CricViz reversed header logo
Insights

CricViz develops new batting and bowling metrics for the World Cup

CricViz Insight Power

Introduction

In preparation for the upcoming ICC Cricket World Cup, CricViz has developed three new bowling metrics – Threat, Control, and Variation – and three new batting metrics – Power, Attack, and Timing. Using a combination of traditional statistics, ball tracking data and CricViz’s state-of-the-art in-house models, these ratings decode players’ attributes and highlight their strengths and weaknesses. Keep an eye out for these metrics in the World Cup: they will accompany each new batter and bowler as part of their player profile. 

Bowling

Measures of threat and control are not new to cricket: strike rate (balls bowled per wicket) and economy rate (runs conceded per over) are two headline numbers studied by players, coaches, and fans alike that give an idea of threat and control respectively. However, both measures have the critical flaw of not accounting for all the variables that go into a bowler taking a wicket or conceding runs, such as the quality of the batter, the grounds played at, or the phase of the match. In short, they do not correctly represent the bowler’s intrinsic threat and control. 

There are two ways of dealing with this. The first option is to use ball tracking data to strip out as many other variables as possible and focus solely on the attributes of each ball bowled, such as speed, swing, line, and length. CricViz’s expected wickets (xW) and expected runs (xR) models take this data and assess the outcomes of similar historic balls to measure the threat and control of an individual ball. Aggregating these across all the balls bowled by a player in a specified timeframe yields an overall measure of their threat and control. Another method is to account for as many of these variables as possible and adjust the bowler’s strike rate and economy rate accordingly. Used by professional teams around the world to compare players and build title-winning teams, CricViz’s player projections models are perfectly placed for this task. 

Each option yields a separate measure of threat and control. Advanced machine learning algorithms then optimise the weighting between the expected runs/wickets model, player projections model, and traditional strike rate and economy rate statistics, and output a rating between 0 and 100 for Threat and Control. 

Bowlers’ variations in the modern game are becoming increasingly important, with seam and spin bowlers regularly employing a wide arsenal of deliveries to keep batters from settling into their innings, trick them into making poor shots, and prevent them from premeditating. However, to date, no widely used metric captures how good bowlers are at this aspect of the game. CricViz’s variation metric puts a number to this, in three steps. First, ball tracking data for every ball bowled by a player in a specified timeframe is collected, and the dispersion of different attributes (such as speed, amount of spin, and length) across these balls is calculated. Comparing these player-specific dispersions to the overall dispersion of these same attributes across all players gives a measure of relative dispersion by a bowler for a specific attribute of bowling. An algorithm then weights and scales the importance of each attribute’s dispersion to arrive at a final rating for Variation between 0 and 100.  

Batting

The Power, Attack, and Timing ratings are slightly different, zooming in on the batters’ shot selection, the quality of contact made in each shot, and the resulting runs scored. Made possible by their enhanced data collection infrastructure, CricViz has built advanced models that map the probability of each shot outcome (e.g. wicket, dot, single) based on the shot being played (e.g. pull, drive, forward defence) combined with the contact made by the batter (e.g. middled, edged, played and missed). For example, making a good connection on a reverse sweep is associated with a relatively high probability of scoring a boundary whilst top-edging the same shot yields a high chance of a wicket from being caught. 

Power is calculated by looking at batters’ ability to score a boundary when they make a good connection. This is done by first filtering down to only the batter’s shots where they make a good connection with the ball. Then, the outcome of the shot (whether a boundary or not) is compared to the probability that the shot chosen would be a boundary if a good connection is made. This reflects, for example, the ability of exceptionally powerful batters to score boundaries from shots that rarely yield boundaries, or that weaker batters might only score 2 or 3 runs off a shot that, if well-timed, most batters would score a boundary.

The Attack rating strips out the outcome of the batter’s shot and focuses on the aggressiveness of their shot selection. For each shot, the model calculates probabilities for the number of runs that are likely to be scored if good contact is made – in other words, the ‘intent’ of the shot choice. The higher this number, the more attacking a shot choice it is. Note that this metric does not look at the outcome of the shot. This delineation is useful in isolating the aggression of a batter away from how capable they are of translating that into runs – for example, many tailenders are very attacking batters and their Attack metric should reflect this, regardless of whether they are good or not.

Finally, Timing compares the probability of a shot outcome given the shot played and the connection made with the probability of a shot outcome in general. In doing so, Timing considers how difficult it is to make a good connection on a specific shot and get runs or avoid losing a wicket by playing that shot. For example, middling a forward defence is much easier than middling a reverse scoop. Furthermore, by aggregating across all shots by a batter, this metric corrects for the overall conservativeness of a batter’s shot choice. 

The final Power, Attack, and Timing ratings are calculated by aggregating the individual Power, Attack, and Timing ratings for each shot by a batter across the specified time period and then scaled between 0 and 100. Ratings for both the batting and bowling metrics can be interpreted according to the following scale:

Image 1

As part of their ongoing partnership with the ICC and in anticipation of the upcoming World Cup, CricViz have looked at every ball bowled in ODIs since the 2019 Cricket World Cup and calculated threat, control, and variation ratings for all bowlers set to appear in the upcoming contest. The top 5 players for each metric are:

image

Looking at the average rating of each country’s players in each metric illuminates each team’s strengths and weaknesses.

image 3

So, expect wickets to tumble when Australia or India bowl, powerful and aggressive batting from England, and plenty of variation from New Zealand’s bowlers.

Share the Post: