Difference between revisions of "BPA"
Kevin Wang (talk | contribs) |
Kevin Wang (talk | contribs) m |
||
| Line 2: | Line 2: | ||
A player who converts zero questions over a [[20/20 format|standard format]] tournament with 10 rounds will have a BPA of 0, while a player who converts 100 questions in the same scenario will have a BPA between 0.5 (if all questions were converted at the end) and 50 (if all of them were gotten in zero seconds). | A player who converts zero questions over a [[20/20 format|standard format]] tournament with 10 rounds will have a BPA of 0, while a player who converts 100 questions in the same scenario will have a BPA between 0.5 (if all questions were converted at the end) and 50 (if all of them were gotten in zero seconds). | ||
| + | |||
| + | A typical BPA for a lead scorer is 5-10; within a specific category [[specialist]]s will typically be 10-20. | ||
==Calculation== | ==Calculation== | ||
Latest revision as of 13:18, 17 January 2025
BPA (buzz point area-under-the-curve) is an advanced stat devised by Ryan Rosenberg[1]. It uses buzzpoint information to weight a player's performance based on how early they convert questions, allowing players with similar PPG or power counts to be distinguished. A player's BPA represents how many total words were not read due to the speed of their buzzes.
A player who converts zero questions over a standard format tournament with 10 rounds will have a BPA of 0, while a player who converts 100 questions in the same scenario will have a BPA between 0.5 (if all questions were converted at the end) and 50 (if all of them were gotten in zero seconds).
A typical BPA for a lead scorer is 5-10; within a specific category specialists will typically be 10-20.
Calculation
As stated in the original post, BPA is calculated as "total area under the curve of [% of tossups gotten successfully] against [% of question elapsed]". In practice this involves calculating cumulative conversion percent at each fractional buzz point of the question being read and then summing all hundred values between 0% and 100%.
Though originally meant to describe individual players across an entire tournament, it can straightforwardly extended to teams (total BPA) over shorter scales like a single game. There are publicly available versions of the code available in R[2], Power Query for Excel,[3] and as a Jupyter notebook.[4].
References
- ↑ Introducing BPA, a new evaluation metric using detailed stats by ryanrosenberg » Wed Oct 17, 2018 12:52 pm
- ↑ Re: Introducing BPA, a new evaluation metric using detailed stats by Jasconius » Sun Apr 12, 2020 6:12 pm
- ↑ Re: Introducing BPA, a new evaluation metric using detailed stats by Alejandro » Tue Apr 14, 2020 11:15 pm
- ↑ Re: Introducing BPA, a new evaluation metric using detailed stats by Smuttynose Island » Tue Apr 14, 2020 11:39 pm