Win Shares: Finding Clutch Performers

[weldone]

No data for pre-2001?

 

[viriya]

No data for pre-2001?

No over-by-over commentary before that so attributing win shares isn't possible.

 

[Victor Ian]

Can you explain this for me Viriya?

cricrate | ODI #2632

I am at this card having seen your Win Share best Performances page and wondering what Bracken did that was so good to be second on there., cricrate | Best/Worst ODI Win Share Performances

How does Bracken get rated so highly to top Murali Kartik's effort who was actually in the winning side?

 

[RossTaylorsBox]

Yeah good catch, seems like a major flaw if that performance is 2nd.

 

[viriya]

Can you explain this for me Viriya?

cricrate | ODI #2632

I am at this card having seen your Win Share best Performances page and wondering what Bracken did that was so good to be second on there., cricrate | Best/Worst ODI Win Share Performances

How does Bracken get rated so highly to top Murali Kartik's effort who was actually in the winning side?

Yeah good catch, seems like a major flaw if that performance is 2nd.

Very nice catch - I've been spending the last few days trying to incorporate fielding win shares so haven't been able to really look at specific examples so this helps a lot.

This is actually an issue with how I calculate odds. When calculating chasing team odds, I was giving a lower and upper bound to the balls remaining. This is fine in theory, but what happened here is that near the end of the match there was very little data to go by so the odds flipped a few times which gave Bracken unwarranted win shares.

I'm putting in a fix that removes the lower bound for balls remaining near the end of a match - if India have 50 balls to get 10 runs every past instance of matches with <~50 remaining should be taken into account instead of something like >45 and <55.

 

[viriya]

Very nice catch - I've been spending the last few days trying to incorporate fielding win shares so haven't been able to really look at specific examples so this helps a lot.

This is actually an issue with how I calculate odds. When calculating chasing team odds, I was giving a lower and upper bound to the balls remaining. This is fine in theory, but what happened here is that near the end of the match there was very little data to go by so the odds flipped a few times which gave Bracken unwarranted win shares.

I'm putting in a fix that removes the lower bound for balls remaining near the end of a match - if India have 50 balls to get 10 runs every past instance of matches with <~50 remaining should be taken into account instead of something like >45 and <55.

This is fixed.

 

[viriya]

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

^ These are both done.
cricrate | Current Best/Worst ODI Win Shares
cricrate | Best/Worst ODI Win Share Careers
cricrate | Best/Worst ODI Win Share Performances

You can select Total/Batting/Bowling/Fielding from the drop-downs.

Fielding win share details:
- Direct hit win share changes are attributed to the fielder
- Great catch win share changes are attributed 50% to the fielder 50% to the bowler
- Dropped catch/missed stumping win share change is attributed to the fielder negatively with the opposite assigned to the bowler (for creating the chance). The end result is a zero sum.
- Runs saved/given away are attributed to the fielder positively/negatively.

Some interesting observations
- Great bowlers (Murali, McGrath, Pollock, Bond) seem to have the most impact in win shares - somewhat surprising since this is for ODIs and they can't make an impact for more than 10 overs.
- Trescothick is a surprise #1 in the career batting win share rankings
- Faulkner is clutch - we knew that already, win shares confirms it.
- Umar Akmal is both a liability with the bat and in the field but more so in the field
- Kamran Akmal is similar

 

[NUFAN]

Is nawaz at number 4 correct with the update? Doesn't seem correct.

 

[viriya]

Is nawaz at number 4 correct with the update? Doesn't seem correct.

Good one. Issue here is when I'm trying to figure out who won the game. One table has "U.A.E." the other has "United Arab Emirates" so I was assuming that this game wasn't won by UAE, which made the bowler in the last over get credit for a fake win

Only affects games involving UAE, USA or PNG. Putting in fix and rerunning.

 

[viriya]

Is nawaz at number 4 correct with the update? Doesn't seem correct.

Good one. Issue here is when I'm trying to figure out who won the game. One table has "U.A.E." the other has "United Arab Emirates" so I was assuming that this game wasn't won by UAE, which made the bowler in the last over get credit for a fake win

Only affects games involving UAE, USA or PNG. Putting in fix and rerunning.

This is fixed.

 

[weldone]

Should replace the word 'All-Time' with '2001-Present'.. 'All-Time' sounds a lot misleading in this case.

Also, unlike total runs and other things, a total measure sounds better in this case than an average measure...will indicate how many matches they won for their team basically

 

[viriya]

Should replace the word 'All-Time' with '2001-Present'.. 'All-Time' sounds a lot misleading in this case.

Good point.

Also, unlike total runs and other things, a total measure sounds better in this case than an average measure...will indicate how many matches they won for their team basically

I think it would be nice to see, it would just be Total aWS * Mat.

 

[weldone]

Good point.



I think it would be nice to see, it would just be Total aWS * Mat.

Yes I'm saying that the lists should be sorted by (Total aWS * Mat)

 

[viriya]

Yes I'm saying that the lists should be sorted by (Total aWS * Mat)

Would benefit players that have been around a long time too much.. I think it's worth showing that in a column, but my rating is a better way to sort (avg Total aWS + longevity bonus).

 

[viriya]

Replaced the unadjusted TotalWSAvg with the TotalWS sum (adjusted). Figured there was no point in keeping the unadjusted odds anyway.
http://www.cricrate.com/odi/winShares/career.php

 

[weldone]

Would benefit players that have been around a long time too much.. I think it's worth showing that in a column, but my rating is a better way to sort (avg Total aWS + longevity bonus).

Will only benefit players with +ve average, who have played for a long time...and rightfully so, imo.

 

[viriya]

Made a change to give fielders 12.5% of the credit for regulation catches (based on average drop rates). The bowler gets 87.5% of the win shares instead of 100%
cricrate | Career Ratings - ODI Win Shares

 

[weldone]

Made a change to give fielders 12.5% of the credit for regulation catches (based on average drop rates). The bowler gets 87.5% of the win shares instead of 100%
cricrate | Career Ratings - ODI Win Shares

Good. I have been thinking about this topic for a few days now. I believe that should also apply when rating bowlers in general, i.e. bowlers shouldn't get 100% credit for caught and stumped dismissals but they should get 100% credit for bowled, lbw etc.

 

[viriya]

Good. I have been thinking about this topic for a few days now. I believe that should also apply when rating bowlers in general, i.e. bowlers shouldn't get 100% credit for caught and stumped dismissals but they should get 100% credit for bowled, lbw etc.

Yea I thought about that too but I feel like it would be unfair for certain bowlers like swing bowlers who are focused on generating slip catches. In the overall scheme of things what's important is getting the guy out not whether he was bowled or caught. Giving bowlers 100% credit for caught dismissals doesn't take away from what the fielder did as long as it's not a zero sum situation like win shares.

Also forgot to mention I also give 5% of the credit of stumping dismissals to the keepers for win shares with the bowler getting 95% of the credit (based on some missed stumping rate numbers - 5% is actually higher than the actual miss rate).