Match Predictions

Introduction

This vignette covers bouncer’s prediction capabilities: - Pre-match predictions: Who’s favored before the game starts? - In-match win probability: Live updates as the game progresses - Score projections: What’s the expected final score?

Pre-Match Predictions

Predict the outcome before a match begins:

library(bouncer)

# Basic match prediction
prediction <- predict_match("India", "Australia", format = "t20")
print(prediction)

The prediction uses: - Team ELO ratings - Recent form - Head-to-head history - Team skill indices

Adding Venue Context

# Prediction with venue
prediction <- predict_match(
  team1 = "India",
  team2 = "Australia",
  format = "t20",
  venue = "Melbourne Cricket Ground"
)
print(prediction)

In-Match Win Probability

Calculate live win probability during a match:

# First innings: batting team at 85/2 after 10 overs
wp <- predict_win_probability(
  current_score = 85,
  wickets = 2,
  overs = 10.0,
  innings = 1,
  format = "t20"
)
print(wp)

Second Innings (Chasing)

# Second innings: chasing 180, currently 100/3 after 12.4 overs
wp <- predict_win_probability(
  current_score = 100,
  wickets = 3,
  overs = 12.4,
  innings = 2,
  target = 180,
  format = "t20"
)
print(wp)

# The required run rate is factored into the probability

Score Projection

Project the final innings score from any game state:

# Scoreboard: "Mumbai Indians 80/3 (10.0 overs)" in T20
projected <- calculate_projected_score(
  current_score = 80,
  wickets = 3,        # wickets fallen (matches scoreboard)
  overs = 10.0,       # overs bowled (matches scoreboard)
  format = "t20"
)
print(projected)

Understanding Resource Percentage

The projection uses a resource-based formula similar to Duckworth-Lewis:

# Calculate resource percentage remaining
resource <- calculate_projection_resource(
  wickets_remaining = 7,  # 10 - 3 wickets fallen
  balls_remaining = 60,   # 10 overs left
  format = "t20"
)
print(resource)  # e.g., 0.65 = 65% of resources remaining

Resources depend on: - Wickets in hand (more wickets = more resources) - Balls remaining (more balls = more resources) - The interaction between them

Understanding ELO and Win Probability

ELO ratings convert to win probability using the standard logistic formula:

Formula: P(A wins) = 1 / (1 + 10^((ELO_B - ELO_A) / 400))

Example calculations:

ELO Difference	Higher-Rated Team Win Probability
0 (equal teams)	50%
+100 points	~64%
+200 points	~76%
+300 points	~85%

The 400 divisor means that a 400-point difference corresponds to roughly 10:1 odds (91% win probability).

In bouncer: - Starting ELO is 1500 for all teams - ELO updates after each match based on result vs expectation - Format-specific ELO tracks performance within T20/ODI/Test separately

Matchup Predictions

Predict specific batter vs bowler outcomes:

# What happens when Kohli faces Bumrah?
matchup <- predict_matchup_outcome(
  batter = "V Kohli",
  bowler = "J Bumrah",
  format = "t20"
)
print(matchup)

Team Comparison

Get a comprehensive comparison of two teams:

comparison <- compare_teams("India", "Australia", format = "t20")
print(comparison)

# Includes:
# - Current ELO ratings
# - Win probability
# - Recent form
# - Head-to-head record
# - Skill comparisons

Prediction Accuracy

The prediction models are calibrated on historical data. Typical accuracy:

Prediction Type	Accuracy
Pre-match winner	~65-70%
Score projection	RMSE ~15-20 runs
Win probability	Well-calibrated (50% events happen ~50% of time)