HLTV Rating 3.0 Updated With Map-Specific Accuracy Improvements

HLTV has released an update to Rating 3.0, its competitive Counter-Strike statistics system, incorporating data from 29,030 maps played since the metric's introduction. The refinements aim to improve accuracy in player performance evaluation across different map types and game scenarios.

A core strength of Rating 3.0 is its map-specific approach to win probability calculations. Unlike earlier versions, the system generates separate averages for each map, recognizing that player impact varies significantly by layout. Win probability percentages can differ by nearly ten percentage points between maps like Anubis and CT-sided maps such as Overpass. This context is essential for fairly assessing round swing and economic adjustments.

New maps presented a challenge for the system. Cache, which was introduced after Rating 3.0's release, initially lacked sufficient data for map-specific calculations. The update brings Cache-specific averages into play now that thousands of matches have been recorded on it. Cache shows a T-side win rate of 47.1% in buy rounds, a meaningful CT advantage that the previous global averages did not capture. The system still falls back to global averages for specific round scenarios on Cache when sample sizes remain low, such as particular eco situations or player counts.

HLTV also recalibrated its round swing calculations following changes Valve made to the in-game economy in July 2025. The original Rating 3.0 data contained insufficient matches under the new economy to fully reflect its impact, making this update necessary for consistency.

Technical improvements address edge cases in round evaluation. Kills that occur when the bomb cannot be stopped (fewer than 10 seconds remaining without a defuse kit, or 5 seconds with one) are now treated as exit frags rather than swing-generating plays. A previous bug where player round swing decreased after a kill has been fixed. In scenarios where win probability would still decrease after a kill, players now receive zero swing rather than a negative value.

The changes produce minor adjustments to historical ratings. Across 6,687 CS2 maps played between July and September 2026, single-map Rating 3.0 values moved an average of only 0.021 points. Approximately 9.4% of player-map ratings shifted by 0.05 or more, with 1% moving by 0.10 or more. The largest individual adjustment was 0.31 points. Cache is the most affected map due to its new specific data, while Inferno saw notable changes from the bomb timer adjustments to round swing.

HLTV stated the improvements will strengthen reliability ahead of the 2026 HLTV Awards and Top 20 Players of the Year announcements.