If you came here to know about me, the person who built this website,
please go to the Contact page instead.
About this website
As I collected feedback on my work, connected with people involved in Cricket,
and with astute observers of the game, I looked for common themes that I could
help address. I was also researching
quantitative work that had been done in other sports. I realized that unbiased
evaluation of player performance (not to be confused with skill)
was a foundational area that addresses several critical questions and opens up paths to more
nuanced ones. At the most basic level, assessing how a player performed in a given appearance
with a consideration of the game situation that they faced helps evaluate the performance fairly
and reasonably. Aggregating player performances and comparing players based on the the strength
and volume of the available evidence allows for judicious player evaluation that can then be
plugged into roster building, risk assessment and other strategic decision making scenarios.
In mid-2020, I
developed a method to use ball by ball event data to produce a prediction of
outcomes for an average player in a given T20 game context and evaluated evidence weighted
contributions for batters and bowlers relative to the expected performance of an average player
playing in their place. I also
developed a win probability model to evaluate player impact on game outcomes.
This website exposes the results of these models in the form of aggregated player level
advanced stats for different splits that should provide a well rounded evaluation of player
performance weighted by evidence and relative to the average player. For more details on
making sense of the metrics, please refer to the Glossary.
I obviously have put a lot more work in the automation and the backend number crunching than I have in the aesthetics of
the site. This is mostly $ & time relateed prioritization that I had to do.
To use the website, go to the home page and navigate to the player page using the player's last name.
Once you are on the player page, scroll vertically to move through the different splits and
horizontally if needed to view different stats for a given split.
You can jump across splits using the index on the right of the player page.
How is this data produced?
Since late 2020, I have automated the ETL, model building and static website generation
process to run at a regular cadence on the AWS cloud. While this makes it a hands-off process
that can scale securely as needed, it does use considerable storage and compute resources
and racks up a monthly bill that I personally pay for.
If you use this website and the data that it exposes regularly, consider buying me a coffee or two
to help pay for it.
Disclaimers
-
The methodology development as well as automation work that
produces this website was done before my time at Zelus Analytics
and my work and association with Rajasthan Royals and Royals teams in other leagues.
Conversely, my work at Zelus Analytics has not influenced this website.
-
I will not be updating any of the underlying model architecture or methodoloigy,
nor adding features as long as I am employed at Zelus Analytics due to conflict
of interest considerations. The automated model training and data updates will continue
without additional development work.
-
Data on this website may become stale for brief periods and may not match official records
or records on other sites due to a variety of reasons. I will fix issues but
I don't claim to have any SLAs/SLOs as far as new data pulls and updated
numbers are concerned.
Acknowledgements
-
I had several discussions on the concept,
the audience & my approach with Jarrod Kimber. This website wouldn't be here without Jarrod's
insightful feedback.
You can follow Jarrod at @ajarrodkimber
-
Jarrod & I leveraged data curated by Andrew Samson for this effort. Andrew makes this data available for a fee.
You can get in touch with Andrew at @AWSStats