cricketr plays the ODIs!


Published in R bloggers: cricketr plays the ODIs

Introduction

In this post my package ‘cricketr’ takes a swing at One Day Internationals(ODIs). Like test batsman who adapt to ODIs with some innovative strokes, the cricketr package has some additional functions and some modified functions to handle the high strike and economy rates in ODIs. As before I have chosen my top 4 ODI batsmen and top 4 ODI bowlers.

Do check out my interactive Shiny app implementation using the cricketr package – Sixer – R package cricketr’s new Shiny avatar

You can also read this post at Rpubs as odi-cricketr. Dowload this report as a PDF file from odi-cricketr.pdf

Take a look at my book with all my articles related to cricket –  Cricket analytics with cricketr!!!. The book is also available in paperback and kindle versions at Amazon  which has, by the way,  better formatting!

Note: If you would like to do a similar analysis for a different set of batsman and bowlers, you can clone/download my skeleton cricketr template from Github (which is the R Markdown file I have used for the analysis below). You will only need to make appropriate changes for the players you are interested in. Just a familiarity with R and R Markdown only is needed.

Batsmen

  1. Virendar Sehwag (Ind)
  2. AB Devilliers (SA)
  3. Chris Gayle (WI)
  4. Glenn Maxwell (Aus)

Bowlers

  1. Mitchell Johnson (Aus)
  2. Lasith Malinga (SL)
  3. Dale Steyn (SA)
  4. Tim Southee (NZ)

I have sprinkled the plots with a few of my comments. Feel free to draw your conclusions! The analysis is included below

The profile for Virender Sehwag is 35263. This can be used to get the ODI data for Sehwag. For a batsman the type should be “batting” and for a bowler the type should be “bowling” and the function is getPlayerDataOD()

The package can be installed directly from CRAN

if (!require("cricketr")){ 
    install.packages("cricketr",lib = "c:/test") 
} 
library(cricketr)

or from Github

library(devtools)
install_github("tvganesh/cricketr")
library(cricketr)

The One day data for a particular player can be obtained with the getPlayerDataOD() function. To do you will need to go to ESPN CricInfo Player and type in the name of the player for e.g Virendar Sehwag, etc. This will bring up a page which have the profile number for the player e.g. for Virendar Sehwag this would be http://www.espncricinfo.com/india/content/player/35263.html. Hence, Sehwag’s profile is 35263. This can be used to get the data for Virat Sehwag as shown below

sehwag <- getPlayerDataOD(35263,dir="..",file="sehwag.csv",type="batting")

Analyses of Batsmen

The following plots gives the analysis of the 4 ODI batsmen

  1. Virendar Sehwag (Ind) – Innings – 245, Runs = 8586, Average=35.05, Strike Rate= 104.33
  2. AB Devilliers (SA) – Innings – 179, Runs= 7941, Average=53.65, Strike Rate= 99.12
  3. Chris Gayle (WI) – Innings – 264, Runs= 9221, Average=37.65, Strike Rate= 85.11
  4. Glenn Maxwell (Aus) – Innings – 45, Runs= 1367, Average=35.02, Strike Rate= 126.69

Plot of 4s, 6s and the scoring rate in ODIs

The 3 charts below give the number of

  1. 4s vs Runs scored
  2. 6s vs Runs scored
  3. Balls faced vs Runs scored

A regression line is fitted in each of these plots for each of the ODI batsmen A. Virender Sehwag

par(mfrow=c(1,3))
par(mar=c(4,4,2,2))
batsman4s("./sehwag.csv","Sehwag")
batsman6s("./sehwag.csv","Sehwag")
batsmanScoringRateODTT("./sehwag.csv","Sehwag")

sehwag-4s6sSR-1

dev.off()
## null device 
##           1

B. AB Devilliers

par(mfrow=c(1,3))
par(mar=c(4,4,2,2))
batsman4s("./devilliers.csv","Devillier")
batsman6s("./devilliers.csv","Devillier")
batsmanScoringRateODTT("./devilliers.csv","Devillier")

devillier-4s6SR-1

dev.off()
## null device 
##           1

C. Chris Gayle

par(mfrow=c(1,3))
par(mar=c(4,4,2,2))
batsman4s("./gayle.csv","Gayle")
batsman6s("./gayle.csv","Gayle")
batsmanScoringRateODTT("./gayle.csv","Gayle")

gayle-4s6sSR-1

dev.off()
## null device 
##           1

D. Glenn Maxwell

par(mfrow=c(1,3))
par(mar=c(4,4,2,2))
batsman4s("./maxwell.csv","Maxwell")
batsman6s("./maxwell.csv","Maxwell")
batsmanScoringRateODTT("./maxwell.csv","Maxwell")

maxwell-4s6sout-1

dev.off()
## null device 
##           1

Relative Mean Strike Rate

In this first plot I plot the Mean Strike Rate of the batsmen. It can be seen that Maxwell has a awesome strike rate in ODIs. However we need to keep in mind that Maxwell has relatively much fewer (only 45 innings) innings. He is followed by Sehwag who(most innings- 245) also has an excellent strike rate till 100 runs and then we have Devilliers who roars ahead. This is also seen in the overall strike rate in above

par(mar=c(4,4,2,2))
frames <- list("./sehwag.csv","./devilliers.csv","gayle.csv","maxwell.csv")
names <- list("Sehwag","Devilliers","Gayle","Maxwell")
relativeBatsmanSRODTT(frames,names)

plot-1-1

Relative Runs Frequency Percentage

Sehwag leads in the percentage of runs in 10 run ranges upto 50 runs. Maxwell and Devilliers lead in 55-66 & 66-85 respectively.

frames <- list("./sehwag.csv","./devilliers.csv","gayle.csv","maxwell.csv")
names <- list("Sehwag","Devilliers","Gayle","Maxwell")
relativeRunsFreqPerfODTT(frames,names)

plot-2-1

Percentage of 4s,6s in the runs scored

The plot below shows the percentage of runs made by the batsmen by ways of 1s,2s,3s, 4s and 6s. It can be seen that Sehwag has the higheest percent of 4s (33.36%) in his overall runs in ODIs. Maxwell has the highest percentage of 6s (13.36%) in his ODI career. If we take the overall 4s+6s then Sehwag leads with (33.36 +5.95 = 39.31%),followed by Gayle (27.80+10.15=37.95%)

Percent 4’s,6’s in total runs scored

The plot below shows the contrib

frames <- list("./sehwag.csv","./devilliers.csv","gayle.csv","maxwell.csv")
names <- list("Sehwag","Devilliers","Gayle","Maxwell")
runs4s6s <-batsman4s6s(frames,names)

plot-46s-1

print(runs4s6s)
##                Sehwag Devilliers Gayle Maxwell
## Runs(1s,2s,3s)  60.69      67.39 62.05   62.11
## 4s              33.36      24.28 27.80   24.53
## 6s               5.95       8.32 10.15   13.36
 

Runs forecast

The forecast for the batsman is shown below.

par(mfrow=c(2,2))
par(mar=c(4,4,2,2))
batsmanPerfForecast("./sehwag.csv","Sehwag")
batsmanPerfForecast("./devilliers.csv","Devilliers")
batsmanPerfForecast("./gayle.csv","Gayle")
batsmanPerfForecast("./maxwell.csv","Maxwell")

swcr-perf-1

dev.off()
## null device 
##           1

3D plot of Runs vs Balls Faced and Minutes at Crease

The plot is a scatter plot of Runs vs Balls faced and Minutes at Crease. A prediction plane is fitted

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
battingPerf3d("./sehwag.csv","V Sehwag")
battingPerf3d("./devilliers.csv","AB Devilliers")

plot-3-1

dev.off()
## null device 
##           1
par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
battingPerf3d("./gayle.csv","C Gayle")
battingPerf3d("./maxwell.csv","G Maxwell")

plot-4-1

dev.off()
## null device 
##           1

Predicting Runs given Balls Faced and Minutes at Crease

A multi-variate regression plane is fitted between Runs and Balls faced +Minutes at crease.

BF <- seq( 10, 200,length=10)
Mins <- seq(30,220,length=10)
newDF <- data.frame(BF,Mins)

sehwag <- batsmanRunsPredict("./sehwag.csv","Sehwag",newdataframe=newDF)
devilliers <- batsmanRunsPredict("./devilliers.csv","Devilliers",newdataframe=newDF)
gayle <- batsmanRunsPredict("./gayle.csv","Gayle",newdataframe=newDF)
maxwell <- batsmanRunsPredict("./maxwell.csv","Maxwell",newdataframe=newDF)

The fitted model is then used to predict the runs that the batsmen will score for a hypotheticial Balls faced and Minutes at crease. It can be seen that Maxwell sets a searing pace in the predicted runs for a given Balls Faced and Minutes at crease followed by Sehwag. But we have to keep in mind that Maxwell has only around 1/5th of the innings of Sehwag (45 to Sehwag’s 245 innings). They are followed by Devilliers and then finally Gayle

batsmen <-cbind(round(sehwag$Runs),round(devilliers$Runs),round(gayle$Runs),round(maxwell$Runs))
colnames(batsmen) <- c("Sehwag","Devilliers","Gayle","Maxwell")
newDF <- data.frame(round(newDF$BF),round(newDF$Mins))
colnames(newDF) <- c("BallsFaced","MinsAtCrease")
predictedRuns <- cbind(newDF,batsmen)
predictedRuns
##    BallsFaced MinsAtCrease Sehwag Devilliers Gayle Maxwell
## 1          10           30     11         12    11      18
## 2          31           51     33         32    28      43
## 3          52           72     55         52    46      67
## 4          73           93     77         71    63      92
## 5          94          114    100         91    81     117
## 6         116          136    122        111    98     141
## 7         137          157    144        130   116     166
## 8         158          178    167        150   133     191
## 9         179          199    189        170   151     215
## 10        200          220    211        190   168     240

Highest runs likelihood

The plots below the runs likelihood of batsman. This uses K-Means It can be seen that Devilliers has almost 27.75% likelihood to make around 90+ runs. Gayle and Sehwag have 34% to make 40+ runs. A. Virender Sehwag

A. Virender Sehwag

batsmanRunsLikelihood("./sehwag.csv","Sehwag")

smith-1

## Summary of  Sehwag 's runs scoring likelihood
## **************************************************
## 
## There is a 35.22 % likelihood that Sehwag  will make  46 Runs in  44 balls over 67  Minutes 
## There is a 9.43 % likelihood that Sehwag  will make  119 Runs in  106 balls over  158  Minutes 
## There is a 55.35 % likelihood that Sehwag  will make  12 Runs in  13 balls over 18  Minutes

B. AB Devilliers

batsmanRunsLikelihood("./devilliers.csv","Devilliers")

warner-1

## Summary of  Devilliers 's runs scoring likelihood
## **************************************************
## 
## There is a 30.65 % likelihood that Devilliers  will make  44 Runs in  43 balls over 60  Minutes 
## There is a 29.84 % likelihood that Devilliers  will make  91 Runs in  88 balls over  124  Minutes 
## There is a 39.52 % likelihood that Devilliers  will make  11 Runs in  15 balls over 21  Minutes

C. Chris Gayle

batsmanRunsLikelihood("./gayle.csv","Gayle")

cook,cache-TRUE-1

## Summary of  Gayle 's runs scoring likelihood
## **************************************************
## 
## There is a 32.69 % likelihood that Gayle  will make  47 Runs in  51 balls over 72  Minutes 
## There is a 54.49 % likelihood that Gayle  will make  10 Runs in  15 balls over  20  Minutes 
## There is a 12.82 % likelihood that Gayle  will make  109 Runs in  119 balls over 172  Minutes

D. Glenn Maxwell

batsmanRunsLikelihood("./maxwell.csv","Maxwell")

oot-1

## Summary of  Maxwell 's runs scoring likelihood
## **************************************************
## 
## There is a 34.38 % likelihood that Maxwell  will make  39 Runs in  29 balls over 35  Minutes 
## There is a 15.62 % likelihood that Maxwell  will make  89 Runs in  55 balls over  69  Minutes 
## There is a 50 % likelihood that Maxwell  will make  6 Runs in  7 balls over 9  Minutes

Average runs at ground and against opposition

A. Virender Sehwag

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
batsmanAvgRunsGround("./sehwag.csv","Sehwag")
batsmanAvgRunsOpposition("./sehwag.csv","Sehwag")

avgrg-1-1

dev.off()
## null device 
##           1

B. AB Devilliers

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
batsmanAvgRunsGround("./devilliers.csv","Devilliers")
batsmanAvgRunsOpposition("./devilliers.csv","Devilliers")

avgrg-2-1

dev.off()
## null device 
##           1

C. Chris Gayle

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
batsmanAvgRunsGround("./gayle.csv","Gayle")
batsmanAvgRunsOpposition("./gayle.csv","Gayle")

avgrg-3-1

dev.off()
## null device 
##           1

D. Glenn Maxwell

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
batsmanAvgRunsGround("./maxwell.csv","Maxwell")
batsmanAvgRunsOpposition("./maxwell.csv","Maxwell")

avgrg-4-1

dev.off()
## null device 
##           1

Moving Average of runs over career

The moving average for the 4 batsmen indicate the following

1. The moving average of Devilliers and Maxwell is on the way up.
2. Sehwag shows a slight downward trend from his 2nd peak in 2011
3. Gayle maintains a consistent 45 runs for the last few years

par(mfrow=c(2,2))
par(mar=c(4,4,2,2))
batsmanMovingAverage("./sehwag.csv","Sehwag")
batsmanMovingAverage("./devilliers.csv","Devilliers")
batsmanMovingAverage("./gayle.csv","Gayle")
batsmanMovingAverage("./maxwell.csv","Maxwell")

sdgm-ma-1

dev.off()
## null device 
##           1

Check batsmen in-form, out-of-form

  1. Maxwell, Devilliers, Sehwag are in-form. This is also evident from the moving average plot
  2. Gayle is out-of-form
checkBatsmanInForm("./sehwag.csv","Sehwag")
## *******************************************************************************************
## 
## Population size: 143  Mean of population: 33.76 
## Sample size: 16  Mean of sample: 37.44 SD of sample: 55.15 
## 
## Null hypothesis H0 : Sehwag 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : Sehwag 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "Sehwag 's Form Status: In-Form because the p value: 0.603525  is greater than alpha=  0.05"
## *******************************************************************************************
checkBatsmanInForm("./devilliers.csv","Devilliers")
## *******************************************************************************************
## 
## Population size: 111  Mean of population: 43.5 
## Sample size: 13  Mean of sample: 57.62 SD of sample: 40.69 
## 
## Null hypothesis H0 : Devilliers 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : Devilliers 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "Devilliers 's Form Status: In-Form because the p value: 0.883541  is greater than alpha=  0.05"
## *******************************************************************************************
checkBatsmanInForm("./gayle.csv","Gayle")
## *******************************************************************************************
## 
## Population size: 140  Mean of population: 37.1 
## Sample size: 16  Mean of sample: 17.25 SD of sample: 20.25 
## 
## Null hypothesis H0 : Gayle 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : Gayle 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "Gayle 's Form Status: Out-of-Form because the p value: 0.000609  is less than alpha=  0.05"
## *******************************************************************************************
checkBatsmanInForm("./maxwell.csv","Maxwell")
## *******************************************************************************************
## 
## Population size: 28  Mean of population: 25.25 
## Sample size: 4  Mean of sample: 64.25 SD of sample: 36.97 
## 
## Null hypothesis H0 : Maxwell 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : Maxwell 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "Maxwell 's Form Status: In-Form because the p value: 0.948744  is greater than alpha=  0.05"
## *******************************************************************************************

Analysis of bowlers

  1. Mitchell Johnson (Aus) – Innings-150, Wickets – 239, Econ Rate : 4.83
  2. Lasith Malinga (SL)- Innings-182, Wickets – 287, Econ Rate : 5.26
  3. Dale Steyn (SA)- Innings-103, Wickets – 162, Econ Rate : 4.81
  4. Tim Southee (NZ)- Innings-96, Wickets – 135, Econ Rate : 5.33

Malinga has the highest number of innings and wickets followed closely by Mitchell. Steyn and Southee have relatively fewer innings.

To get the bowler’s data use

malinga <- getPlayerDataOD(49758,dir=".",file="malinga.csv",type="bowling")

Wicket Frequency percentage

This plot gives the percentage of wickets for each wickets (1,2,3…etc)

par(mfrow=c(1,4))
par(mar=c(4,4,2,2))
bowlerWktsFreqPercent("./mitchell.csv","J Mitchell")
bowlerWktsFreqPercent("./malinga.csv","Malinga")
bowlerWktsFreqPercent("./steyn.csv","Steyn")
bowlerWktsFreqPercent("./southee.csv","southee")

relBowlFP-1

dev.off()
## null device 
##           1

Wickets Runs plot

The plot below gives a boxplot of the runs ranges for each of the wickets taken by the bowlers. M Johnson and Steyn are more economical than Malinga and Southee corroborating the figures above

par(mfrow=c(1,4))
par(mar=c(4,4,2,2))

bowlerWktsRunsPlot("./mitchell.csv","J Mitchell")
bowlerWktsRunsPlot("./malinga.csv","Malinga")
bowlerWktsRunsPlot("./steyn.csv","Steyn")
bowlerWktsRunsPlot("./southee.csv","southee")

wktsrun-1

dev.off()
## null device 
##           1

Average wickets in different grounds and opposition

A. Mitchell Johnson

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
bowlerAvgWktsGround("./mitchell.csv","J Mitchell")
bowlerAvgWktsOpposition("./mitchell.csv","J Mitchell")

gr-1-1

dev.off()
## null device 
##           1

B. Lasith Malinga

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
bowlerAvgWktsGround("./malinga.csv","Malinga")
bowlerAvgWktsOpposition("./malinga.csv","Malinga")

gr-2-1

dev.off()
## null device 
##           1

C. Dale Steyn

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
bowlerAvgWktsGround("./steyn.csv","Steyn")
bowlerAvgWktsOpposition("./steyn.csv","Steyn")

gr-3-1

dev.off()
## null device 
##           1

D. Tim Southee

par(mfrow=c(1,2))
par(mar=c(4,4,2,2))
bowlerAvgWktsGround("./southee.csv","southee")
bowlerAvgWktsOpposition("./southee.csv","southee")

avgrg-4-1

dev.off()
## null device 
##           1

Relative bowling performance

The plot below shows that Mitchell Johnson and Southee have more wickets in 3-4 wickets range while Steyn and Malinga in 1-2 wicket range

frames <- list("./mitchell.csv","./malinga.csv","steyn.csv","southee.csv")
names <- list("M Johnson","Malinga","Steyn","Southee")
relativeBowlingPerf(frames,names)

relBowlPerf-1

Relative Economy Rate against wickets taken

Steyn had the best economy rate followed by M Johnson. Malinga and Southee have a poorer economy rate

frames <- list("./mitchell.csv","./malinga.csv","steyn.csv","southee.csv")
names <- list("M Johnson","Malinga","Steyn","Southee")
relativeBowlingERODTT(frames,names)

relBowlER-1

Moving average of wickets over career

Johnson and Steyn career vs wicket graph is on the up-swing. Southee is maintaining a reasonable record while Malinga shows a decline in ODI performance

par(mfrow=c(2,2))
par(mar=c(4,4,2,2))
bowlerMovingAverage("./mitchell.csv","M Johnson")
bowlerMovingAverage("./malinga.csv","Malinga")
bowlerMovingAverage("./steyn.csv","Steyn")
bowlerMovingAverage("./southee.csv","Southee")

jmss-bowlma-1

dev.off()
## null device 
##           1

Wickets forecast

par(mfrow=c(2,2))
par(mar=c(4,4,2,2))
bowlerPerfForecast("./mitchell.csv","M Johnson")
bowlerPerfForecast("./malinga.csv","Malinga")
bowlerPerfForecast("./steyn.csv","Steyn")
bowlerPerfForecast("./southee.csv","southee")

jsba-pfcst-1

dev.off()
## null device 
##           1

Check bowler in-form, out-of-form

All the bowlers are shown to be still in-form

checkBowlerInForm("./mitchell.csv","J Mitchell")
## *******************************************************************************************
## 
## Population size: 135  Mean of population: 1.55 
## Sample size: 15  Mean of sample: 2 SD of sample: 1.07 
## 
## Null hypothesis H0 : J Mitchell 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : J Mitchell 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "J Mitchell 's Form Status: In-Form because the p value: 0.937917  is greater than alpha=  0.05"
## *******************************************************************************************
checkBowlerInForm("./malinga.csv","Malinga")
## *******************************************************************************************
## 
## Population size: 163  Mean of population: 1.58 
## Sample size: 19  Mean of sample: 1.58 SD of sample: 1.22 
## 
## Null hypothesis H0 : Malinga 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : Malinga 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "Malinga 's Form Status: In-Form because the p value: 0.5  is greater than alpha=  0.05"
## *******************************************************************************************
checkBowlerInForm("./steyn.csv","Steyn")
## *******************************************************************************************
## 
## Population size: 93  Mean of population: 1.59 
## Sample size: 11  Mean of sample: 1.45 SD of sample: 0.69 
## 
## Null hypothesis H0 : Steyn 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : Steyn 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "Steyn 's Form Status: In-Form because the p value: 0.257438  is greater than alpha=  0.05"
## *******************************************************************************************
checkBowlerInForm("./southee.csv","southee")
## *******************************************************************************************
## 
## Population size: 86  Mean of population: 1.48 
## Sample size: 10  Mean of sample: 0.8 SD of sample: 1.14 
## 
## Null hypothesis H0 : southee 's sample average is within 95% confidence interval 
##         of population average
## Alternative hypothesis Ha : southee 's sample average is below the 95% confidence
##         interval of population average
## 
## [1] "southee 's Form Status: Out-of-Form because the p value: 0.044302  is less than alpha=  0.05"
## *******************************************************************************************

***************

Key findings

Here are some key conclusions ODI batsmen

  1. AB Devilliers has high frequency of runs in the 60-120 range and the highest average
  2. Sehwag has the most number of innings and good strike rate
  3. Maxwell has the best strike rate but it should be kept in mind that he has 1/5 of the innings of Sehwag. We need to see how he progress further
  4. Sehwag has the highest percentage of 4s in the runs scored, while Maxwell has the most 6s
  5. For a hypothetical Balls Faced and Minutes at creases Maxwell will score the most runs followed by Sehwag
  6. The moving average of indicates that the best is yet to come for Devilliers and Maxwell. Sehwag has a few more years in him while Gayle shows a decline in ODI performance and an out of form is indicated.

ODI bowlers

  1. Malinga has the highest played the highest innings and also has the highest wickets though he has poor economy rate
  2. M Johnson is the most effective in the 3-4 wicket range followed by Southee
  3. M Johnson and Steyn has the best overall economy rate followed by Malinga and Steyn 4 M Johnson and Steyn’s career is on the up-swing,Southee maintains a steady consistent performance, while Malinga shows a downward trend

Hasta la vista! I’ll be back!
Watch this space!

Also see my other posts in R

  1. Introducing cricketr! : An R package to analyze performances of cricketers
  2. cricketr digs the Ashes!
  3. A peek into literacy in India: Statistical Learning with R
  4. A crime map of India in R – Crimes against women
  5. Analyzing cricket’s batting legends – Through the mirage with R
  6. Mirror, mirror . the best batsman of them all?

You may also like

  1. A closer look at “Robot Horse on a Trot” in Android
  2. What’s up Watson? Using IBM Watson’s QAAPI with Bluemix, NodeExpress – Part 1
  3. Bend it like Bluemix, MongoDB with autoscaling – Part 2
  4. Informed choices through Machine Learning : Analyzing Kohli, Tendulkar and Dravid
  5. TWS-4: Gossip protocol: Epidemics and rumors to the rescue
  6. Deblurring with OpenCV:Weiner filter reloadedhttp://www.r-bloggers.com/cricketr-plays-the-odis/
Advertisements

8 thoughts on “cricketr plays the ODIs!

  1. Pingback: cricketr adapts to the Twenty20 International! | Giga thoughts …

  2. Hi Ganesh,

    When I tried downloading the ODI stats for Viru and AB using the commands you have explained above, I don’t seem to get the complete list of games played by both. For example, You have shown that Viru has played 245 innings and scored 8586 runs. When I use the same commands, I get a total of 169 innings, with total runs of 5823.

    According to my list, the last game Viru played was on 03-Jan-13, which agrees with whats on statsguru. But some games are missing. For example, statsguru says there was a game with the following details: v West Indies Bulawayo 30-Jun-01 ODI # 1731. That doesn’t show up in the downloaded list. The downloaded list also misses the next 3 games. So I am not sure whats happening. Have you come across such discrepancies during the development? Please let me know.

    • Srisai – Thanks for pointing this out. I will check and fix this over the weekend. I think I have an idea what the problem is. ODIs probably have matches that end in tie or no result. Will fix if needed. – Ganesh

  3. Pingback: Introducing cricketr! : An R package to analyze performances of cricketers | Giga thoughts …

  4. Pingback: Sixer – R package cricketr’s new avatar | Giga thoughts …

  5. Pingback: The making of cricket package yorkr – Part 2 | Giga thoughts …

  6. Pingback: cricketr plays the ODIs! | Giga thoughts … – Unstable Contextuality Research

  7. Pingback: The making of cricket package yorkr – Part 2 – Mubashir Qasim

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s