HOW IT WORKS: THE ‘ALGORITHM’ CHOOSING BETWAY SA20 GROUP STAGE PLAYERS OF THE MATCH

Betway SA20 has partnered with CricViz to use their MVP data science model, a performance evaluation measure (aka ‘The Algorithm’ by the commentators), to assess the performance of players in each match and Season 2 in its entirety.

MVP allows comparisons of player performance across different roles and phases of the game. The model uses CricViz’s Match Impact model which takes into account a wide range of factors to determine how valuable a performance is. The MVP determines what an expected outcome would be for each delivery and then evaluates the realised score against that.

“We at Betway SA20 are always striving to be at the cutting edge of what’s happening in cricket, which is why we decided to embrace analytics to help us choose our individual player awards throughout the season,” says Graeme Smith, League Commissioner.

To give a more rounded picture of a player’s performance in a match, CricViz’s MVP model calculates the expected score considering the following factors:

WHERE IS THE MATCH BEING PLAYED?

The first thing accounted for is the ground where the match is being played, as this has a large effect on the value of a given innings. An inning at Centurion is expected to have a higher strike rate than that at Boland Park in Paarl for example. Similarly, Wanderers is the ground where we see the highest wickets per ball.

 The MVP model accounts for this by rewarding players who play faster innings more at lower scoring grounds and will reward wicket-takers at grounds with a lower wicket-taking threat. This feature allows for players playing in very different conditions to be fairly compared to each other.

EXPECTED RUNS AND WICKET PROBABILITY PER GROUND

Centurion | 1.40 expected runs | 4.6% wicket probability

Wanderers | 1.21 expected runs | 4.9% wicket probability

Newlands | 1.20 expected runs | 4.2% wicket probability

Kingsmead | 1.18 expected runs | 4.6% wicket probability

Boland Park | 1.17 expected runs | 3.7% wicket probability

St George's Park | 1.14 expected runs | 3.9% wicket probability

THE VALUE OF PACE VS SPIN

In general, in white ball cricket, pace bowling is the more aggressive bowling type, it takes wickets more frequently but goes at a higher economy rate. This effect varies by ground, for example at St George’s Park in Gqeberha we see that spin is much easier to bowl (historically), with it taking more wickets and being less expensive, whereas at Centurion spin is very expensive as well as being less threatening. The MVP scoring model accounts for this in its expected score calculations, this allows bowlers playing in different conditions to be compared, a spin bowler bowling in favourable conditions will not be overly rewarded due purely to the conditions they play in.

WHAT STATE IS THE GAME AT?

The other factor that the model accounts for is the game situation, this is the over number and the wickets lost at that point in the match. These factors have two effects on the MVP score, the first being that it affects the expected score.

This allows innings that occur at different times in the game to be more easily compared. A strike rate of 140 is excellent in the early middle overs but would be a poor performance in the final overs. 

The game state also allows CricViz MVP to account for the value of a wicket, wickets in the Powerplay at the start of the game are worth significantly more than at the end of the match.  

The MVP model accounts for this for both batters and bowlers. Bowlers who take the majority of their wickets at the death will see less benefit from this as the wickets have less of an effect on the match. Similarly, wicket retention is a more valuable skill for opening batters as the value of their wicket is significantly higher. This allows the scoring system to compare players who operate in different phases of the match to determine which had the higher overall value.

OTHER CONSIDERATIONS

The MVP scoring system also rewards a player's fielding with points awarded for catches, run-outs and stumpings. Additionally, players are rewarded for winning the match with the captain of the winning team receiving an additional bonus.

HOW OFTEN DOES THE MVP MODEL ANALYSE PLAYER SCORES?

The model runs the above calculations on every delivery which allows us to give an MVP score for an individual match and also across a whole competition. These numbers can be displayed as leaderboards for the whole competition or on a match-by-match basis – both of which South Africa’s premier T20 league have adopted for Season 2.

Our Player Rankings leaderboards are keeping a running total of the players as the season unfolds.