Median Pick/Ban Rate of All Heroes: 8.00%

Compare this to prior years:

  • 2022: 12.99% (62.3% more hero viability)
  • 2021: 13.66% (highest in recent history, 70.8% more hero viability)
  • 2020: No TI. Notably, Patch 7.27 increased bans from 6-per-team to 7-per-team, which should increase overall pick/ban rate for all heroes in later years
  • 2019: 10.36%
  • 2018: 9.33%
  • 2017: 13.61%
  • 2016: 10.38%
  • 2015: 6.13% (lol yikes)

So what does this mean? Basically there was such a strong pool of must-pick and must-ban heroes (e.g. Treant, Dazzle, Kunkka, PB, Muerta, Invoker, Bristle, CK) that the rest of the pool was disproportionately ignored. If you look at only the number of unpicked/banned heroes, it’s not so bad compared to prior years. But there are so many more heroes that only saw 1-3 games played because the same 20 heroes were soaking up the pick/bans.

NOTE: Due to event schedule changes, there were fewer games played this year (150 as compared to 231 in 2022, 194 in 2021), and that can lead to greater extremes in data. However, 2016 also had <160 games played and has dramatically higher hero diversity at 10.38% (~29% more hero diversity). Games played is not the driving factor behind TI 2023’s hero puddle.

NOTE ALSO: Adding new heroes will decrease P/B rate of all heroes across the board. But with the addition of a 7th ban for each team in 2020, the ratio of heroes to picks/bans has stayed about steady.

Source: Liquipedia post-tourney stats pages (e.g. https://liquipedia.net/dota2/The_International/2023/Statistics)

  • _NINESEVENBEnglish
    1·
    1 year ago

    Nice! Median pick/ban rate is way better for this than the usual “highest number of unique heroes” or “number of uncontested heroes”.

  • The_Wata_BoyBEnglish
    1·
    1 year ago

    My stance on this being the worst TI patch continues to grow. This has been the most unfun 2 months of Dota I’ve ever played.

  • ileamareBEnglish
    1·
    1 year ago

    It feels like looking at median pickrate directly and comparing them is a flawed approach to evaluate the diversity of a meta.

    I have a metric based on the number of the heroes present in the “core” of the hero pool of a report (aka how many heroes are close to average), basically showing how many heroes were actually played, were viable and were balanced. There are three versions of this metric, focusing on different stats (pickrate, winrate, banrate) and one is the combined version. The metric is not tied to the number of matches played, but rather is using calculations based on quartiles and median pick/ban/winrate values (I had a blog post with a more detailed explanation).

    I will use the combined and winrate versions here, but you can check all of them in my reports at any time.

    Based on that :

    • TI12: 70.5% / 92.3%
    • TI11: 64.0% / 85.5%
    • TI10: 63.8% / 78.6%
    • TI9: 63.1% / 73.5%
    • TI8: 50.8% / 56.0%
    • (not going further as it won’t really count as recent history at this point, but the data is still present)

    Based off that, it seems like TI12 was actually the most diverse and balanced in terms of meta in recent history.