Skip to main content

Regulators Not Happy With Guy Whose Algorithm Tricked Some Other Algorithms

  • Author:
  • Updated:

An important element of any Wall Street education is figuring out what shady practices will win you a reputation as a genius, what shady practices will win you a reputation as a scumbag, and what shady practices will win you a prison sentence. There is substantial overlap!1 That education is extremely contextual, and your intuitions about what shadiness flies in one business won't necessarily help you in another, or in court for that matter. For instance I grew up in a corporate equity business, so I'd be happy to tell you why Yahoo!'s share repurchase from Dan Loeb wasn't insider trading but you can probably figure that out on your own. Meanwhile I have no idea what to make of spoofing, but it seems like Panther Energy Trading did some of it, and now they are in trouble:

The Commodity Futures Trading Commission said on Monday that it had fined Panther Energy Trading $2.4 million for a trading practice known as “spoofing,” in which bogus orders are used to draw in other traders. The firm and its owner were also barred from trading for a year. The agency said it was the first case to be brought using new rules against “spoofing” contained in the Dodd-Frank financial reform legislation.

The CFTC's press release is here, and various other regulators - the UK FCA and the CME among them - have also weighed in. The specific bad thing that Panther did went like this:2

  • Let's say there's a thing with a market at like 95/100.
  • Panter had an algorithm that would put in a smallish sell order at 99, just below the lowest standing offer, and would then put in a bunch of huge buy orders just above the highest standing bid, first at 96 then 97 then 98.
  • All the other algorithms in the market would see all those huge buy orders and be all "ooh guess the price is going up, aren't we so smart to notice that," so they'd bid higher.
  • Their higher bids would lift Panther's offer at 99.
  • Then Panther would cancel all its bids and would have just sold at 99, which is better than selling at 95 I guess, though also of course worse than selling at 100.
  • Then they'd reverse the process and try to buy at like 96 and put in a bunch of sell orders at 98, 97 etc., and the price would collapse and they'd buy at 96.
  • So effectively they'd just mess around within the bid/ask spread all day and make a lot of riskless-ish money by just deceiving other algorithms into chasing them.

Or something? My intuition for how scummy this is is undeveloped, my concern about "ooh sometimes they were able to sell below the offer and buy above the bid" is limited, and my sympathy for the algorithms that chased them is near zero. The complete process listed above would happen in about two-thirds of a second, meaning that the only people who had time to be fooled by this strategy were, y'know, not people. Really my assumption would have been that all the other algorithms were doing the exact same thing. Here is the actual law against spoofing, Commodity Exchange Act section 4c(a)(5)(C):

It shall be unlawful for any person to engage in any trading, practice, or conduct on or subject to the rules of a registered entity that ... (C) is, is of the character of, or is commonly known to the trade as, “spoofing” (bidding or offering with the intent to cancel the bid or offer before execution).

And here is a fact, or fact-ish statement, about high-frequency trading:

An estimated 95% to 98% of orders submitted by high-speed traders are canceled as the firms rapidly react to shifts in prices across the stock market, according to Tabb Group, which tracks trends in electronic trading.

If you put in an order knowing that there is a 98% chance that you'll cancel it before execution, do you intend to cancel it before execution? Like, statistically you sure do. Executing your order is a two-standard-deviation event. The CFTC's example of Panther's trading - where they canceled three orders for every one they executed - suggests that Panther executed something like 5-12x as many of its orders, proportionally, as the average stock-market HFT firm.3

So is every algorithmic trader spoofing all the time? I dunno, I'm sure you can find someone to tell you that the answer is yes. A more reasonable answer might be "well they are cancelling a lot of orders but not in this way where they are doing it to artificially inflate the price" and sure, though the law doesn't say that. (Interestingly the CFTC does, though it's also a little vague.4)

Why go after Panther and not all the other algorithmic traders cancelling all their orders? The answer appears to be that Panther, and its sole owner Michael Coscia, were completely nuts; as Paul Murphy points out, Panther would bid for $100 million of oil in order to make $340 on its actual trades. It's not entirely clear what Panther would have done if it'd been hit on its $100 million of fake bids but that sort of risk/reward ratio suggests the answer is "nothing good"; from a market stability perspective it's probably for the best that they were canceled. Also from a market stability perspective it's probably for the best that the CFTC and FCA shut this guy down.5

Still it's hard to know what to make of it: is this the start of a broader crackdown on HFT and its attendant order cancellations and illusory market depth? Or is it just like "this dude is beyond-the-pale nuts"? The CFTC is vague:

David Meister, the CFTC’s Enforcement Director, said, “While forms of algorithmic trading are of course lawful, using a computer program that is written to spoof the market is illegal and will not be tolerated. We will use the Dodd Frank anti-disruptive practices provision against schemes like this one to protect market participants and promote market integrity, particularly in the growing world of electronic trading platforms.”

As, actually, is Bart Chilton, doing rather less than I would have expected with the cat metaphor that Panther served up to him:

[T]hese types of violations of the law are becoming more common with the advent of high frequency traders (HFTs)—traders I’ve termed “cheetahs” due to their incredible speed. The cheetahs are to be commended for their innovative strategies, at the same time, when they violate the law, regulators need to be firm and resolute in our desire to deter such activities. Regulators already have a tough time keeping up with the cheetahs. Without sufficient deterrents, such as meaningful trading bans, many trading cats will simply find other ways to get back to their market hunting grounds.

Possibly not naming your firm after a big cat is step one in that process.

High-Speed Trading Firm Is Fined and Barred [DealBook]
Should Michael Coscia have been fined, or medicated? [FTAV]
CFTC Orders Panther Energy Trading LLC and its Principal Michael J. Coscia to Pay $2.8 Million and Bans Them from Trading for One Year, for Spoofing in Numerous Commodity Futures Contracts [CFTC]

1.The Venn diagram is like:

2.From the CFTC's complaint:

First, the algorithm placed a relatively small order on one side of the market at or near the best price being offered to buy or sell, in this instance a sell order for 17 contracts at a price of $85.29 per barrel, which was a lower price than the contracts then being offered by other market participants. Thus, the Respondents' offer was at the lowest, i.e. best, offered price. Second, within a fraction of a second, the Respondents entered orders to
buy a relatively larger number of Light Sweet Crude Oil futures contracts at progressively higher prices: the first bid at $85.26, the second bid at $85.27, and the third bid at $85.28. The prices of Respondents' bids were higher than the contracts then being bid by other market participants. Thus, Respondents' placed their bids at the highest, i.e. best, prices. By placing the large buy orders, Respondents sought to give the market the impression that there was significant buying interest, which suggested that prices would soon rise, raising the likelihood that other market participants would buy the 17 lots the Respondents were then offering to sell. Although Respondents wanted to give the impression of buy-side interest, Respondents entered the large buy orders with the intent that these buy orders be canceled before the orders were actually executed.

In the above example, the program sought to capture an immediate profit from selling the 17 lots. Thus, if the program successfully filled the small 17-lot sell order, the large buy orders were immediately cancelled and the algorithm was designed to promptly operate in reverse. That is, the algorithm would then enter a small buy order in conjunction with relatively large sell orders at progressively lower prices, which sell orders Respondents intended to cancel prior to

3.Ooh those numbers are pretty casual. The CFTC's numbers are "illustrative," and the canceled trades are bigger than the executed ones. The FCA order has more specfic numbers, with him faking orders on 261 contracts to actually trade 17, meaning that he's cancelling 94% of his orders or just a smidge less than the average. (The average for stocks, I mean; these are oil contracts though.) Also later in the order they have a little table; he cancelled a total of 88.5% of his orders without execution, by their count, including 96.5% of his large lots.

4.Here's their interpretive release on spoofing:

A section 4c(a)(5)(C) violation occurs when the trader intends to cancel a bid or offer before execution. CEA section 4c(a)(5)(C) “spoofing” prohibition covers bid and offer activity on all products traded on all registered entities, including DCMs and SEFs. “Spoofing” includes, but is not limited to: (i) submitting or cancelling bids or offers to overload the quotation system of a registered entity, (ii) submitting or cancelling bids or offers to delay another person’s execution of trades, (iii) submitting or cancelling multiple bids or offers to create an appearance of false market depth, and (iv) submitting or canceling bids or offers with intent to create artificial price movements upwards or downwards. The Commission further interprets the CEA section 4c(a)(5)(C) prohibition to include all bids and offers in pre-open periods or during other exchange-controlled trading halts. The Commission interprets that a CEA section 4c(a)(5)(C) violation requires a market participant to act with some degree of intent, or scienter, beyond recklessness to engage in the “spoofing” trading practices prohibited by CEA section 4c(a)(5)(C). Because CEA section 4c(a)(5)(C) requires that a person intend to cancel a bid or offer before execution, the Commission believes that reckless trading, practices, or conduct will not constitute a “spoofing” violation.

5.Not entirely fair? The FCA order notes that his algorithm "required certain conditions to be met including: (i) the bid/offer price having remained constant for a certain minimum period of time; (ii) a minimum size bid/offer spread; and (iii) the quantity of lots comprising the best bid and offer being within certain parameters"; presumably he took some precautions not to be hit on his fake bids.