long after I left SIG. I taught in different schools and was given license to try different things with my math students. I came up with the mechanics of the cards and the currency trading hands, which must have been lifted from a concept that I saw somewhere.

In 2017, I watched a video of Sal Khan from Khan Academy — it’s my goal to gamify learning like he does. He ran a separate event from his normal online teaching, let me try to set it up. Imagine a game of Risk played by eight different colored groups. There’s a yellow person controlling all the yellow pieces, a red person controlling all the red pieces and so forth. The goal is world domination by one color — one color will win. He played this prediction market game based on Risk.

Seeing what he did, I realized that was what I had to do, which was make an eight-way race between eight different colors that people can trade on in a prediction market. That eight-way race had to be repeatable, unique every time and probabilistic in nature.

I wrote the program in Python — I still have the code somewhere. The difference between the initial version and what it eventually became is like the difference between a monkey and a fully-formed human. I took free online courses to learn programming in Python, stuff involving graphing. Sometimes I look back at the code and don’t remember writing it. But once I had this repeatable, unique and probabilistic game, everything else fell into place.

Imagine we’re watching a race between bunnies running 400 meters across a field — on a screen. Bunnies will run sideways, sometimes they’ll run backwards. One bunny might tackle another that’s getting too far ahead. A spinner with colors on it determines how the different colored bunnies will run — which ones will move and how far in which direction. One of the bunnies wins the race. How does a prediction market work? The shares of the winning bunny will settle to $100, and the shares of all the other colored bunnies settle to zero — it’s a very simple, straightforward prediction market.

That’s the essence of my game Pitbulls.

KD: How has Pitbulls evolved over time?

MS: It’s been a total trip. The mechanics: Each trader starts with a portfolio of cards, which represent the bets, and currency, which provides liquidity. Imagine each trader with a handful of cash and and different colored cards. Originally, every trader was going to start with $400 and four of each colored share for a total of 32 cards. When I started playing out this game, I’d have to rebuild that portfolio for each student every single time. It took so long — it was one of the biggest hurdles to playing the game.

Somebody mentioned that if everybody has same portfolio, there might not be a huge desire to trade. With that comment came the following things.

Every color has an equal chance of winning. Every card before the race starts — whether it be gray, yellow, pink etc. — has an equal expected value of $12.50, which is $100 divided by eight. I tell the players that, and I tell them go build a portfolio at that table over there that’s worth $400 total. For example, you can get $200 and 16 colored cards or $100 and 24 cards. After everyone has built their portfolio, I ask how much each person’s portfolio is worth, and they all respond with $400. I respond that new traders on Wall Street usually don't start by managing their own portfolio — they wind up taking someone else's. Much to their dismay, I tell the players to hand their portfolio to the person on their left — your job by the end of this game is to make your new portfolio worth as much as possible. Then we start the game. This big twist has paid off dividends.

In the beginning, people told me when I ran a session that it was a shame Steiner doesn’t ship in a box — he needles and encourages people, gets them to do stuff and jumps up and down. Part of it is performance, but mostly it’s teaching. Anyways, I came up with the idea of broker cards, which tell you what to do. For example, buy a share of green. I went from fanatic jumping up and down to quiet teacher.

I hand the broker cards out randomly to people. I hand “buy a share of green” or “sell a share of blue” to someone and tell them, do as the card says and I’ll give you $1. I drop these cards around the room, and after someone executes their card, they bring it back to me and ask for their dollar. We then have a conversation. You bought a share of yellow, what did you play? They say they paid $18, and I look up and ask, why did you pay $18? We go back-and-forth, and I say, let's see how good of a trade it is now by figuring out the price of yellow. I show them how to ask — I yell out “has yellow, what's the bid in yellow?” If someone replies they’ll pay $25, l turn to the person who just paid $18 and say good job.

Since the broker cards tell the traders what to do, I’m not necessary anymore. You just need a teacher walking around the room, giving these cards out randomly and saying, come back to me after executing what’s on the card, I'll give you $1 and then we'll talk about it.

Suppose I give someone a card that says “buy a share of red”. They’re just standing there, not knowing what to do. I come up to them and say the first thing we’ll do is come up with an idea for the price of red. It must be between $1 and $100. Then I give some ideas on how to come up with a value. I say this isn’t real money — let's suppose it’s worth $30, then we definitely want to buy it for $25. The person replies yes, and I say, let's make sure everybody knows that we're willing to pay $25. How do we do that? By telling everyone real loud — we yell out “pay $25 per red”. Sometimes the person gets to buy it, and I say, isn’t that great? You bought something for $25 that we think is worth $30 — sounds like you made $5. Let’s see if we can sell it for $35 and make $10. Sure enough, if you start negotiating, screaming and getting enough attention, sometimes you can get someone to pay $35 for something you just paid $25 for, in which case you get $10.

KD: What do the players find hard?

MS: One thing that trips people up is coming up with estimates — any estimate, doesn’t have to be perfect. Once you have an estimate, then you build a confidence interval around it — a low number and a high number. We like to say that we’re 90% confident that the value falls within that range. Some people will hesitate with coming up with that estimate.

I ask people, how many inputs does it take to come up with a good estimate? An infinite number of lines pass through one standalone point. You can draw a lot of things through two points — a parabola, a sine wave — but it’s still quite ambiguous. But three data points is pretty good to use for an estimation.

In Pitbulls, different colors have a relative value to each other. Suppose you’re trying to figure out a value for pink because the broker card tells you to do something with pink. You look at where pink is in the race relative to the other colors, and it’s right in line with blue. If you know blue’s price, that’s one data point. If you know what someone just paid for pink — they just bought pink and someone else agreed to sell at that price — that's another data point, even if you think the real price is lower or higher. You have two points, now you need a third. You ask someone, what would you pay for a share of pink? If they loudly proclaim they’ll pay $12 for pink and nobody sells to him, now you have a third data point — you know it’s worth at least $12. If you ask if anyone wants to sell pink and someone responds they’ll sell at $18, now you know it’ll be less than $18 — a fourth data point. The data points aren’t too hard to find out if you just ask questions.

The other thing is that people won’t listen to the full market. Instead of making the market compete for their love, they sometimes take the market for granted. For example, say I’m willing to sell at $12 and someone says they’ll pay $12. Why didn’t they bid $10 first to see if I’d sell at $10? Was there nobody else around? Could they ask somebody else? There are all these things that people can use to their advantage. Steiner, you’re willing to sell at $12? I’ll pay $8. Can you do any better than $12? That might not be the best they can do, but the negotiation adds to the whole process. Let’s say the two of us are very loud with $8 bid at $12. We get someone’s attention, and then someone says wait a minute, Kevin wants to pay $8? I’ll sell at $8. Then you’re happy — you didn’t have to pay $12, you saved $4.

That’s another thing that’s exciting about my game. It teaches negotiation, price discovery, supply and demand — things like that.

KD: Do you plan on expanding Pitbulls to options? 

MS: Right now, all the colors trade like a prediction market. One settles to $100, the others settle to zero. We can change that to first place settling to $50, second to $30, third to $20 — that happens to add up to $100. The person running the session could set the parameters at the beginning of the game to anything — first place settling to five, second to one hundred, third to a million. Why someone would want to do that, I don’t know.

Options are a bit different. We’d use them as just another thing to trade. It would only be on the screen, it wouldn’t be in the open outcry pit. Options are surprisingly not tricky to code into what I already have, they just require a different screen. The mechanics are easy: You have all the colors on your screen. If you want to trade options on purple, you click on purple, and now your entire screen fills up with option prices for purple. It could be options that settle at the end of the next round or cash-settle at the end of the game, depending on the parameters set up for the game. Options will be either cash-settled or stock-settled. ■

TO BE CONTINUED…