Cricket Data - Explanation of Concepts

Skype: Sports Analytics Advantage

There have been many attempts to grade batsmen in the various formats in cricket, and the T20 format is no different.

The ICC, as the ruling authority of cricket, have their own rankings here, but these rankings are far from perfect given that they only take into account T20 internationals, and not the various T20 domestic competitions throughout the world.

Taking into account T20 domestic matches should be mandatory, as it has a variety of benefits:-

1) It allows for a much bigger sample of results for individual players and therefore avoids scenarios such as Hamilton Masakadza being graded as better than AB de Villiers, as the ICC current rankings at 28/7/16 suggest.

Not many cricket fans would grade AB de Villiers as a worse T20 batsman than Hamilton Masakadza, as the ICC rankings suggest...

2) It gives the opportunity for players without international experience to be assessed, because, as the data repeatedly illustrates, international selectors are far from perfect.  A further advantage of this is that this allows us to work out how these non-international players are likely to perform on the international stage.  It would also allow T20 teams to find hidden talent which they could pick up for relatively cheap sums, similar to the 'Moneyball' approach in US sport.  In all sports, marginal gains are key and data such as this will help teams achieve this.

3) Not all formats/opponents are equal.  It would have been a much easier project to complete if I had just compiled data on domestic and international matches and worked out player averages and strike rates from these, but this would have been largely flawed given that not all formats and opponents are equal.

Without giving away too much of my calculations, I can use an example of how I approached grading players.  An average player who has played in both the T20 Blast (in England) and the IPL (in India), will average a multiplier of 0.86 in the IPL than they would in the T20 Blast.  

So, for example, a player who averages 30.00 in the T20 Blast, will be expected to average 25.80 (30.00*0.86) in the IPL, because historical data has shown that batting in the T20 Blast is easier.  Therefore, their batting performance in the T20 Blast is likely to make them look a better player than if they had played in the IPL.

In fact, my research showed that batting in the T20 Blast was the easiest of all the major domestic T20 leagues in the world.  The average T20 Blast player will see their average reduced by a multiplication factor of 0.82 across a standardised worldwide mean, and their strike rate (runs per 100 balls) will be reduced by a factor of 0.92.  Either the bowling in the T20 Blast is extremely weak, or batting conditions are extremely good - a combination of the two is possible, and quite likely.

There are many examples of strong T20 Blast players failing abroad.  Ryan ten Doeschate is a good example - from 1/1/2014 to 24/7/16, ten Doeschate averaged 38.23 in T20 Blast innings with a strike rate of 143.52 - certainly very impressive data indeed.  However, from innings abroad in franchise leagues in this period of time, he averaged just 17.00 with a strike rate of 112.89.

Ryan ten Doeschate has performed superbly in the T20 Blast in recent years, but in other tournaments around the world has struggled...

Grading T20 performances in the major domestic leagues and T20 internationals enabled me to build a comprehensive database of player abilities, including adjusted averages for each player given the difficulty of each innings that they played.  

For overall quality, multiple of mean differential was used.  

So, for example, at the time of writing, Virat Kohli's adjusted average was 62.30 (very similar to his actual average of 61.51) and this was 2.49x the overall T20 time period mean of 25.06 runs per wicket.  

In addition, we could do the same for strike rate.  A mean time period strike rate of 134.50 was average for T20 matches, and Kohli's adjusted strike was 133.55, which was 0.99x the T20 mean strike rate.

Therefore, the average player would have an average mean differential of 1 (they'd score at an average of 25.06 and at 134.50 strike rate), whereas Kohli's was ((2.49+.99)/2) = 1.74, showing him to be significantly better than average, indeed, world class.