Skip to content
eStatcast

Preview data

Round Win Probability Added

RWPAVALv0.1.0 · experimental · validated 2026-07-16

Computed · experimental. Values are computed, but the definition or model may still change.

What it measures

How much a player's kills and spike actions moved their team's chance of winning rounds, per round played.

Formal definition

Round win probability is evaluated before and after each attributable event (kills, spike plants and defuses); positive ΔWP is credited to the acting player and averaged over rounds played.

Inputs

  • kill events
  • spike events
  • round state timeline
  • round_wp_cs2 model

Exclusions

  • rounds with missing side or start-time state

Sample & population

Minimum sample
60 rounds
Baseline population
Players in the same game/window with ≥ minimum rounds
Direction
higher better
Unit
win prob/round

Percentile interpretation

90th percentile ≈ consistently decisive in rounds, not merely high-fragging.

Confidence = source confidence × model confidence. Values from baseline models are capped at 0.7. Below-minimum samples display the value grayed out with no percentile.

Example

Killing the last defender in a 1v1 post-plant moves win probability from 0.62 to 0.95 → +0.33.

Known limitations

  • The round-win-probability model is the shared CS2 baseline mapped onto attack/defense; its round and spike clocks are CS2's, so the timing term is approximate for VALORANT
  • Ability usage does not move the model — only kills and spike events are attributable, so utility-led rounds under-credit initiators and controllers
  • Victims are not debited, so this reads as value created rather than net value

Current distribution

career window · n=30 eligible players

Percentile spread

Each published percentile placed on the value axis. Only these points were measured — the strip draws no curve between them.

p10.047p100.064p500.084p750.100p900.122p990.150

The shaded band is the middle half of the population, p25 to p75. Points between two markers were never published, so nothing is drawn there.

Show the numbers
Percentile spread — underlying values
PercentileValuevs median
p10.047-0.037
p20.048-0.036
p30.049-0.035
p40.051-0.032
p50.055-0.029
p60.058-0.025
p70.062-0.022
p80.063-0.021
p90.063-0.020
p100.064-0.019
p110.065-0.018
p120.066-0.018
p130.067-0.017
p140.067-0.016
p150.068-0.015
p160.069-0.014
p170.070-0.013
p180.071-0.013
p190.072-0.012
p200.073-0.011
p210.074-0.010
p220.075-0.009
p230.077-0.007
p240.078-0.006
p250.078-0.005
p260.078-0.005
p270.078-0.005
p280.078-0.005
p290.079-0.005
p300.079-0.004
p310.079-0.004
p320.080-0.004
p330.080-0.004
p340.080-0.003
p350.080-0.003
p360.081-0.003
p370.082-0.002
p380.082-0.001
p390.082-0.001
p400.082-0.001
p410.082-0.001
p420.082-0.001
p430.083-0.001
p440.083-0.001
p450.083-0.000
p460.083-0.000
p470.083-0.000
p480.083-0.000
p490.083-0.000
p500.0840.000
p510.0840.000
p520.0840.000
p530.0840.000
p540.0840.000
p550.0840.000
p560.0840.001
p570.0850.001
p580.0850.001
p590.0860.002
p600.0880.004
p610.0900.006
p620.0910.008
p630.0920.008
p640.0920.008
p650.0920.009
p660.0930.010
p670.0950.011
p680.0960.013
p690.0980.014
p700.0980.015
p710.0990.016
p720.1000.016
p730.1000.017
p740.1000.017
p750.1000.017
p760.1010.017
p770.1040.021
p780.1070.024
p790.1100.027
p800.1110.028
p810.1120.028
p820.1120.028
p830.1120.028
p840.1130.029
p850.1130.030
p860.1140.030
p870.1160.032
p880.1180.034
p890.1200.037
p900.1220.039
p910.1250.041
p920.1270.044
p930.1300.046
p940.1300.047
p950.1310.047
p960.1310.047
p970.1350.051
p980.1430.059
p990.1500.067
Published metric distribution for the stated window.

Population

every qualified player's season value

win prob/roundseasonn=30

Q1 0.078 · Median 0.084 · Q3 0.100 — half of the 30 observations fall between Q1 and Q3. qorren sits at the 98th percentile: higher than 29 of them, level with 1. 8 bins, sized from the spread and the sample.

Show the numbers
Population — underlying values
Round Win Probability AddedObservationsShare
0.046 – 0.06026.7%
0.060 – 0.074516.7%
0.074 – 0.0881136.7%
0.088 – 0.102516.7%
0.102 – 0.116310.0%
0.116 – 0.13026.7%
0.130 – 0.14413.3%
0.144 – 0.15813.3%
Computed from player metric summaries past each metric's minimum sample

Model card

round_wp_cs2 v0.1.0
MODELlogistic_regression · baseline
Training data — read this

Hand-specified priors, sanity-calibrated on synthetic fixture corpus v1. NOT fitted on production data.

Calibration

Terminal states pinned to truth; mid-round states are directional priors pending real-data fit.

Features
  • alive_diff(CT alive − T alive) / 5
  • alive_ratio_edgeNear-wipe indicator (±0.5 when a side is down to 1)
  • equip_diff(CT equipment − T equipment) / 5000, clamped
  • bomb_plantedBomb planted indicator
  • bomb_time_pressurePlanted × (1 − bomb-time fraction remaining)
  • late_no_plantNo plant × (1 − round-time fraction remaining)
Not intended for

Live betting · Cross-game comparisons

Change log

  • v0.1.0 · 2026-07-16Initial baseline release

→ Rank players by RWPA