Direct answer: key data inputs to assess low liquidity pairs
To assess low liquidity pairs, collect data that lets you (1) define liquidity conditions, (2) measure trading frictions and execution sensitivity, and (3) validate that the data is timely, complete, and comparable across sources. Because liquidity changes with market regimes, the most reliable assessments separate stable measurement mechanics from variable market and provider conditions.
At a minimum, you need inputs from four areas: liquidity/flow, price/quote behavior, execution cost components, and data provenance plus quality checks.
Mechanism and definition: what “low liquidity” means in practice
Low liquidity pairs are currency pairs where trading interest and available orders are insufficient for smooth price discovery. In practical terms, this often shows up as thin trading, quote gaps, and larger sensitivity to order size.
When liquidity is low, small trades can move prices more than they would in a highly liquid market. Quoted spreads may widen, and the “spread you see” may differ from the “spread you pay” when actual execution finds fewer counter-orders. Therefore, your data should capture not just a single snapshot, but how quotes and transactions behave across time windows.
Evidence and example: what to collect and how to check it
1) Liquidity and activity metrics
Collect measures that reflect how many participants are active and how often the pair trades. Useful examples include:
- Trade frequency and/or the number of executed trades over defined time windows.
- Traded volume distribution (not just totals), including whether activity concentrates at certain times.
- Depth proxies, such as availability of quotes around the mid price (if your data source provides it).
2) Quote behavior and price stability
Low liquidity often creates irregular quote dynamics. Collect:
- Bid–ask spread time series (not a single average).
- Mid-price changes and the distribution of short-horizon price moves.
- Evidence of quote staleness (for example, repeated timestamps or long gaps between updates, if available).
3) Execution cost and trading friction inputs
Even without real-time data, you can assess “likely friction” using historical execution-relevant fields from your source, such as:
- Spread-related cost estimates from the observed bid–ask spread distribution.
- Slippage proxies: how far the execution price deviates from the observed quote at the moment (only if the dataset includes execution vs quote).
If you do a calculation, state assumptions explicitly, such as using a selected time window, a specific percentile of spread, and whether you assume execution at the quoted bid/ask or using observed execution prices.
4) Provenance, timeliness, and comparability
Before interpreting any metric, record:
- The data source category (public market data vs provider feed vs internal execution logs).
- Timestamp definition and timezone handling, including whether quotes and trades are aligned to the same clock.
- Sampling frequency and coverage: are there missing intervals, holidays, or instrument-specific outages?
A simple worked example (with assumptions)
Assume you compute a “typical cost” measure using the 80th percentile spread observed over a chosen 30-minute window in a historical dataset. You then compare it to another window with higher activity.
This separates mechanics from variation because:
- The spread percentile summarizes observed quote conditions.
- The window choice is an explicit assumption.
- Any difference in results can be attributed to different liquidity regimes within the history you analyzed.
Limitations and risks: at least one material failure mode
A major failure mode is relying on stale or incomplete data. In low liquidity conditions, quotes may update less frequently, and data feeds may have gaps. If your dataset under-reports updates or mixes instruments/time ranges with different coverage, you can incorrectly conclude that spreads are stable or that price moves are small.
Other common limitations include:
- Regime shifts: historical relationships may not hold when volatility or participation changes.
- Provider effects: your visible quotes can reflect venue choice, matching rules, or internal routing differences rather than the broader market.
- Cost measurement gaps: without execution-vs-quote fields, spread-based estimates may miss additional slippage.
Verification and next question: how to validate your assessment
Use a checklist of verifications you can apply to any dataset:
- Confirm timestamps and timezone consistency across quotes and trades. - Check missing-data rates and whether the instrument was active during all windows you compare. - Inspect spread and quote-update irregularities (especially long gaps).