There is a lot to not like IPL. There are also a few things that I like about a club based league. A dual round-robin home/away format is the obvious candidate. Combining players from across the globe is another. Above all it allows rookies to ply their trade in company of the veterans. I have maintained an IPL league table for the year 2012, 2013 & 2014 based on Ball Difference. Recently I noticed that Ball Difference page was referred by Wikipedia. The revision history shows:
Net run rate
In Apr 2012, user Mmitchell10 added a new section ‘Alternatives to NRR‘ to the NRR page where BD appears along with D/L, Average of the Match NRRs etc. That page does not refer to Computations for Ball Difference which clearly explains how to compute this value for an interrupted match. I conceived BD as an alternative to NRR during 2011 World Cup. Over the years, it has become the basic building block of my evolving player performance program. There are 100 points for every completed match. The points for each team are decided based on Ball Difference. In case of a tie both teams get 50 points. In case of a victory off the last delivery, the winning team earns a little over 50 points based on my ‘Points for 1 Run’ metric. After that an equation determines whether the match was won by batting side or bowling side. If both then margin for each unit is quantified. Team points are divided as Team Batting, Bowling and Fielding points. Batting points are distributed based on runs scored and balls used. Bowling points are based on overs bowled, runs conceded, number of wickets taken and when a wicket was taken. Fielding points are based on number of catches, stumpings and run outs. Instead of tabulating the results for each match, I create a chart and publish it on Twitter:
All charts for IPL2014 matches are here. 8 teams have played 14 matches each and it is time to represent the summary tables by team to supplement the match contribution tables. There is a vanilla and a value-added table for each team. The vanilla table is a bar chart representing total points earned by each player over 14 matches. It is ordered by typical batting position.
Blue bars represent batting points and red ones are for bowling. A typical winning combination will show some tall blue bars either for openers or the top middle order. Most of the runs should be scored by top 6. KXIP won 11 out of 14 matches where late order generally did not have to bat. In case 9, 10 & Jack were required to bat, they did not achieve much as can be seen by the absence of blue bars on the right hand side. The value added chart shows the same information on secondary axis. The value added primary axis is reserved for average points per match. A black box with two whiskers is nothing but the Open, High, Low & Close chart for stocks. The four values used are Q1, Q3, G & µ. These 4 values were discussed in – Vijay Merchant’s G and Cyril Walters’ µ – my earlier post about dealing fairly with not outs in computing batting averages. Maxwell was the top performer with the tallest bar. He contributed heavily in a quarter of matches. The Q3 value represented by the top whisker is above 14 and stretches beyond the chart. He also failed in a quarter of matches contributing less than 3 points which is depicted by the lower whisker or the Q1 value. He earned above 108 points in 13 matches which means that he averaged 8.3 points per match. This is the Arithmetic Mean (µ) represented by the top of the box. But his Geometric Mean G is less than half at 3.5 which indicates that his performance varied a lot between matches. Vohra who played only 5 matches scored 40.7 points. His G & µ are much closer at 7.3 and 8.1 which indicates better consistency. A team needs an in-form batsman, bowler and all-rounder. KKR found these in Uthappa, Narine and Shakib. Yusuf Pathan who scored over 50 points in 13 matches, earned most of those in a single match. His blitzkrieg in the 14th match ensured that KKR finish above CSK giving them two chances to qualify for the final instead of winning two play-off matches to reach the final. An inform batsman may add 100 or even more runs in a match with no restriction on the number of deliveries. But bowlers are restricted to 24 deliveries. This means that a very good performance by a bowler is worth only 10-15 points but a batsman can earn 20 or more in a match. This also means that the gap between G & µ will be wider between a batsman and a bowler. Notice the narrow box for Narine and compare it with the middle sized box for Shakib. Uthappa did remarkably well as a consistent batsman finishing with the highest value of G. Yet the difference between G & µ is wider than other two. 5 bowlers must bowl in a match where 2-3 batsmen can score all th runs. That is the nature of a limited overs match. Smith, Jadeja, Ashwin, Raina and Mohit Sharma appeared in all 14 matches and performed above average in respective roles. Smith was their star performer and a little more contribution from the top order would have helped. Ashwin was very consistent similar to Narine – see how close G & µ are in his case. He took only 14 wickets compared to Jadeja’s 17 and Mohit’s 19. But the equations used in this model account for economy with respect to match figures. Hence he gets 76.5 bowling points compared to 67 and 64 points by Jadeja and Sharma. The tallest bar is only 80 points and that too from a bowler. This indicates that MI batsmen struggled to assert themselves. Just like Yusuf, Corey Anderson scored most of his batting points in a single but vital match. All four teams had a star bowler who was consistent in his 24 ball spell. Harbhajan Singh’s contribution was lower in terms of overall points because his team did not do well enough. Had MI won more matches or improved on margin of victory, he would have earned more bowling points for exactly same bowling performance. His case highlights that this model emphasises us over me. Multiple players in a team performing well improve the total contribution for each performer – see Uthappa, Shakib and Narine above. 1 ball separated MI and RR which means not much separates the two teams. No one scored 80 points and the best pure batsman got less than 60 batting points. Bowlers can do only so much with the 4 over restriction so the top order must get in and then get going to consistently win matches. Tambe was the best bowler for RR though less consistent than the other 4 discussed so far. But his achievements are of the highest order. He truly outperformed everyone else when we account for his age and international experience. Bhuvnesh Kumar was the best bowler till the 13th round – Narine went past him after contrasting performances by these two in the final round. Steyn and Karn Sharma did very well to support him. SRH did not qualify because of the lack of blue colour in the middle of the chart. Warner did not get much support. If FInch and Dhawan had occupied the crease longer then things may have been different. Gayle is the Bradman of Club T20. Perhaps not because he had such a poor series. Chahal was the standout performer for RCB in a team of superstars along with Starc who contributed a little with bat too. Yuvraj Singh was exceptinal at times and was the inspiraton to revive this IPL performance model after more than half the matches were concluded.
A G value around 2 points per match on the left side of chart underlines why RCB campaign failed. Yuvraj and de Villiers were better in the middle order but more T20 matches are won when top 3 contribute heavily.
Duminy was the Shakib for DD but he did not get help from a star batsman and a star bowler. Shami was fairly consistent for DD. Unfortunately he was consistently average. He did what was par but not as much as was expected of him.