From Pits to Pixels — 30 Years of Markets

It’s a big number. 30 years. That’s about how long I’ve been involved in the markets.

I started running paper in the meats at Chicago Mercantile Exchange (CME) in January of 1991. Now, nearly 30 years later, I work for one of the leading software providers in the electronic trading space, Trading Technologies. There were stops along the way working in the back-office of a few banks, trading for a Commodity Trading Advisor (CTA), trading my own account full-time and working for an investment research company.

Running paper is the pits is as basic of a position that you could have on the trading floors. Desk clerks answered calls from traders and brokers, wrote the order down on a paper ticket, time stamped the ticket, and handed it to a runner. Runners would take the paper order to a clerk or broker in the appropriate pit. Desk clerks could also submit orders through hand signals, AKA arb, and that practice was more popular in the bigger pits.

After a few weeks of running paper, I talked my way into a brokerage clerk position in the Eurodollar pit. As a brokerage clerk, the role generally involved quoting the markets to customers who ring the pit. Quotes came from your brokers in the pit and those brokers would execute the orders that clerks received from customers. Quotes also came from our heads. We would calculate tradable markets based off current market conditions across an array of contracts and imply where the market should trade. This process of retrieving quotes and farming orders could be a glaringly clear example of turnaround time as clerks would often have to turn their heads away from their customers to speak with brokers and relay order information. Often, we’d literally turn around and yell at the broker to get their attention and relay the order information.

Clerks were the order servers of their day while brokers were human matching engines. The brokerage firm I worked for filled orders in spreads. A simple spread would be selling contracts in the front month, perhaps March, and buying contracts in a deferred month, perhaps June. Generally, the spread prices were easy to calculate as it was nothing more than the price price differential between the two markets. You’d factor in the quantities of each and provide tradable quotes to the customers.

It would get more challenging when customers were looking at complex spreads involving three or more contract months. When these complex spreads extended into the deep deferred contract months, this got more tricky as the markets tend to be less liquid as you move out the curve. Bid-ask spreads can get wider and quantities can be significantly lower than what you find in the more liquid front contracts which are the contracts closer to their expiration. Clerks would have to “eye-ball” many of these spreads into the deferred contracts and make a best market. We’d assess the markets involved and calculate that considering their current bid-ask, it would imply that the market we are quoting should trade at a bid of “x” and ask of “y” even though there are now known orders at those prices. As clerks were providing tradable quotes, for safety sake, we would provide the worst possible prices with bid-ask spreads a few to several ticks wide. Then the race was on to tighten the quote by using your resources…your contacts in the back months who could provide tighter markets. Not all bids and offer were necessarily public…brokers and clerks would hold orders out of the market and wait to reveal when the market was optimal for their execution. This is similar to “ghost” orders that you see today or “sniper” orders.

All of this spread quoting math, the implied pricing calculations and bravado of providing the best quote has all been replaced by spreading applications like TT’s Autospreader and Spread Matrix. Autospreader and Spread Matrix provide tradable quotes even there’s not necessarily another trader on the opposite side of that spread quote. Much of the liquidity is implied, or calculated. These are the quotes clerks and brokers would calculate in their head. Market conditions in month A, when considered with market conditions in month B, would imply that you may execute a trade in spread AB at price X. Who owns that data? Who owns that quote in spread AB if there is no other trader making a market in spread AB? Think about those quotes that clerks and brokers provided. Who “owned” those quotes? The clerks and brokers were at risk. Did they own those quotes as they were at risk?

“Back in the day” brokers standing in the pits were the matching engines. The brokers did the work of matching orders against each other. This included matching direct orders against other direct orders as well as matching direct orders against implied prices. The exchange would clear the trades but the brokers matched. The brokers were at risk for the quotes they provided, as well as for the quotes their clerks provided, and on top of that were at risk for settlement as sometimes their trades didn’t get matched. Despite the head-nod in the pit, the broker or trader on the other side of the transaction could DK. DK = Don’t Know as in I don’t know what this guys is talking about…I don’t know this transaction.

Look back at the image that opens this blog. See the people sitting/standing at the elevated desks? Those desks were along two sides of the pit. The pit was oblong and along the long edges were desks of customers representing the leading investment banks and FCMs. Their job was to relay information and orders from their customers and colleagues on the other end of their phones. They’d generally communicate through a series of hand signals and shouting to the clerks who rang the outside of the pit. The majority of clerks were easy to find because of their yellow jackets. Those clerks effectively served as the modern day ISV. Clerks provided quotes while relaying orders and market information back and forth from customers to their brokers. These quotes were direct quotes, where we know there were resting order on paper order tickets in the pit, and implied quotes which required us to calculate the price differential between contracts and determine available quantity. The brokers, in the middle of the pit and wearing colored jackets, mostly red in this instance, were the matching engine. They paired off the trades.

Now, when a clerk received an order from a customer, as noted previously, the clerk would turn around and relay the information to their brokers to execute that trade. As you can imagine, there’s some lag there. There’s the turnaround time that delays the order from getting executed. The flow was desk, to yellow-jacketed clerk, to red-jacketed broker. What if we took a layer out of that? What if the filling activity happened on that outside layer? The ISV level. Turnaround time will have been slashed. Those yellow-jackets, the ISV level, may not have all of the information that the red-jacket/matching engine level had but they likely had enough to get the job done and get it done more quickly. That yellow-jacket/ISV level has information before the red-jacket/matching engine does and could reflect that information back out the their customers. Not to all customers, but just to their select customers. Now, information is viewed, processed and relayed back to select customers before the actual matching engine, the red-jackets there at the middle of the pit, even know what’s going on.

Sounds pretty advantageous.

Sounds like something TT is kicking around.

Sounds like Echo Chamber.