gem
v0.13.0

Experimental Features

The records in this section join several replay facts into one record per route, Roshan, smoke, fight or point. They hold no tags, scores or verdicts: gem reports what the replay shows, and answering questions from it is left to you (see the Recipes).

What makes them experimental

  • They join facts that the replay records separately, such as a camp catalog and sampled positions, or a Roshan kill and the fights that followed.
  • Some of their choices are analytical rather than read from the replay: window lengths, segment boundaries, and evidence categories such as strong_farm_evidence.
  • Those choices, and the record fields, may still change between releases.

Available experimental features

FeatureWhat it tries to answer
Farming PatternsWhich camp-local routes were observed, and what supports farming rather than transit
Farming Route CalibrationWhich real-replay facts and targeted boundaries the farming-route corpus guards
Roshan ConversionWhat happened after each Roshan kill: the Aegis lifecycle, and the fights, structures, economy, wards and Tormentors in the window
Roshan Conversion CalibrationWhich real-replay attribution, lifecycle and fight-association facts the Roshan corpus guards
Smoke AnalysisWhen each smoked hero gained and lost the modifier, what the enemy could see, and what happened next
Fight PositioningHow both teams were arranged at four bounded fight moments, with position freshness and opposing-team visibility kept explicit
Fight TimelineWhat each hero cast in a fight, who it hit and for how much, what hit each hero before it died, and the gold and XP paid for each kill
Point-Vision EvidenceBounded hero/observer geometry with explicit support, incompleteness, provenance, and separate target evidence
Vision ModifiersWhich reveal-style modifier windows gem tracks, how they are derived from combat-log events, and how they feed later vision analysis
  1. Bits & Bytes Primer
  2. Parser Internals
  3. Reports
  4. Farming Patterns
  5. Farming Route Calibration
  6. Roshan Conversion
  7. Roshan Conversion Calibration
  8. Smoke Analysis
  9. Fight Positioning
  10. Fight Timeline
  11. Point-Vision Evidence
  12. Vision Modifiers

The first three tell you where the underlying replay data comes from. The pages in this section explain how gem joins that data into each record, and what the record does and does not establish. The map regions and camp catalog these records use are on Map Regions and Camps.