Statistics

liquefry
Posts: 339
Joined: Wed May 18, 2016 10:10 am

Statistics

Postby liquefry » Wed Aug 14, 2019 2:32 am

Sorry this is a super long post. The TL;DR is some new stats I'd like to suggest that better reflect performance in limited overs than the traditional average/strike rate measures, and allow for all rounder stats comparisons. I'd personally love to see the "batting score", "bowling score", and "allround score" as calculated below added to the stats pages...

So I've recently become a little obsessed with cricket stats thanks to the Short League friendlies - I wanted to replicate the Stumped league stats for those games so created my own little spreadsheet. Unfortunately this interest rapidly became an obsession, and has now spawned a suggestion for new stats in the main interface! I thought I'd open a conversation here about this approach to see what others think.

Basically I've been thinking of a way to judge allround contributions - both batting and bowling. This led me to some research and particularly the discussion here.

I really like the second approach suggested in that page, from the cricinfo stats editor S Rajesh. I reckon it's a really clever idea. And it's one I could easily do from scorecards which is a bonus! The other approaches use in-game data which would require a scraping in-game info, which is more than I (or realistically Rob) would likely be prepared to do!

All the approaches are trying to judge individual performances with the bat and ball in the broader context of limited over matches - where "average" is not sufficient to describe contributions, and "economy rate" is also misleading if you look at it in isolation. And being a scoring machine on a flat pitch may not be as impressive as scoring fewer runs on a real turner.

So the idea proposed by Rajesh is to weigh each player's individual performance relative to the average performance in the games that they have played. Here's my summary:

Batting score
A player's "batting score" is an indication of how many runs a player makes per wicket relative to others in the games they have played in (their "batting average index"), and how fast they made those runs relative to other batsmen in those games (their "batting strike rate index").

The player's "batting average index" is their overall batting average divided by the overall batting average in the games they played. So if a player has an average of 50, and in the games they played the overall average is 25, they'd have a "batting average index" of 50/25=2. That is, the player scores twice as many runs for their wicket as everyone else they played with and against, on average. An index greater than 1 means they have performed better than others, less than one means they're not as good.

Their "batting strike rate index" is the rate they scored those runs compared with the other runs scored in their games (ie their average strike rate divided by the total strike rate for runs scored in the games they played). If a player has a strike rate of 100, and the overall strike rate is 50, their "batting strike rate index" would be 100/50=2. In plain English, this means that the player scores runs 2 times faster than everyone else they play with and against, on average.

Their total batting score is:

Code: Select all

Batting score = (batting average index) x (batting strike rate index) x 100
where:
(batting average index) = (individual average runs / wicket) / (total runs / wicket for the games they played in)
(batting strike rate index) = (individual average runs / 100 balls) / (total runs / 100 balls for the games they played in)


Bowling score
Similarly we work out bowling indices based on bowling average and economy rate. In this case lower is better, so it's the inverse of the batting ones: "bowling average index" is the overall bowling average in the games the player participated in, divided by their personal bowling average. So if the bowler average 10, and the overall average in those games was 20, their bowling average index is 20/10=2. Same for "bowling economy rate index" - if they conceded 3 runs per over, and the average is 6 runs per over, their bowling economy rate index is 6/3 = 2. The scores are combined in the same way:

Code: Select all

Bowling score = (bowling average index) x (bowling economy rate index) x 100
where:
(bowling average index) = (total runs / wicket for the games they played in) / (individual average runs / wicket)
(bowling economy rate index) = (total runs / over for the games they played in) / (individual average runs / over)


All round score
I've come up with this one - just the addition of Batting score and Bowling score. There's two ways to look at this one - either the absolute best individual contributions (eg a bowler that's so much better than other performers in their games that their batting score is irrelevant), or the best "all round" contributions (eg someone who is expected to contribute with both bat and ball). For an "all rounder" contribution it kind of makes sense to limit the comparison to those who've batted and bowled a certain amount (say scored 100 runs, and bowled 20 overs). But it's also interesting to compare the individual contributions of pure batsmen and bowlers using these scores.

Thoughts?

as a postscript, I've also had some thoughts about fielding (eg using "net runs saved" and "fielding dismissals" indices something like the scores above) but the first of those is impossible to do unless you track fielding contribution in games. The commentary does show when a fielder misses an opportunity or saves runs, but it doesn't quantify these chances so it could only be done in the database itself. I couldn't be bothered scraping the commentary for it, and I doubt there's a lot of appetite to make coding changes for this sort of flavour enhancement.

bumpuss
Posts: 378
Joined: Sat Apr 23, 2016 12:30 am

Re: Statistics

Postby bumpuss » Wed Aug 14, 2019 9:56 am

I dont see the benefit.

Generally a player with a higher batting average is a better batsmen than somoene with lower average. We also have visible to us the number of matches played and not outs - so we can guage the "true" strength of the player.

The player's "batting average index" is their overall batting average divided by the overall batting average in the games they played. So if a player has an average of 50, and in the games they played the overall average is 25, they'd have a "batting average index" of 50/25=2. That is, the player scores twice as many runs for their wicket as everyone else they played with and against, on average. An index greater than 1 means they have performed better than others, less than one means they're not as good.


With above example, if the overall average is 25 then that means the rest of the players batting avergae will be lower anyway. So this player average is going to higher than the rest anyway.

Importantly this doesn't take into match situation (for e.g players batting down the order are more likely to have lower batting averages and higher SRs as they dont bat for as long and often at the end they need to hit out). I dont think this statistic shows the value of such a batsman.

The other thing is this statistics also ignores matches in which the batsmen does not take part in, why? Will this favour flat track bullies (aka Warner?) If they only play in flat pitches, the rest of the batsmen may do well or may not?

I do not think this provides any additional insight.

liquefry
Posts: 339
Joined: Wed May 18, 2016 10:10 am

Re: Statistics

Postby liquefry » Wed Aug 14, 2019 1:55 pm

Generally a player with a higher batting average is a better batsmen than somoene with lower average. We also have visible to us the number of matches played and not outs - so we can guage the "true" strength of the player.

I'm not suggesting removing the existing stats, just supplementing them. I'm not sure what you mean by the "true" strength but yes you can see raw numbers. It's not always true that a player with a high batting average is better - the example the original stats guy uses is Michael Bevan (average 54) vs Sehwag (average 35). How do you even compare? Few people would suggest Bevan was almost twice as good as Sehwag.

With above example, if the overall average is 25 then that means the rest of the players batting avergae will be lower anyway. So this player average is going to higher than the rest anyway.


Exactly the point. Remember it's only the overall average for games they were in. Imagine that guy plays on a team whose home pitch is very hard to score on, giving that low 25 average. But every other pitch in the comp is a road, so the overall average for the comp is actually 50. The first batsman would just be average compared to everyone else, but the index factors in how hard his runs were to get so you get a better idea of how good he is compared to everyone else.

Importantly this doesn't take into match situation (for e.g players batting down the order are more likely to have lower batting averages and higher SRs as they dont bat for as long and often at the end they need to hit out). I dont think this statistic shows the value of such a batsman.

This sort of situation is exactly what the batting index does show. It's average index x SR index - so a fast low average batsman might have the same score as a slower high average batsman. At the moment you can't tell the value of these batsmen at all.

The other thing is this statistics also ignores matches in which the batsmen does not take part in, why? Will this favour flat track bullies (aka Warner?) If they only play in flat pitches, the rest of the batsmen may do well or may not?

No, again this is exactly what this score corrects for. If you only ever play flat tracks, you will have a high average but so will everyone else (both your team and the opposition). So your index will be lower than it would be if you made the same average on difficult pitches.

Your criticism could be used to directly point out the shortcomings of traditional stats like average and SR that favour players who play more often on favourable pitches (batsmen on flat pitches, and bowlers on seamers). This approach effectively corrects for those problems, and combines the two factors so you have a better idea of how effective the batsman is both in scoring runs, and scoring them rapidly, in the context of the pitches and bowlers that they are playing against.


Return to “Suggestions”

Who is online

Users browsing this forum: No registered users and 1 guest