By Hitesh Mittal, Founder & CEO, BestEx Research

A major trend we increasingly see is that fundamental discretionary managers are focusing most of their attention on their “hard orders”, those greater than 5% of a stock’s average daily volume. While there is always room to improve, even on lower difficulty trades, those are moving toward schedule-based algos like VWAP, while traders’ intellectual energy is preserved for the hard orders. The hard orders are generally mid- and small-cap names with sizes of 10, 20, or 30% of a day’s volume, or large-cap orders where the notional size creates real execution risk and portfolio managers are demanding speed. These are the trades where the gap between “good” and “bad” execution is measured in real money.

For these orders, finding natural liquidity with minimal impact in Alternative Trading Systems (ATSs) is more of a pipe dream than a reality. There is increasingly more volume in ATSs, but the diverse mix of market participants and focus on market share make information leakage virtually guaranteed. And resting orders get picked off with a regularity that should concern anyone paying attention.

The Manual Workaround

In our conversations with fundamental discretionary asset managers, we have found the inability to rest orders in dark pools without information leakage to be the single most frustrating aspect of algorithmic execution. Traders are finding that the moment they put orders into “dark” algorithms, prices move immediately. Orders fill suspiciously quickly when prices are moving in the client’s favor, and the reversion after execution completion is obvious to anyone looking at a chart.

So traders have adapted. Because they cannot trust most dark algos, they use them as tools rather than algorithms. They manage orders with tight limits, large minimum quantities, or both, submitting only a small slice at a time. The trader is doing the real work of execution.

The industry has spent two decades building increasingly sophisticated execution algorithms, but for the hardest orders, the ones that matter most, many experienced traders find them ineffective and use them strictly as smart order routers.

The Segmentation Mirage

The ATS transparency initiative through ATS-N filings was a welcome step toward solving this problem. In theory, better disclosure gives traders the information they need to choose venues wisely.

Segmentation, the mechanism by which ATSs separate different types of flow to protect institutional orders, is only as good as its implementation. And increasingly, we find that segmentation is only effective at some ATSs and is often designed to serve HFT market makers rather than institutional asset managers (as detailed in the filings and visible in the results).

The execution markout mechanisms used to fine-tune segments typically measure over millisecond horizons, with the exact time window often undisclosed. This captures the tip of the iceberg at best. A fill that looks clean at 100 milliseconds can look catastrophic after 10 minutes. A counterparty that shows neutral markouts in the venue’s scoring system can be systematically extracting profit from other counterparties on a longer time horizon.

Internal policing at ATSs often ignores obvious metrics. For example, how often is a conditional order indication faded before it results in a fill? Is there systematic impact from the counterparties fading, or not? What if the fading was due to race conditions due to an algorithm receiving multiple invitations to firm up rather than an intent to exploit information? These are answerable questions, but they require a willingness to look at the data honestly. Not every venue operator has the incentive to do so.

The ATS filings themselves are a wealth of information for those willing to read them carefully. Buried in the rule books are “opt out” features, order type nuances, and segmentation exceptions that materially affect execution quality. But it is practically impossible for a buy-side trader to read 30-plus ATS rule books, evaluate them against one another, and ask the right questions of their brokers. The information asymmetry is real, and it favors the operators and the participants with the most resources to analyze the rules, similar to the advantages gained from utilizing complex exchange order types.  Algo brokers are in the best position to help institutional clients mitigate that asymmetry, and especially those not operating their own venues.

Narratives Versus Data

The market structure conversation around dark liquidity is dominated by narratives. Some are useful, but many are not. Almost none of them are tested with sufficient rigor before they become conventional wisdom.

“Higher spread capture is always better.” “Trade more aggressively when the price moves in your favor.” “Use a large minimum size on every child order to reduce impact.” “Trajectory crosses are the gold standard.” “Conditional orders help you access passive resting liquidity.”

Each of these sounds reasonable, and each has a kernel of truth. But each, when tested against actual execution data, turns out to be far more conditional than the narrative suggests. Better execution markouts can coexist with terrible information leakage. Large minimum quantities reduce certain types of adverse selection but can create others. Conditional orders can be among the most toxic order types in some venues and among the most effective in others, depending on the segmentation, the counterparty mix, and the specifics of how the venue handles indication flow.

Accessing the “best segment” at a given ATS is not always the most effective strategy. In some cases, ATSs without offering any segmentation offer better execution than the best segment in other ATSs. In some ATSs, the fill rate difference between the best segment and the entire ATS is 90%, and in other cases both the quality and fill rates are virtually unchanged.

From Aggregation to Curation

Aggregating dark liquidity is no longer enough. The era in which connecting to more venues and sweeping more pools reliably improved execution is over. What is needed instead is curation – the disciplined combination of market structure knowledge and empirical data analysis. Neither is sufficient on its own. Data alone falls short because there are hundreds of ways to access liquidity and never sufficient data to evaluate all of them. Market structure knowledge alone falls short because most of the prevailing narratives don’t hold water when tested against actual results.

What does curation look like in practice? It starts with reading every ATS filing, speaking with ATS operators, and optimizing the configuration of each access point based on the findings. Then separate tactics should be designed for each access point: liquidity providing, liquidity taking, and conditional orders each behave differently at the same venue and need to be evaluated independently. Each tactic must be measured through the lens of information leakage and adverse selection. Some will have worse markouts but lower impact, others the opposite, and some are good at both. Then you optimize the mix, both across venues and within each tactic, calibrating minimum quantities, order duration, and timing to the specific characteristics of each access point. Measurement tools that allow you to evaluate the time series of the performance of each tactic and tactic-venue easily are key because the landscape continues to shift as the participant mix in ATSs does.

We published two papers that describe this Curation Framework in detail and an algorithm called Curator that implements it. We continue to evolve our framework as we run more experiments and gather more data.

What This Means for the Buy Side

For buy-side traders managing hard orders, the goal should be simple, strive for tools that you do not have to babysit. But in order to get there, you really need to be able to trust your algorithms and that requires asking questions.

Ask your brokers to articulate how they configure each ATS. What options do they opt in for and opt out of, and why? How do they separate information leakage from adverse selection when the measurements work in the opposite sense? Do not tie your brokers with restrictions like minimum quantity across the board, as these may be effective for some tactics but not others. And continue to study how market structure has evolved, beyond the narratives.

If traders are having to manage slices, set tight limits, and babysit dark algorithms, the problem is not the traders. The problem is that the algorithms are not built with sufficient rigor for the hardest use case.

The hard orders are not getting easier. Dark liquidity is not getting simpler. But the tools for navigating it can get meaningfully better, if the industry is willing to replace narratives with data and aggregation with curation.

Hitesh Mittal is the Founder and CEO of BestEx Research, a financial technology firm and agency broker-dealer specializing in the design of highly sophisticated execution algorithms for equities and global futures helping institutional asset managers, hedge funds, and broker-dealers reduce their transaction costs.