Win Shares: Finding Clutch Performers

[viriya]

I'm interested to see aggregated results over a few matches - because although some people are saying that finishers get benefitted from these, I'm interested to see if in the long-run that kinda offsets (because finishers get large penalties for failed chases too).

I ran it for all WC matches up to now then setup an AvgWS measure. Here are the top 20 MVPs/LVPs in terms of win shares for players that have played >5 games in this WC:

Best AvgWS in WC 2015:

Player AvgWS
Daniel Vettori 0.122
Mahmudullah 0.095
Martin Guptill 0.084
Chris Gayle 0.078
Mohammed Shami 0.076
AB de Villiers 0.076
David Warner 0.069
Tendai Chatara 0.067
Tillakaratne Dilshan 0.067
Shikhar Dhawan 0.066
Shaiman Anwar 0.058
Misbah-ul-Haq 0.055
David Miller 0.055
Brendan Taylor 0.051
Aaron Finch 0.049
Mushfiqur Rahim 0.044
Faf du Plessis 0.042
Dale Steyn 0.041
Steven Smith 0.039
Soumya Sarkar 0.038

Worst AvgWS in WC 2015:

Player AvgWS
Nawroz Mangal -0.105
Tinashe Panyangara -0.103
Sohail Khan -0.088
Stuart Broad -0.077
Ahmed Shehzad -0.076
Kevin O'Brien -0.064
Thisara Perera -0.064
Taskin Ahmed -0.059
Matthew Cross -0.058
Josh Davey -0.058
Tamim Iqbal -0.057
Amjad Ali -0.055
Sean Williams -0.052
Kyle Coetzer -0.052
Hamid Hassan -0.051
Mohammad Nabi -0.038
Afsar Zazai -0.036
Quinton de Kock -0.036
Preston Mommsen -0.034
Swapnil Patil -0.033

 

[viriya]

A few interesting picks from the above. Dave Warner is a surprise in the best list since it doesn't seem like he did much except for his ton. I think this is a function of the WS measure favoring openers slightly because they have the opportunity to affect the starting 50-50 odds more than others. In the worst list, Josh Davey shows up - even though he was momentarily the top wicket-taker in the tournament. I think this makes sense though since his wickets were not impactful and he was expensive.

 

[Athlai]

Yeah it looks a bit like a random number generated two lists of names there, might need some work.

 

[viriya]

Yeah it looks a bit like a random number generated two lists of names there, might need some work.

Uh, how does it look random? Aside from a few names like I pointed the others look on point.

 

[weldone]

Why are you starting with 50-50 odds? Is it too much of work if you try starting from the real starting odds? That will correct those 2 lists to an extent. (e.g. Warner, Tamim)

 

[NUFAN]

A couple of questions on this interesting exercise: do run outs or great fielding help the fielder and/or the bowler?

Is it easy to break the contributions through batting and bowling? Sean Williams looks odd on the list above, but I'm curious what his batting and bowling looks like broken up.

 

[viriya]

Why are you starting with 50-50 odds? Is it too much of work if you try starting from the real starting odds? That will correct those 2 lists to an extent. (e.g. Warner, Tamim)

It's not that simple. Say I start with 80-20 odds for a NZ vs Ban game. NZ bats first and Baz gets out first over. Odds would be 45-55 if you ignored teams. What are the right odds now? I have to figure out a way to "adjust" the right odds properly first before implementgin pre-adjusted starting odds or Baz gets -0.35 just for getting out early.

 

[viriya]

A couple of questions on this interesting exercise: do run outs or great fielding help the fielder and/or the bowler?

Is it easy to break the contributions through batting and bowling? Sean Williams looks odd on the list above, but I'm curious what his batting and bowling looks like broken up.

I haven't implemented fieldingWS yet but I should be able to since I've already done the fielding stats exercise using commentary. I've left that for last till I resolve all other major issues and run a first 10 year or so run without issues.

Right now run outs are being assigned to the bowler which is obviously wrong.

I already have battingWS and bowlingWS separated - just showing the totalWSAvg in the last list.
Sean Williams has +0.01 battingWSAvg and -0.06 bowlingWSAvg.

 

[weldone]

It's not that simple. Say I start with 80-20 odds for a NZ vs Ban game. NZ bats first and Baz gets out first over. Odds would be 45-55 if you ignored teams. What are the right odds now? I have to figure out a way to "adjust" the right odds properly first before implementgin pre-adjusted starting odds or Baz gets -0.35 just for getting out early.

yes you have to find a way to adjust that, obviously (betting sites do that)

 

[marc71178]

Uh, how does it look random? Aside from a few names like I pointed the others look on point.

So where's the man of the series then?

 

[viriya]

Finally ran the first iteration for all possible ODIs (~2001 onwards). Here are the top 20 best and worst in win share averages for players that have played 50+ ODIs during the period:

Best Win Share Averages:

Player Matches TotalWSAvg
Hashim Amla 108 0.075
Glenn McGrath 109 0.074
Muttiah Muralitharan 159 0.072
Shikhar Dhawan 59 0.064
Saeed Ajmal 104 0.063
Adam Gilchrist 163 0.061
Sunil Narine 51 0.060
Shane Bond 79 0.056
Jason Gillespie 72 0.056
Shane Watson 180 0.055
Virender Sehwag 222 0.052
Marcus Trescothick 93 0.052
Sachin Tendulkar 178 0.051
Virat Kohli 146 0.050
David Warner 59 0.045
Alastair Cook 90 0.041
Graeme Swann 74 0.038
Gautam Gambhir 139 0.037
Michael Hussey 155 0.036
Chris Gayle 223 0.035

Worst Win Share Averages:

Player Matches TotalWSAvg
Tapash Baisya 53 -0.093
RP Singh 57 -0.070
Sreesanth 51 -0.069
Sohail Tanvir 59 -0.066
Chamu Chibhabha 60 -0.064
Anil Kumble 54 -0.061
Tim Bresnan 83 -0.058
Darren Gough 62 -0.055
Dwayne Bravo 158 -0.053
Habibul Bashar 78 -0.052
Ishant Sharma 71 -0.050
Daryl Tuffey 71 -0.048
Dion Ebrahim 60 -0.046
Dwayne Smith 97 -0.044
Harbhajan Singh 139 -0.043
Naved-ul-Hasan 72 -0.043
Asad Shafiq 51 -0.038
Mohammad Ashraful 151 -0.038
Elton Chigumbura 153 -0.036
Clint McKay 57 -0.036

Hashim Amla's win share average of 0.075 says that on average, he increases his team's winning odds by 7.5%. Negative win share odds point to players who decrease their team's chances of winning.

Next steps:
- Adjust starting odds to reflect team ratings and adjust match odds accordingly for a more fair match to match comparison of win shares (currently all matches start at 50-50)
- Separate fielding win shares for run outs etc (currently the out is credited to the bowler).. also great catches/dropped catches can be identified and valued

 

[Debris]

I knew it. Shane Watson is better than Tendulkar. Respect.

 

[Daemon]

Gough decreases his teams chance of winning :o

 

[NUFAN]

Gillespie the big surprise for me.

 

[viriya]

Gough decreases his teams chance of winning :o

Gillespie the big surprise for me.

One result of every game starting at 50-50 odds in the current iteration is that opening batsmen/bowlers have more of the win shares attributed to them (because it's easier to change your team's win odds from 50 to 60 or from 50 to 40 compared to 65 to 75). This seems to affect Gillespie positively but negatively for Gough. A lot of the batsmen in the best list are openers as well - Trescothick and Cook are examples of names that you wouldn't expect to be right up there.

I'm currently rerunning with starting team odds taken into consideration (Aus vs Afg match doesn't start at 50-50 but more like 75-25), so being the opening batsmen/bowler won't make as much of a difference.

 

[RossTaylorsBox]

Not sure if I'm interpreting it correctly, but would the win shares for every player add to zero?

 

[viriya]

Not sure if I'm interpreting it correctly, but would the win shares for every player add to zero?

In a single game? You would think yes, with the winning team's players getting +0.5 and the losing team -0.5 (in the current setup where starting odds are 50-50), but because of extras that's actually not the case.

Once I implement team odd adjustments (aus 75 vs afg 25 starting odds for example) the less favored team's players will have a chance to gain more win shares than the other team. If Afghanistan win vs Australia their players can cumulatively get up to +0.75 but Australian players can only gain +0.25 for the win. This will make it so that comparing win shares between matches becomes more meaningful.

I'll be storing both unadjusted and adjusted win shares - it will at least tell you who the minnow-bashers are.

 

[weldone]

Next steps:
- Adjust starting odds to reflect team ratings and adjust match odds accordingly for a more fair match to match comparison of win shares (currently all matches start at 50-50)
- Separate fielding win shares for run outs etc (currently the out is credited to the bowler).. also great catches/dropped catches can be identified and valued

looking forward to this

 

[viriya]

looking forward to this

- Adjust starting odds to reflect team ratings and adjust match odds accordingly for a more fair match to match comparison of win shares (currently all matches start at 50-50)

^ This is done, I'm trying to get it all ready to just present it all in the site.

 

[viriya]

First implementation is live:

cricrate | ODI Win Shares
cricrate | Current Best/Worst ODI Win Shares
cricrate | Best/Worst ODI Win Share Careers
cricrate | Best/Worst ODI Win Share Performances

WS values are with 50-50 starting odds for all matches, AdjWS values are with odds adjusted based on team current ratings. AdjWS makes more sense imo.

Currently FieldingWS is ignored and attributed to the bowler, only major step left is to separate that properly.