Win Shares: Finding Clutch Performers

[viriya]

I'm thinking of implementing a win shares measure in cricket - not necessarily similar to the one in basketball even though the I'm borrowing the name.

I can already compute odds for each team after each over based on historical matches - this is used for live odds based on current match situation. The idea is to attribute changes in win odds to the players involved in an over. As an example, say Steve Smith and James Faulkner are batting and Vettori is bowling with Australia at 220/5 at 40 overs chasing 305. The odds as it stands are Aus 50% and NZ 50%. Steve Smith proceeds to take Vettori apart for 15 in the 41st over, and the odds change to Aus 60% and NZ 40%. In this case Smith will be attributed +10% in win shares while Vettori will get -10%. If Faulkner got 5 of the 15 runs scored in that over, he gets +3.33% (5/15) while Smith gets +6.67%.

After a completed match (assuming no tie/NR), the winning team's players will have a sum total of +100% win shares while the losing team will have -100%. Clutch performers who win games late such as Dhoni would get a big portion of the win shares in close games. Say a game is 50-50 going into the 50th over and Dhoni gets 10 runs to win the game on his own. He will be attributed +50% in win shares while the bowler gets -50%.

Over the course of a career, I would be able to calculate the winners and losers and clutch performers. The shares attributed would probably need to be adjusted based on team strengths - so a +20% win shares vs Bangladesh isn't equated to be the same as vs Australia.

Any ideas/suggestions/comments appreciated.

 

[hendrix]

cool idea.

I'm not too keen on the win = + 100 and loss = -100 but that's how it has to work really, and it just means you hae to understand the stat.

 

[Adders]

I must be missing something here.

So the maximum team total is +100% win shares, and if Dohni scores 10 in the last over to win the game he'll get 50% of that.......leaving only 50% left for the rest of the side?? What if Kohli has made a hundred earlier in the innings and (as unlikely as this may sound) an Indian bowler has taken a 5fer to restrict the total required??

50% for MSD to polish off the innings seems extraordinarily high...........or was that just a loose example?

 

[viriya]

I must be missing something here.

So the maximum team total is +100% win shares, and if Dohni scores 10 in the last over to win the game he'll get 50% of that.......leaving only 50% left for the rest of the side?? What if Kohli has made a hundred earlier in the innings and (as unlikely as this may sound) an Indian bowler has taken a 5fer to restrict the total required??

50% for MSD to polish off the innings seems extraordinarily high...........or was that just a loose example?

Kohli might have something like +25%.. that Indian bowler might have +20%.. the rest of the Indian team might have 5%. It only has to add up to 100% - Dhoni won them the match and gets the bulk.

This can be an issue possibly where finishers benefit significantly - I think this will mean that when comparing win shares you have to compare like for like - Dhoni would not be a good comparison for Kohli, Sanga would be.

 

[viriya]

That 5% for the rest of the team can have negatives and positives - just because the win % has to add up to +100% doesn't mean that Yadav doesn't get -15% after going for 75 off his 10 overs.. it just means that there's someone else that has a positive to counter that.

 

[Dan]

How does this take into account who has overs left? 85 in 10 overs changes from a 50/50 proposition based on whether you've got Southee and Corey Anderson bowling death overs compared to Ishant and Bhuvi, for example.

 

[Burgey]

Y U do this viriya? Y?

 

[viriya]

How does this take into account who has overs left? 85 in 10 overs changes from a 50/50 proposition based on whether you've got Southee and Corey Anderson bowling death overs compared to Ishant and Bhuvi, for example.

It wouldn't. But I'd imagine any win shares you get for a game will be adjusted by the opposition strength (say opposition team ranking) so that a +10% vs India's bowling is differentiated from +10% vs NZ's.

 

[Burgey]

I actually thought this thread was about a competition to win some BHP shares or something. I dunno.

 

[viriya]

It wouldn't. But I'd imagine any win shares you get for a game will be adjusted by the opposition strength (say opposition team ranking) so that a +10% vs India's bowling is differentiated from +10% vs NZ's.

This is a good point, I'll probably work on getting the odds calculator a little more robust by taking team strength and momentum into account before diving into the win shares implementation.

 

[hendrix]

It wouldn't. But I'd imagine any win shares you get for a game will be adjusted by the opposition strength (say opposition team ranking) so that a +10% vs India's bowling is differentiated from +10% vs NZ's.

Nah, this is unnecessary.

Win shares shouldn't be adjusted by the strength of the opposition. Just the actual contribution.

 

[viriya]

Nah, this is unnecessary.

Win shares shouldn't be adjusted by the strength of the opposition. Just the actual contribution.

It wouldn't be necessary if the odds calculator already takes that into account which it doesn't currently. At the moment it doesn't so it tends to be inaccurate vs minnow teams especially.

 

[Dan]

I actually thought this thread was about a competition to win some BHP shares or something. I dunno.

Would be nice if they gave away some better shares than BHP tbh.

 

[Riggins]

Would be nice if they gave away some better shares than BHP tbh.

Just about the best time to get them imo.

 

[viriya]

Actually I should clarify something - if done properly it won't be +100% for the winning team and -100% for the losing. It would be dependent on what the team odds were at the start of the game.

So for the SL vs Ban game coming up tomorrow, SL starts with 80% win odds, so if SL end up winning, SL players will only be able to gain +20% win shares total. Bangladesh however, if they pull off a win can game +80% win shares (-80% for SL players in that case).

 

[Riggins]

I actually really like this concept, and have thought about similar ideas. If you can accurately generate the odds taking into account the opposition strength I think it would be a really worthwhile statistic, especially over the course of a tournament like the WC.


One thing, I wouldn't represent the numbers as a percentage. I feel like it's pretty misleading. Just represent it as a decimal. For example there is one win available for each match and one loss (obviously. If you increase the odds by 10% or whatever, you gain 0.1 win shares. I think generating decimals is easier to follow for people who don't necessarily know the methodology and also easier to sum to keep track of wins added across a period of time.

 

[hendrix]

Actually I should clarify something - if done properly it won't be +100% for the winning team and -100% for the losing. It would be dependent on what the team odds were at the start of the game.

So for the SL vs Ban game coming up tomorrow, SL starts with 80% win odds, so if SL end up winning, SL players will only be able to gain +20% win shares total. Bangladesh however, if they pull off a win can game +80% win shares (-80% for SL players in that case).

This is wrong.
A player's impact on the match doesn't change by the strength of an opposition. You can leave further variables to alter that. But have the fundamental value still there.
I.e. we can record runs scored as runs scored. Further down the line if we want to analyse we can weigh runs scored vs opposition with different values. I think the same concept should be applied here.

The "winningness" of the innings is defined by the performance.

 

[viriya]

I actually really like this concept, and have thought about similar ideas. If you can accurately generate the odds taking into account the opposition strength I think it would be a really worthwhile statistic, especially over the course of a tournament like the WC.


One thing, I wouldn't represent the numbers as a percentage. I feel like it's pretty misleading. Just represent it as a decimal. For example there is one win available for each match and one loss (obviously. If you increase the odds by 10% or whatever, you gain 0.1 win shares. I think generating decimals is easier to follow for people who don't necessarily know the methodology and also easier to sum to keep track of wins added across a period of time.

Yes it would be in decimal format - I went with % here to make it easier to connect with actual odds.

 

[viriya]

This is wrong.
A player's impact on the match doesn't change by the strength of an opposition. You can leave further variables to alter that. But have the fundamental value still there.
I.e. we can record runs scored as runs scored. Further down the line if we want to analyse we can weigh runs scored vs opposition with different values. I think the same concept should be applied here.

The "winningness" of the innings is defined by the performance.

The "pure" value would not take that into account. But the one that takes starting odds into consideration would be more valuable in evaluating players. Otherwise minnow-bashers might be overrated. I'll have both numbers as separate stats which will give more information.

 

[hendrix]

The "pure" value would not take that into account. But the one that takes starting odds into consideration would be more valuable in evaluating players. Otherwise minnow-bashers might be overrated. I'll have both numbers as separate stats which will give more information.

cool.