lev-os Analyzing Dark Pool And Alternative Venues
Evaluates alternative trading systems with fill rate analysis, information leakage assessment, and venue toxicity measurement.
When To Use
- Comparing execution quality across dark pools and ATS venues for order routing decisions
- Investigating information leakage or adverse selection on a specific venue
- Performing periodic venue analysis required under Rule 606 or best execution obligations
- Evaluating whether to add, remove, or re-weight a venue in smart order router (SOR) configuration
- Assessing venue toxicity after noticing degraded fill quality or increased markouts
Inputs To Gather
- Execution data: Fill rates, fill sizes, time-to-fill distributions, and midpoint vs. far-touch fill ratios per venue over the analysis window (minimum 20 trading days recommended)
- Markout data: Short-term price reversion at standard intervals (e.g., 1s, 5s, 30s, 1min, 5min post-fill) per venue
- Order flow profile: Breakdown of order types routed (midpoint peg, limit, IOC, conditional) and average order size relative to venue ADV
- Venue metadata: ATS Form ATS-N filings, subscriber segmentation rules, anti-gaming controls, minimum quantity thresholds, and crossing logic (continuous vs. periodic) [VERIFY against current SEC EDGAR filings]
- Benchmark data: Arrival price, VWAP, or interval VWAP for the relevant period to contextualize fill quality
- Market context: Average daily volume, volatility (realized and implied), and spread regime for the securities analyzed
Workflow
Define scope and segmentation
- Specify the universe of securities (large-cap, mid-cap, small-cap, ETFs) and time window
- Segment analysis by market-cap tier, spread bucket (sub-penny, 1–3¢, 3¢+), and volatility regime — venue performance varies materially across these dimensions
- Identify the set of venues to compare (include lit exchanges as a control benchmark)
Calculate fill quality metrics
- Fill rate: Orders filled / orders routed, segmented by order type and urgency
- Effective spread: Execution price vs. midpoint at time of fill, expressed in bps
- Price improvement: Percentage of fills that improve on the NBBO, and average improvement in mils per share
- Fill size ratio: Average fill size / average order size — flags venues that consistently partial-fill
- Time-to-fill: Distribution of latency from order arrival to execution; flag venues with bimodal distributions (may indicate information-dependent crossing)
Assess information leakage and adverse selection
- Compute markout curves per venue at 1s, 5s, 30s, 1min, and 5min intervals post-execution
- Negative markouts (price moves against you after fill) indicate adverse selection; compare venue markout to lit-exchange baseline
- Flag venues where markout deteriorates sharply beyond 5 seconds — suggests informed flow or latency arbitrage
- Review reversion asymmetry: if buys mark out worse than sells (or vice versa), investigate whether venue subscriber mix is skewed
Measure venue toxicity
- Calculate VPIN (Volume-Synchronized Probability of Informed Trading) or similar toxicity proxy per venue if tick data is available
- Track the ratio of aggressive-to-passive fills — high aggressive ratios may signal predatory flow
- Evaluate the venue's anti-gaming controls: Does it offer minimum rest times, order randomization, periodic auctions, or size priority? Cross-reference with Form ATS-N disclosures [VERIFY current ATS-N filings for each venue]
- Compare toxicity metrics to the venue's stated subscriber segmentation (e.g., does the venue claim to exclude high-frequency participants but show HFT-consistent markout patterns?)
Benchmark and rank venues
- Normalize all metrics to a common scale (e.g., z-scores within each spread/size bucket)
- Produce a composite venue scorecard weighting: fill rate (20%), effective spread (25%), markout at 1min (25%), fill size ratio (15%), toxicity score (15%) — adjust weights to reflect desk priorities
- Rank venues within each security tier; highlight venues that rank well on fill rate but poorly on markout (classic "toxic fill" pattern)
Formulate routing recommendations
- Recommend SOR weight adjustments: increase allocation to venues with favorable markout-adjusted fill rates, reduce or eliminate venues showing persistent adverse selection
- Identify conditional routing rules (e.g., route to Venue X only for orders >5,000 shares where its periodic auction adds value)
- Flag venues warranting further investigation or a probationary period before removal
Output
Deliver a Dark Pool & ATS Venue Analysis Report containing:
- Executive summary: Top-line findings, worst/best performing venues, and headline routing changes recommended
- Venue scorecard table: Composite scores and component metrics per venue, segmented by security tier
- Markout curve charts: Overlay markout profiles across venues at standardized intervals
- Information leakage flags: Specific venues and security segments where adverse selection exceeds threshold (e.g., >1.5× lit baseline)
- Routing recommendations: Specific SOR configuration changes with expected impact on execution cost (bps saved per share)
- Data limitations: Note any gaps in execution data, short sample windows, or venues with insufficient fill counts for statistical significance
Quality Checks
- Confirm markout calculations use the correct midpoint timestamp (execution time, not order arrival) — a common source of error
- Verify fill data excludes auction prints, odd lots, or other non-representative executions unless explicitly in scope
- Ensure statistical significance: venues with fewer than 100 fills per bucket should be flagged as low-confidence
- Cross-check venue-reported fill statistics (from ATS quarterly reports on FINRA) against internally computed metrics — discrepancies may indicate data feed issues [VERIFY FINRA ATS transparency data for current quarter]
- Validate that spread and markout calculations account for tick-size regime (sub-penny venues vs. tick-constrained names)
- Confirm that the analysis period does not overlap with abnormal market events (e.g., volatility spikes, exchange outages) without explicit notation