Thursday 29 April 2010

IPL 2010 - The Most Valuable Player

As anybody who has ever been a fan of any sports will tell you, heated debates over the merits and demerits of one's favourite players are an intrinsic part of following the sport. The debates centre around a lot of things, but the crux of several arguments is a "my guy has better numbers than your guy" line.

Having done a Most Valuable Player analysis after last year's IPL [Link], it seemed only logical to go one better this year, and tweak the formulas involved to produce a more comprehensive analysis this time around. After all, everything about IPL-3 has been bigger than IPL-2 - the crowds in the stadiums, the noise, the number of advertisement breaks, and the off-field drama.

Thus here is the analysis of IPL-3's Most Valuable Player.

Just as equations off the field in the corridors of the BCCI have gotten more complex and unfathomable, so I found this year, my equations for calculating the Batting and Bowling indexes for players getting more complex. There was one key difference though: I did not foresee any problem in actually explaining my equations.

Let us start with the batting. Batting, by its nature lends itself to greater manipulation by numbers because there exists a proportional relationship in the numbers one measures and the performances associated with them. Thus the more runs you score or the higher your strike rate - the better you have performed.

Amongst the simplest ways of measuring a Batting Index for players is to simply multiply the runs scored by the strike-rate. While this satisfies the basic criteria that one must account for while measuring a Batting Index in a limited overs game - runs scored and the speed with which runs are scored - it is a little too simple, and offers potential for tweaking.

The Batting Index for this rating, is calculated not by taking strike-rates into account, but taking relative strike-rates. The relative strike-rate of a batsman is simply his strike-rate divided by the tournament's average strike-rate. Thus if the tournament's average strike-rate has been 100, and Batsman A scores his runs at a strike-rate of 120, his relative strike-rate will be 1.2. If Batsman B scores his runs at a strike-rate of 90, his relative strike-rate will be 0.90.

When one uses relative strike-rates, it is easier to see batsmen who have scored at above-par rates getting benefited, while those who have been slow get penalized. However, in a Twenty20 format, the strike-rate is an extremely important measurement criterion and needs to be given added weightage. A score of 25 off 15 balls can often swing a match towards the team, while a score of 40 off 38 balls, while being numerically superior, will not help the team as much.

While increasing the weightage of the strike-rate, due care must be exercised so that rewards and penalties are automatically in-built into whatever formula is used, so that a Keiron Pollard who has scored at much above the normal scoring rates gets doubly benefited, but a Jacques Kallis who has scored at much below the normal scoring rate gets doubly penalized. The way to do this is by using the square of the relative strike-rate.

Go back to the example earlier of Batsmen A and B. Their relative strike-rates were 1.2 and 0.9. The square of their relative strike-rates will thus be 1.44 and 0.81. In one stroke, the man who has scored faster than par, gets a higher coefficient (1.44), while the man who has scored runs slower, gets a lower coefficient (0.81).

Thus the number of runs a batsman has scored will be multiplied by the square of the relative strike rate.

The only problem with squaring relative strike-rates is that it reduces the importance of runs scored very drastically. Therefore a new factor is added to the mix: the relative average. Multiplying still further by the relative average rewards the batsmen who have been consistent, since scoring runs is the basic unit of measurement that batsmen must excel at.

For the purposes of calculating the relative average, each Not Out for a batsman was considered as adding 10 runs to his total. The figure of 10 is a rounded one, arrived at after studying when the batsmen remained not out, and how often they did so.

The complete formula for calculating the number of points a batsman has got is thus:
Runs Scored * Relative Average * Relative Strike Rate squared.

For IPL-3, the tournament strike rate has been 126.76, while the tournament average (after accounting for Not Outs as mentioned above) has been 21.29.

It will, of course, come as no surprise to people to see who tops the list of batsmen in IPL-3, but it is still interesting to see who the other men are who make up the list of top batsmen. For the purpose of brevity, only the top-20 batsmen are shown in the list below:

RankPlayerTeamRunsBallsStrike RateBatting Points
1Sachin TendulkarMumbai618466132.621351.30
2Suresh RainaChennai520364142.861105.07
3Murali VijayChennai458292156.851049.53
4Mahela JayawardenePunjab439298147.321004.66
5Robin UthappaBangalore374218171.56905.52
6Jacques KallisBangalore572494115.79857.42
7Sourav GangulyKolkata493419117.66716.77
8Virender SehwagDelhi356218163.30705.65
9Chris GayleKolkata292184158.70697.40
10Yusuf PathanRajasthan333201165.67673.62
11Saurabh TiwaryMumbai419309135.60644.05
12Kumar SangakkaraPunjab357257138.91599.04
13Kevin PietersenBangalore236157150.32592.31
14Kieron PollardMumbai273147185.71576.00
15Ambati RayuduMumbai356246144.72569.72
16Rohit SharmaDeccan404302133.77560.02
17Andrew SymondsDeccan429341125.81556.95
18Naman OjhaRajasthan377285132.28546.80
19Shane WatsonRajasthan185114162.28526.91
20David WarnerDelhi282191147.64476.98

It is instructive to note that even though Jacques Kallis has been the second highest scorer in the IPL, his batting rank is 6. Robin Uthappa, who has scored nearly 200 runs less than Kallis has pipped him because of his fantastic strike-rate. The achievements of Shane Watson and Kevin Pietersen in making the list are also note-worthy, since they both had far fewer inning to play than the others, and had they batted more, would have ended up probably near the top of the table.

Bowling points are relatively harder to fix than batting points, since the bowlers have an inversely proportional relationship between their figures and how well they have done. Thus, a bowler has done better if his strike rate and economy rate are lower. The lower the measures, the better the bowler - which is the complete opposite of a batsman, in that while batting the higher the number of runs scored, the strike rate, or the average the better the batsman has done.

Therefore, for bowling a straight multiplication does not work. There are also several more parameters to bowling than batting, with any numbers-based analysis bound to account for wickets taken, economy rate and strike rate.

For wickets taken, the way to assign points is fairly simple - each wicket is worth the tournament average. That is, if across the IPL, the bowlers conceded 3000 runs while taking 100 wickets, each wicket would be worth 30 points. The tournament average for IPL 2010 has been 29.38, making each wicket worth that many points.

The next thing to consider is the strike rate. Just as was done for batting, the relative strike rate is used for bowling, with the relationship inversed.

That is to say, if the tournament's strike rate for bowling has been 20, and Bowler A has picked up 12 wickets in 50 overs bowled, while Bowler B has picked up 15 wickets in 40 overs bowled, their relative strike rates will be as follow:


Bowler A has a strike rate of 25 (300 balls bowled divided by 12 wickets), which would give him a relative strike rate of 0.8 (20 divided by 25), while Bowler B has a strike rate of 16 (240 divided by 15), which would give him a relative strike rate of 1.25 (20 divided by 16). Thus instead of dividing the bowler's individual strike rates by the tournament strike rate, for the Bowling Index, the tournament strike rate is divided by the individual bowlers' strike rates, thus ensuring that the lower the strike rate of a bowler, the higher the value of his relative strike rate.
The same principle is followed for economy rates, with the tournament average economy rate being divided by the bowlers' individual economy rates to arrive at the relative economy rate. With Twenty20's emphasis on giving as few runs as possible, the relative economy rate is then squared, following the same principle as the batting strike rate.

In IPL 2010, the tournament bowling strike rate has been 22.25, while the tournament economy rate has been 7.92.

The bowler, then, will get his points thus: (No. of Wickets taken * 29.38 * relative strike rate) + (No. of balls bowled * relative economy rate squared).

The relative economy rate is multiplied by the number of balls bowled for two reasons:

a) It ensures that a part-time bowler who might have bowled just 4-5 overs for very few runs does not get an undue advantage over a regular bowler who has bowled a lot more, under more trying and varying conditions,

and

b) It is of a greater credit to a bowler to sustain a given economy rate for a greater number of overs. For example, if Bowler A has bowled 6 overs and conceded 30 runs, his achievement is not as noteworthy as Bowler B who has bowled 15 overs and conceded 75 runs. They both have the same economy rate (5 runs an over), but Bowler B has been better because he has been able to sustain it longer.

However, this formula remains a tad incomplete, because it sometimes unfairly rewards bowlers who might have been very poor with their economy rates, but got to bowl a lot of overs simply because there weren't other bowlers in the team.

To correct this anomaly an additional parameter was added, the idea for which came from Dr. Srinivas Bhogle, who had done a similar ratings exercise. Many thanks to him for letting me use it in my analysis.

The correction factor is this: Take the number of balls bowled by a bowler, and based on the tournament economy rate, calculate the number of runs he ought to have conceded if he was bowling at par. Thus if Bowler A has bowled 20 overs (120 balls), he ought to have conceded 160 runs (if the tournament economy rate is rounded to 8 runs an over to simplify the example). Let us call this figure the par-runs conceded. Now take the actual amount of runs that Bowler A has conceded and subtract this figure from his par-runs conceded. Therefore if Bowler A has bowled his 20 overs for just 140 runs, he would get +20 points, while if he has been expensive and conceded 190 runs from his 20 overs, he would get -30 points.

This correction factor goes a large way towards making the Bowling Points more comprehensive. All that remains is to multiply the bowling points of each bowler by a constant, so that they are on the same scale as the batting points.

The list of the top bowlers in IPL 2010 is:

RankPlayerTeamOversWicketsStrike RateEcon. RateBowling Points
1Pragyan OjhaDeccan58.52116.817.29776.53
2Anil KumbleBangalore63.21722.356.43714.86
3Amit MishraDelhi531718.716.85657.98
4Harbhajan SinghMumbai53.31718.887.05636.38
5R AshwinChennai481322.156.10584.39
6MuralitharanChennai481519.206.85578.70
7Dale SteynBangalore591523.606.88578.36
8Kieron PollardMumbai371514.807.41571.92
9Lasith MalingaMumbai491519.607.02561.67
10Ryan HarrisDeccan30.41413.147.60553.93
11Zaheer KhanMumbai48.21519.337.78498.07
12Vinay KumarBangalore46.11617.318.58495.18
13Doug BollingerChennai311215.506.68492.95
14Andrew SymondsDeccan531226.507.02457.90
15Shadab JakatiChennai381317.547.66451.45
16Irfan PathanPunjab46.21518.539.19413.51
17RP SinghDeccan421418.008.81412.54
18Piyush ChawlaPunjab491224.507.49409.76
19Chaminda VaasDeccan22914.676.32393.60
20Siddharth TrivediRajasthan35.31119.367.32392.31
21Murali KarthikKolkata39926.006.49385.89

The first thing immediately apparent from the table is that spinners have ruled the roost, with all top-6 spots going to them. Dale Steyn is the highest ranked pacer at Number 7, though Ryan Harris and Doug Bollinger could have potentially gotten many more points if they had been available throughout the tournament instead of just in the latter half.

The list included 21 names, because Kolkata's highest ranked bowler - Murali Karthik - makes an appearance at that number. The lack of a strike bowling option for Kolkata has been pointed at before in this article, but it comes out starkly when an analysis is done.

In passing, Anil Kumble showed once again, just what India took for granted when he was playing, and what they are missing when he's retired, with his outstanding show.

Pragyan Ojha is the deserving table-topper though, and the table above merely reinforces the fact that the selectors erred big time by selecting Piyush Chawla for the T20 World Cup instead of Ojha, Mishra or even R Ashwin.

Now all that remains is for the Batting and Bowling Indexes to be combined, adding the fielding points and arriving at the Most Valuable Player. However, before arriving at the final list, it is worth noting what this analysis does not measure.

The ratings do not take into account any subjective criteria, for instance innings played under pressure. This is because it is impossible to quantify pressure. Is it harder to bat coming in at 20/3 in 3 overs against a Chennai attack or at 34/4 in 7 overs against a Kolkata attack? Similarly for bowling, there is no additional weightage given to performance in power-play overs or overs 16-20, simply because in those overs, although batsmen do score more runs, more wickets are also lost, and by contrast in overs 7-16, although runs may be scored at a relatively slower pace, fewer wickets will be lost.

Neither are points awarded for captaincy, since there is no way to judge what percentage of a win has come about by captaincy, or indeed, even whether every captaincy decision that bears fruit or backfires is respectively the result of great/flawed thinking.

The other thing the ratings don't measure at present is the direct hit run-outs. While it would be good to include that, run-outs are sadly not kept track of in any compilation of statistics and to keep a track of which fielders were involved in a run-out, I'd have to make notes during each match - which is practically not possible. The same goes for runs saved while fielding. A Raina or an AB de Villiers will be worth more to their sides than the runs they score simply because of the runs they prevent. But until scoreboards start reflecting the runs saved/conceded alongwith the batting and bowling figures, this is also difficult to incorporate.

With that in mind, the fielding points have been given for catches and stumpings, with each catch/stumping fetching 15 points.

With these points added to the tally that players have accumulated with their batting and bowling, it is now possible to put forth the final list of the Most Valuable Player of IPL 2010.


RankPlayerTeamBatting PointsBowling PointsFielding PointsTotal Points
1Suresh RainaChennai1105.07205.571501460.64
2Sachin TendulkarMumbai1351.300.00451396.30
3Jacques KallisBangalore857.42370.691051333.11
4Kieron PollardMumbai576.00571.92901237.92
5Murali VijayChennai1049.530.001651214.53
6Andrew SymondsDeccan556.95457.901801194.85
7Mahela JayawardenePunjab1004.660.00751079.66
8Robin UthappaBangalore905.520.001501055.52
9Yusuf PathanRajasthan673.62212.801351021.42
10Virender SehwagDelhi705.65198.3330933.97
11Irfan PathanPunjab445.45413.5160918.96
12Harbhajan SinghMumbai164.42636.3890890.80
13Pragyan OjhaDeccan0.07776.5360836.59
14Sourav GangulyKolkata716.771.06105822.83
15Christ GayleKolkata697.4091.9830819.37
16Rohit SharmaDeccan560.0279.98135775.00
17Kumar SangakkaraPunjab599.040.00165764.04
18Anil KumbleBangalore0.58714.8645760.45
19Shane WatsonRajasthan526.91175.4445747.35
20Kevin PietersenBangalore592.31100.1245737.43


Even though Sachin Tendulkar towered over the others with his supreme batting, the all-round skills of Suresh Raina - who scored less than Tendulkar but whose strike rate was 10 points above that of the Little Master, besides picking up 6 wickets and 10 catches - have pipped the batting skills of Tendulkar, and Raina has emerged as the worthy Most Valuable Player.

It seems only fitting that the MVP should belong to the winning team. Also remember, that though there is no way to quantify pressure, it feels right that the man who has out-performed all others in the biggest match of the IPL (the final), is also t
The table showcases a nice blend of people who have brought multiple skills to the field - such as Kieron Pollard, Jacques Kallis, Andrew Symonds, Robin Uthappa (remember he kept wickets for Bangalore) - and those who have made largely by virtue of a single skill, such as Tendulkar, Mahela Jayawardene, Pragyan Ojha and Anil Kumble.



Is this a final ratings to end all arguments about who was the best player in IPL 2010? Of course not. This is merely a numbers-based guide. But it is a useful tool if you want to settle an argument over who was the better player, since it lets you say, "My guy has better numbers than your guy!"

Photos courtesy CricBuzz. This article has also been published on CricBuzz in three parts (Part 1, Part 2, Part 3).

4 comments:

Offshore ASP .Net Developers said...

Hey I am agree with you. It is a very nice post. thanks a lot for this interesting post. Keep up the good work.

Vasu said...

Great analysis...I like your methodology, and while no methodology would be 100% foolproof, what you have adopted is pretty good.

It is interesting to know that, even a simplistic methodology, shows up results that appear consistent to what a good cricketing brain will come up with, without the aid of numbers.

Suresh Raina at the top is a great example. Everyone went on and on about Sachin, and much as I admire him, and his IPL performance this year, I think Raina deserved the Man of the series ahead of Sachin. Raina came in the middle order, and hence less chance to score as much as Sachin, struck at a far better strike rate then Sachin, again batting outside power plays, took some outstanding catches, and played a terrific knock in the finals, when it mattered the most.

At least happy he came out on top in your analysis :)

Another example is Ganguly. A lot of people raved about his aggregate, but in reality, his slow batting affected his team badly. So to a neutral observer, he is not one of the top 2-3 performers in the IPL.

Also like how you kept subjective things out, as thee are key factors, but it typically leads to meaningless discussions where everyone plays up / down players according to their biases.

Great stuff!

Cheers!
Vasu

http://vasusworld.wordpress.com/

Saurabh Somani said...

@Vasu: thanks, glad you liked the analysis. :)

Onion Insights said...

Hi,

Nice post! For this wonderful read, we present you with a unique opportunity of mystery shopping. Onion Insights Pvt. Ltd. is a Global Customer Experience Measurement (CEM) company that uses the tool of Mystery Shopping wherein one poses as a customer to a particular product or service and anonymously evaluates his/her shopping experience and provides a well-written structured report on the findings. The assignment lets you combine two of your favourite things – shopping and writing!

Being a Mystery Shopper not only helps you to become a Better Customer, you also get paid for shopping! You will be paid a nominal fee for your time and efforts and additionally reimbursed for the products you purchase to complete the Mystery Shop. To become a Mystery Shopper, you just have to be of a minimum age of 18, have good command over English and have Internet accessibility.

To begin with, register with us for free at www.onioninsights.info as New Mystery Shopper and start your journey into the world of Mystery Shopping!

Warm Regards,
Onion Insights