Author: Deep Tide TechFlow

When discussing the big winners in cryptocurrency, people typically think of exchanges, market makers, or the "diamond hands" who became wealthy overnight during a bull market.

Retail investors experience drastic ups and downs during market transitions, with some buying at peaks and others selling at troughs; the volatility of the market becomes their nightmare.

However, in this market filled with speculation and frenzy, a group of people consistently maintains stable profits: quantitative trading teams.

These mysterious winners rarely appear in the public eye; they do not flaunt their earnings on social media, do not participate in KOL promotions, and seldom accept media interviews. They act like the "shadow force" behind the market, silently harvesting profits during every fluctuation in cryptocurrency.

What exactly enables them to make stable profits?

In this uncertain market, how do they manage to attack when necessary and defend when needed? They have turned trading into a science.

The World of Quantitative Trading: From Arbitrage to Arms Race

The history of crypto quant trading condenses half a century of evolution in traditional finance.

The Arbitrage Era (2017-2018) had simple and brutal rules: the price difference for the same cryptocurrency between different exchanges could reach 5-10%; a programmer could use a few computers to arbitrage across exchanges, yielding annual returns in multiples. BitMEX founder Arthur Hayes and FTX founder SBF both made their first pot of gold through arbitrage. In 2018, SBF noticed a 10% premium for Bitcoin in Japan and gathered a few friends to start arbitraging, earning about $20 million in just a few weeks, which led to the founding of Alameda Research.

It was an era of wild growth. "It was really bearish back then, first-level valuations crashing, second-level crashes, and many TokenFunds shifted to do quantitative trading," recalled former crypto VC practitioner Leo. "The early crypto quant scene mainly comprised three groups: elites returning from Wall Street, those who previously traded A-shares, and pure crypto wildcards." In the lawless era, everyone was exploring in the bear market; some practiced with hundreds of Bitcoins, while others cursed at exchange APIs that crashed frequently.

The Professional Era (2020-2023) was ignited by DeFi Summer. Numerous teams with backgrounds in traditional finance and the internet flooded in, and crypto quant began to truly professionalize. An essential change occurred on the funding side: family offices and institutional asset management replaced miners to become new financiers for quantitative teams.

"Now, the prominent quantitative teams are essentially in a 'kept' state, serving large family offices or asset management institutions. They don't lack financial backers, hence they purposefully maintain a low profile," described Grace, a BD from a crypto asset management firm, regarding the ecology of this circle.

The partner of multi-strategy quant fund Target Capital, Stephanie, stated that their clients are mainly renowned family offices from Singapore and Hong Kong. For family offices, the allure of crypto quant is immediate: they are not afraid of missing out on rampant growth, but they cannot endure crashes. A stable yield of 15-25% annually is far more attractive than a potential zero-valued coin. Therefore, crypto quant has become the first stop for many traditional family offices entering the crypto world.

The Institutional Era (2024 to present) has seen the approval of BTC spot ETFs, gradual establishment of global regulatory frameworks, and large traditional financial institutions entering the market. The "retail player table" in the crypto market is turning into an "institutional battlefield."

But the problem that all practitioners feel is this: strategies are becoming increasingly competitive and similar.

Leo’s judgment is straightforward: "More than 80% of secondary teams are employing very neutral arbitrage strategies, and the homogenization of strategies is quite severe."

Oliver Chen, CFO and COO of QSG, observed this trend from the perspective of an infrastructure service provider: "A clear example is funding arbitrage. Large funds, family offices, and LPs typically prefer strategies with low drawdowns and stable yield curves, making such strategies highly crowded over the past few years. The question is, as everyone's signals and trading logic become more similar, what really determines the difference in yield is often the infrastructure."

Even more awkwardly, the high risk-free interest rates in the crypto market have created a dimensionality reduction attack on traditional quant strategies. During a bull market, the risk-free interest rates on Pendle could soar to 30%, and the Alpha from your painstakingly developed model might yield less than lying flat in the blockchain.

As strategies become unyielding, the dimensions of competition begin to migrate.

From Competing Strategies to Competing Weapons: Why Now is the Window for Infrastructure?

This shift has already occurred in traditional finance.

In 2010, a company called Spread Networks spent $300 million to lay fiber optic cables across the Appalachian Mountains, simply to save 3 milliseconds between Chicago and New York. Jump Trading was even more aggressive, directly erecting a microwave tower on the roof of the Chicago Mercantile Exchange to replace fiber optics with near-light-speed wireless transmission. The arms race in Wall Street high-frequency trading lasted nearly twenty years, ultimately forming a consensus: when everyone's strategies are smart enough, the victor is determined by whose infrastructure is faster, more stable, and closer to the exchanges.

The crypto market is accelerating the replay of this path. The years 2024-2026 happen to be the window for this arms race to speed up, driven by three forces concurrently.

First, after the ETFs, the structure of market participants has changed. BTC and ETH spot ETFs have allowed traditional funds to enter the market on a large scale, making the trading structure more institutional. Institutional funds do not chase hundredfold coins; they focus more on stable yields, drawdown control, and execution quality. This directly raises the requirements for infrastructure.

Second, the complexity of trading opportunities is increasing. Nowadays, Alphas are not just hidden in price differentials between centralized exchanges; they are also found between CEX and DEX, between perpetual and spot, and between on-chain yields and centralized market spreads. Capturing these cross-market, cross-protocol opportunities requires far greater demands on market acquisition, routing, execution, and risk control than before.

Third, AI has accelerated the generation of strategies, further increasing the scarcity of infrastructure. In the past, an idea for a strategy from conception, backtesting, to launch could take researchers and engineers several weeks of repeated iterations. Now, AI is compressing the front-end cycle of strategy research: data cleaning, factor hypothesis, code generation, backtesting frameworks can all be built faster. However, as strategy generation accelerates, strategy homogenization also accelerates. The real gap lies no longer just in "who first thought of a factor," but in "who can more quickly access real markets with signals and convert theoretical profits into actual profits under latency, slippage, and access constraints."

More importantly, AI is giving rise to an entirely new trading form: AI Agents. In the past, the chain of trading decisions was "researchers generate ideas → engineers implement strategies → systems execute trades."

Now, AI Agents are starting to try to compress these three steps into one: autonomously perceiving market conditions, generating trading decisions, and directly calling execution channels to complete trades. As algorithmic trading proliferates exponentially, with more trading actions initiated by AI Agents rather than humans, the requirements for underlying infrastructure will suddenly rise. AI Agents will not call exchanges to negotiate VIP status, will not manually switch AWS nodes, and will not rely on experience to judge whether to cancel an order in extreme situations. What they need is a standardized, low-latency, highly reliable infrastructure interface that can be called and responded to at any time.

Tommy Ho, CSO of QSG, made an even more straightforward judgment: "The importance of strategies has not diminished, but strategies are increasingly dependent on infrastructure. Many crypto-native traders are very knowledgeable about the market and sharp, but they find it hard to compete with large institutions that possess complete infra teams in terms of low-latency situations, order execution, and AWS environment optimization. It's not that strategy is no longer important; it's evolved from being the 'only core' to being 'one of the cores.'"

Breaking Down the Technology Stack of Quant Teams: How Many Layers Must a Trade Cross?

How many layers of technical checkpoints must a quantitative team pass from discovering an opportunity to making a profit?

Most people think quant trading is just "write a strategy and run it." The reality is far more complex. A complete quantitative trading process, from signal generation to profit realization, requires crossing at least four technical layers. Each layer has its own thresholds and cost black holes, and the shortcomings at each layer can directly consume the Alpha generated by the strategy.

First Layer: Faster Market Acquisition

This is the starting point of the entire trading chain and the most easily overlooked aspect.

Most quantitative teams acquire market data through the public WebSocket interfaces of exchanges. The problem lies in that this channel is essentially "retail-level." There is a delay difference of several times between the public market data and the internal market-making channels of exchanges. In calm markets, this gap is not fatal. But the crypto market has never lacked extreme moments.

During extreme market conditions and surges in news flow, the public market data delay can balloon from milliseconds to seconds. In the high-leverage contract market, this is enough to cause prices to cross multiple quoting levels and even trigger a series of chain liquidations or stop losses. The model still makes judgments based on delayed market data while the real market has already transitioned to another state.

This is also why, in the past two years, some crypto quantitative teams have begun outsourcing their "market channels" from internal engineering problems to professional infrastructure service providers. QSG is one of the representatives of this trend.

They are not offering "a faster API," but instead productizing the low-latency trading capabilities that were previously hidden within top-tier market makers for more small to medium-sized quantitative teams. For example, with their market product (Sytus Feed), in extreme market conditions, the delay can be compressed from seconds to hundreds of milliseconds compared to public channels, significantly reducing jitter.

What makes QSG unique is that it does not approach trading from a traditional SaaS perspective, but instead reverse products from frontline quantitative and market making experience. Its core members come from institutions like Kronos Research, Jane Street, WorldQuant, and maintain high-level VIP and market-making qualifications with Binance.

Oliver admitted that other teams in the market also provide similar low-latency services, but QSG's barriers come from the combination of "trading understanding + engineering capability + execution experience." "If it were just selling colo data, the foundational conditions across the board wouldn’t differ much. The true challenge lies in further creating a performance advantage of 30% to 50% by understanding the receiving end, sending end, network path, system core, and exchange interface on top of having colo," he stated.

The infrastructure problem in the crypto market is not only a technical issue, but also a trading problem and an extreme market problem. In many past extreme market conditions, systems are most likely to expose issues: market delays, disconnections, matching congestion, and failed cancellations. These details can hardly be perfectly inferred without firsthand experience.

Second Layer: More Stable Order Execution

Even if you see the right price, slow order placement is useless.

The bottleneck in order execution, for most teams, lies not in the network but in their engineering capabilities. Building a low-latency execution channel that has been deeply optimized requires engineers who understand Linux kernel tuning, network card driver optimization, and user-space network stacks—such talents are scarce globally and even rarer in the crypto space.

A scenario that illustrates the importance of infrastructure: during a highly volatile period in December 2024, the order placement delay for most trading institutions surged above 120 milliseconds, while teams using institutional-level execution channels stably operated around 40 milliseconds. In client tests, the improvement in latency for some exchange scenarios exceeded 90%.

Tommy told us that many clients have encountered the same pitfall in both layers of market data and execution: "Many teams only looked at average latency at first and thought the system was fast enough. But when the market becomes highly volatile, the tail latency ultimately determines whether you can timely receive market data, send out orders, and control risks. We often say internally: P50 determines how fast you look on average, but P99 determines whether you can survive in the fiercest markets."

The combination of market acquisition and order execution means the entire link from "seeing" to "eating." Every additional millisecond of delay in this link discounts the actual profits of the strategy. Teams doing inter-exchange arbitrage feel this the most: between Tokyo and Hong Kong, the physical distance itself incurs latency. Some infrastructure service providers offer cross-region connection tools that can reduce such round-trip delays by over 30%, while others automatically seek the optimal cloud nodes for target exchanges. These are optimizations for the "last mile" that might seem insignificant individually, but collectively determine whether a team can outperform others in homogeneous strategy competition.

Third Layer: Higher Exchange Permissions

There is an open secret in the quant industry: running the same strategy on a VIP3 and a VIP9 account can yield a difference of up to 100% in returns.

The reason is straightforward. The higher the VIP level, the lower the transaction fees (the highest VIP can have negative maker fees, meaning the exchange pays you to trade), the looser the API rate limits, and the better the borrowing rates. More crucially, the low-latency dedicated endpoints that come with market maker qualifications are far faster than public APIs.

However, obtaining high-level VIP status comes with a very high barrier. For example, to reach Binance VIP9, a monthly futures trading volume of $25 billion is required. If the strategy itself does not make money, just to maintain that level of trading, the friction cost could lead to tens of millions in losses in a year. It's a classic "which came first, the chicken or the egg" dilemma: you need the cost advantage of top VIPs to make the strategy profitable, but you first need to achieve that volume to qualify.

The services provided by QSG at this layer do not follow the broker model. Against the backdrop of top exchanges explicitly prohibiting broker business, QSG has found a win-win path with exchanges: by utilizing high-frequency trading technologies to help clients generate real incremental trading on their own accounts, thus reaching the thresholds for market maker qualifications and high-level VIPs. The fee discounts, low-latency endpoints, and institutional-level borrowing privileges that clients receive are all tied to their own accounts, not through any form of sub-account or agency relationships. This model means genuine liquidity growth for the exchange and tangible cost advantages for the client, aligning the interests of both parties.

Fourth Layer: Lower Slippage for Large Trades

A quantitative fund managing tens of millions of dollars often finds its biggest headache is not the strategy but the execution.

When you need to establish or close a large position in one go, the market's liquidity may not suffice. A large market order could eat up several basis points in slippage. Executing dozens of times a month, the accumulated slippage costs could wipe out the profits of the strategy. Traditional finance has mature large block trading channels and dark pools, while the crypto market has long lacked in this respect.

Some infrastructure service providers are starting to introduce large execution, smart routing, and price-matching mechanisms from traditional finance into the crypto market. QSG's large execution service is one case: using algorithmic trading and execution optimization to try to reduce the impact of large orders on public markets. Tommy mentioned that in certain scenarios, they helped clients improve execution costs by about 3 basis points compared to standard TWAP strategies. "Not all clients care about a few hundred microseconds," he said, "some CTAs, long-short or large fund adjustment teams are more concerned with slippage during the execution of large orders. Over the long term, this translates to very tangible profit improvements."

With all four layers completed, a key question emerges: to build in-house or rely on external access?

Early quantitative teams favored a "full-stack model," building everything from strategies to infrastructure in-house. However, the cost of this model is becoming unbearable.

Tommy shared an impressive client case: a small but excellent live trading quant team with strong strategy capabilities, but lacking the time to re-create the VPC, network, market data formats, ordering protocols, and trading paths from scratch. "Their pain point is not that they can't write strategies, but that the opportunities for strategies are cyclical. If a strategy is profitable right now, but the team still has to spend a few months building the AWS environment, market data system, and ordering system, by the time the infrastructure is ready, the market opportunity may have already passed."

For many small to medium-sized quantitative teams, the real question is not "can they build in-house," but "is it worth building in-house?"

If the core strengths of a team lie in strategy research, capital management, and risk control, spending a year on low-latency networks, exchange permissions, execution engines, and large transaction systems may not be the optimal solution. Building in-house infrastructure means greater control, but also higher fixed costs, longer deployment cycles, and greater maintenance pressure. For teams of 5-15 people, professional division of labor may be more realistic than full-stack building.

More fatal is the time cost. Building a complete low-latency trading system from scratch conservatively estimates 6-12 months. During this time, your competitors have already run the same Alpha using off-the-shelf infrastructure. Alphas inherently have a shelf life; the window of market inefficiency rapidly narrows as more participants enter. Every day spent reinventing the wheel degrades the precision of the strategy.

Of course, externalizing infrastructure does not mean a team can relinquish its engineering capabilities. Strategy logic, risk control frameworks, capital management, and exception handling must still be firmly in their own hands. Third-party infrastructure addresses the efficiency issues at the execution layer, not replacing the core investment capabilities of the team.

QSG has now been listed on the AWS Marketplace, utilizing a standard enterprise-level SaaS model. For traditional financial institutions entering the crypto market, this means a seamless compliance procurement path, standardized billing, with no need to engage with any complexities involving tokens or crypto-native processes.

The crypto market is rapidly entering the "professional division of labor" era. Just like quantitative funds in traditional finance outsource trade execution to prime brokers and data to Bloomberg, crypto quant teams are beginning to delegate their infrastructure needs to specialized third parties. Strategies are their core assets; infrastructure does not necessarily have to be.

Conclusion

Oliver has a clear judgment of the future: “The infrastructure for crypto quant will shift from being an 'optional tool' to the 'standard configuration for professional trading teams.' As AI increasingly integrates into strategy research, signal generation, and parameter optimization, the barriers on the strategy side will lower, and more teams will generate similar trading ideas at lower costs. The strategies themselves will become more crowded; the real differentiation will revert to the capability of underlying execution.”

He summarizes the future competitive landscape as a formula: AI Strategy + Data Quality + Execution System + Low-Latency Infrastructure + Risk Control Capability. This is not a comparison of singular dimensions, but a competition of comprehensive capabilities.

In Oliver's planning, one of QSG's future evolutionary directions is to connect market data, order placement, node optimization, data, large execution, and risk control monitoring into a complete intelligent trading infrastructure through AI Agents. "In the future, AI Agents could serve as co-pilots of trading infrastructure, helping trading teams monitor the market, diagnose system issues, and optimize execution paths," he stated. Tommy added a more vivid analogy: "Just as the explosion of applications on the App Store occurred after the proliferation of smartphones, as more and more traders begin to use programmed and AI-assisted trading, they will need not to build their own infrastructure from scratch but a trading infrastructure network that can be directly called upon."

This path has already been traversed by traditional finance. Bloomberg Terminals, various Prime Brokers—these infrastructures ultimately defined the rules of the game on Wall Street. The crypto quantitative track is waiting for its own "infrastructure moment."

For infrastructure service providers like QSG, the opportunity lies not just in selling a set of tools to quantitative teams, but in participating in the rewriting of the fundamental standards of cryptocurrency trading: how markets are accessed, how orders are executed, how large trades are matched, and how exchange permissions are serviced.

As the crypto market transitions from its grassroots beginnings to institutionalization, the capabilities previously hidden within top-tier market makers are being dismantled, packaged, and gradually transformed into public infrastructure accessible to more teams.

The next time BTC experiences extreme volatility, while most are waiting for market updates in those few seconds, the real battle may have already been won.