Advanced analysis guides
Run controlled GPU, cross-venue, latency, and performance investigations without confusing a display change with a market-data result.
Use these procedures only after each input feed passes the checks in Run and operate VisualHFT. Feature availability depends on your plan and installed plugins.
Feature availability
Performance Counters and Data Feed Monitor are included with Core. Cross-Venue Analytics is a Marketplace add-on. GPU-Accelerated Charts and Replay Engine are included with Pro. Confirm the current pricing and add-on list before planning a workflow.
Analysis tracks
Preserve the input conditions
Before changing a chart or study, record the provider, venue symbol, normalized symbol, configured depth, aggregation, capture window, VisualHFT version, and plugin version. A result without those inputs cannot be compared reliably.
GPU-accelerated charts
GPU-backed charts change the rendering stage, not the connector or calculation inputs. Record the graphics adapter, driver, active rendering path, number of visible series, depth, and update rate. Compare the same saved or replayed interval when evaluating a rendering change.
Cross-venue analysis
Bring venues online one at a time. Before using Cross Venue Analytics, confirm that instruments share the same underlying, quote currency, contract type, and timestamp basis. Different aggregation or feed depth can create an apparent divergence even when both connectors are working correctly.
Market-event analysis
For an event study, record the event timestamp and timezone, retain a pre-event interval, and use one clock basis across feeds. Keep market-event time, connector receipt time, application processing, UI scheduling, and rendering as separate stages.
Latency and performance analysis
Use Data Feeds Monitoring for provider health and feed observations, and Performance Counters for application resource behavior. Report a distribution—at least median and tail percentiles—together with sample count and workload. A single average conceals stalls and burst behavior.
Recommended workflow
- State the question in measurable terms.
- Capture a baseline with one provider, one symbol, and the smallest useful workspace.
- Save the configuration and observation window.
- Change one variable.
- Repeat the same interval with Session Recorder and Replay Engine when reproducibility matters.
- Compare values only after confirming identical units and timestamps.
- Report what changed, what did not, sample count, percentiles, and any missing data.