Conversion stats

From QBWiki
Revision as of 01:29, 25 April 2021 by Ophir Lifshitz (talk | contribs) (wahever)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
Not to be confused with Conversion metrics.

Conversion stats (or conversion statistics or conversion data) measure the conversion of individual questions or categories (sometimes called category stats or cat stats). Conversion stats require tabulating entire scoresheets, which involves more fine-grained data than conventional stats (which only records summary data for each game – player statlines and final team scores), but less than detailed stats (which records much more data than is possible on a conventional paper scoresheet).

Until 2016, only a small number of tournaments had created conversion stats, usually via the tedious task of transcribing scoresheets row-by-row. Purpose-built electronic scoresheets can allow creating conversion stats automatically.

Assessment

Conversion stats can offer decent insight into how questions and teams play. For example, conversion stats can show which questions had a higher rate of powers or negs, or how much consistency or variance in difficulty exists across bonuses. It can also show which categories a team is strong or weak at.

However, conversion stats frequently suffer from low sample sizes, which contributes to misleading or misuse of stats. Conversion stats also don't tell you anything about how interested or frustrated players felt about a question.

Publishing conversion stats for unclear sets is a major responsibility and question security issue. It gives an unfair advantage to teams who play in later mirrors, it rewards many undesirable ways of metagaming, and can cause unintended psychological consequences.

History

Not much information is known about the first instances of conversion stats. They were likely to have been novelties created by one person dedicated to manually entering around 10,000 numbers into Excel (which involves on the order of 100 paper scoresheets, each containing 50–100 pieces of data).

Because large national tournaments use several thousand paper scoresheets, painstakingly creating conversion stats for them is beyond the practical scope of a single human's patience.

List of known tournaments before 2017 with conversion stats (not automated)

This table includes mainstream tournaments with published conversion stats only; side events and others are mentioned below the table.

Date Tournament Tournament type Created by Tabulated using Types of data Scope of data Status
Oct 26, 2007 EFT 2 College, housewrite Dennis Jang Excel, unknown Unknown, bonus parts 90 scoresheets, 1 site
(10 rooms × 9 rounds)
Published in thread, mentioned[1][2][3], dead since 2010
Jul 21, 2014 2014 PACE NSC High school, championship Fred Morlan Excel, manual PACE format (steals conflated), no bonus parts 96 scoresheets, 1 site
(48 rooms × 2 rounds)
Cited in thread, unfinished
Jul 28, 2014 2014 Chicago Open Open, packet submission Andrew Hart and Kay Li? Google Sheets, manual Conversion outcome only, no bonus parts; bonus categories 105 scoresheets, 1 site
(7 rooms × 15 rounds)
Published in thread, live
Jul 28, 2015 2015 Chicago Open Open, packet submission Jerry Vinokurov Google Sheets, manual Conversion outcome only, no bonus parts 136 scoresheets, 1 site
(8 rooms × 17 rounds)
Published in thread, live
Jul 30, 2016 2016 Chicago Open Open, packet submission Tejas Raje Google Sheets, manual No bonus parts; (sub)categories, teams, players 140 scoresheets, 1 site
(9 rooms × 10 rounds + 5 × 2 × 2 + 10 × 3)
Published in thread, live

In May 2010, Charlie Dees cites several dozen answerlines with low conversion from an HSAPQ tournament.[4] Since 2012, announcements for SCOP Novice and MS offer a small discount per attending team to tournament directors who send the raw scoresheets to the editors.[5] In March 2014, Marianna Zhang mentions the existence of conversion stats for the high school housewrite Prison Bowl VII.[6][7] In March 2015, Ben Smith posted a comparison of conversion stats for the extra Canadian questions at the ONQBA Provincial Championship.[8] Further information about all of the above is unknown.

NAQT collects proprietary conversion stats from some of its tournaments, including national championships like HSNCT and ICT, and some mirrors of SCT and IS sets. Perhaps an estimated 100,000 scoresheets have cumulatively been transcribed since around 2005 by people whom NAQT has hired to do such manual data entry during the summer. Representatives of NAQT may post selected conversion information upon request,[9] or at their discretion,[10] but NAQT policy is to not publish its conversion stats in full.[11]

In May 2016, Eddie Kim tabulated conversion stats for the audio-tossup-only music side event Imaginary Landscape No. 3, which was mirrored at HSNCT (up to 4 rooms × 8 rounds) and NSC (up to 3 rooms × 8 rounds). [12] Buzz times, measured in elapsed seconds, were also collected (in a form of detailed stats).

Andrew Hart published conversion stats for 2016 Chicago Open Trash (10 rooms × 12 rounds)[13] and 2018 CO Trash (12 rooms × 11 rounds).[14] As of 2021, the discussion threads for these stats have still generated more engagement than most other conversion stats or detailed stats, combined.

Beginning in 2017, Jacob O'Rourke published limited conversion stats for some sites of WHAQ, CAST, LOGIC, CALISTO, and MWT.[15]

Automation

Since November 2008, the tossup-only high school singles tournament Scobol Solo has tracked some conversion stats via software created by Jonah Greenthal.[16] It has also tracked limited detailed stats since November 2016.[17]

Even though electronic scoresheets have existed for many years, most of them did nothing with the low-hanging fruit. More recently, some electronic scoresheets may automatically tabulate conversion stats.

The April 2016 tossup-only science side event AVOGADROS NUMBER used Google Sheets to collect conversion stats. Raw normalized data, standings, and aggregate stats by team, player, category, and author were published (4 rooms × 5 rounds + 5 rooms × 3 rounds).[18]

References

  1. https://hsquizbowl.org/forums/viewtopic.php?p=60487#p60487
  2. https://hsquizbowl.org/forums/viewtopic.php?p=95022#p95022
  3. https://hsquizbowl.org/forums/viewtopic.php?p=104193#p104193
  4. https://hsquizbowl.org/forums/viewtopic.php?p=186099#p186099
  5. https://hsquizbowl.org/forums/viewtopic.php?f=20&t=13243&hilit=conversion+statistics
  6. https://hsquizbowl.org/forums/viewtopic.php?p=278529#p278529
  7. https://hsquizbowl.org/forums/viewtopic.php?p=265568#p265568
  8. https://hsquizbowl.org/forums/viewtopic.php?p=299309#p299309
  9. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
  10. [11] [12] [13] [14] [15]
  11. https://hsquizbowl.org/forums/viewtopic.php?p=306008#p306008 etc.
  12. https://hsquizbowl.org/forums/viewtopic.php?p=318027#p318027
  13. https://hsquizbowl.org/forums/viewtopic.php?f=21&t=18849
  14. https://hsquizbowl.org/forums/viewtopic.php?f=286&t=21530
  15. [16] [17] [18] [19] [20]
  16. https://www.scobolsolo.com/stats/conversion-by-question.jsp?object=2008
  17. https://hsquizbowl.org/forums/viewtopic.php?p=322675#p322675
  18. https://www.hsquizbowl.org/forums/viewtopic.php?f=9&t=18515&p=326194#p314805