Market Tech: How To Spot an FTX

By Sam Gaer

I called out anomalies in FTX’s technology and order handling practices over a year ago. And no one seemed to care, as long as the music kept playing. Then the music stopped. 

Mature, deep, liquid markets depend on a certain feature set: liquidity, competition among exchanges and a multitude of market participants (brokers, market makers, speculators, hedgers, for instance). They also require critical in-built risk-mitigation protections to assure users of the market that they are getting access to a level playing field, where no single operator or group has an unfair rules-based or risk-based advantage. Clearly with hindsight, this feature set was absent in FTX’s trading relationships — and discernible from the data that I outline below. If you find this helpful, stay tuned for more deep dives on some of the above feature sets, risk mitigants and tools that the FTX experience reminded me we deeply need, in order to cultivate liquid markets that we can all rely upon.

For context, I’ve spent my 30-year career as CIO/EVP of NYMEX and FINRA, founder of several technology and high-volume trading firms, and as former Global Vice Chairman of FIX Protocol Derivatives Committee— humbly, it’s all this time in the trenches that makes me feel fit to assess the robustness of trading technology, whether in the crypto markets or otherwise, and to share my observations here, in case helpful. I also suspect I’m not alone — and hope this sparks an open dialogue among market technicians and participants on how we might continue to add depth and high-fidelity to existing and new asset classes.

Years of running high-frequency trading technologies and intimate knowledge of market microstructure and the intricacies of various exchange protocols, including FIX, revealed, upon some surface data diligence, that FTX, an exchange claiming to have the best technology in the cryptocurrency industry, was actually engaging in unethical practices such as front-running and exploiting traders.

What We Discovered: FTX was 800% slower than Binance

Over a year ago, I conducted an analysis of samples of FTX’s FIX messaging logs and discovered that the average “round trip” latency was 85 milliseconds. In contrast, other “middle of the road” exchanges like OkEx and Bybit had latencies averaging 45 milliseconds, while Deribit, on a non-co-located (cloud-based) basis, had latencies averaging about 500 microseconds to 2 milliseconds, and rival exchange Binance averages 5 to 10 milliseconds. To add fuel to the fire, FTX had some of the most restrictive API messaging policies in the industry.  Which begs the question- if the technology stack at FTX was as good as touted, why would they need to restrict messaging to such an intrusive degree, and why would response times be so poor?

To put this in perspective, it took 800% more time to exchange messages with FTX than Binance.  This was an immediate red flag once we ascertained that this was not a flaw in our hardware, software or network setup.

For clarity, the term “round trip” latency was measured as the time difference between when an order leaves our server, is acknowledged by the exchange, and is received back at our server. I won’t get into the deep-dive technicals but this is a generally accepted measure of latency, though not exhaustive. I used this measure as a control methodology- so the exact same measurements were taken on the exact same servers, located in AWS regions where the respective exchanges recommended for lowest latency. (I know most to most in the HFT community, the concept of measuring in milliseconds is like a time-warp back to the early 2000’s, but this basically status quo in crypto.)  For those unfamiliar with “FIX” it is short for Financial Information eXchange, and is an industry standard communication protocol used by nearly every bank, broker, and exchange.

FTX’s high relative latency, restrictive messaging rate policies, non-existent customer support (which could not answer basic technology/latency/network questions despite billions of dollars of venture capital), collectively led me to the conclusion that the only explanation for such poor performance was a “cherry-picking” system that front-ran or took the opposite side of incoming orders that were instantly profitable (known as “riskless principal”), or worse, front-ran those orders. I also presumed that some participants must have had better/faster access and bypasses to certain gating features that were encountered by nearly everyone else who traded on the system. This was further exacerbated by the fact that Alameda, the trading firm owned by FTX founder and majority shareholder Sam Bankman-Fried, was operating as a proprietary trading vehicle on FTX — this was a huge red flag that would not be tolerated in any mature, regulated exchange purporting to provide a level playing field for all participants. Notwithstanding my alarm, the industry and media were infatuated with SBF’s aura and FTX. Many even laughed at the suggestion of unethical practices.

This experience serves as a valuable reminder that, perhaps especially in a rapidly-growing and innovative industry like cryptocurrency where the regulatory rails are still in flux, it is essential to remain diligent, vigilant and thoroughly evaluate the technology and practices of exchanges and their counterparties.

The Good Part

Candidly, there was a silver lining to my learnings: it was tech I built that made it relatively easy to spot these issues early and avoid collateral damage: Katana, a quantitative investment manager and software developer focused on energy, equity index, and cryptocurrency markets, relies on in-house end-to-end software designed specifically for ultra-low latency trading, typically recording single-digit microsecond latencies, using the fastest available tools and processors. Our tooling makes market gaps and counterparty inadequacies easy to spot and price while our trader-facing systems are state of the art and highly performant, capable of trading multiple asset classes on multiple screens with consistent performance and high availability.  We utilize state of the art techniques for message handling, compression, and encryption, making our front-end technology second to none in terms of performance. 

Stay Tuned

If you found this helpful, stay tuned for more market structure reflections and what we are up to by following us

This content will also be posted on LinkedIn. My profile is located at https://www.linkedin.com/in/samgaer/