Why didn't @OpenAI use fair drafts against @OGesports like they did in the TI 2018 games? Is it fair to expect humans to know the 18-hero draft meta without having any practice? Draft does seem to matter, since humans were able to win with fair drafts.
1
9
Replying to @sherjilozair @OpenAI @OGesports
.@Blitz_DotA and others had figured out the 18 hero meta as of Benchmark. It's certainly different from the full meta, but not really hard for pros to figure out. It's actually much less fair to force the *bots* to play according to the human meta!

Apr 15, 2019 · 4:20 PM UTC

3
6
Replying to @gdb @OpenAI @OGesports @Blitz_DotA
They can figure it out if they have any incentive to figure it out. They are Dota 2 players, not 18-hero Dota 2 players. Was there a prize for winning OpenAI Finals? If not, then I don't see why they would spend any time analyzing 18-hero.
1
It's pretty clear that the bots won due to superior drafting than gameplay (they made huge mistakes throughout the game). Since drafting is the main difference between 18-hero and 117-hero, I expected it to be removed like in TI8 so that we get closer to the game the pros play.
1
1
more replies
Replying to @gdb @sherjilozair @OpenAI @OGesports @Blitz_DotA
Will you be hoping to include all 117 at some point or do you think it's too difficult?
Replying to @gdb @sherjilozair @OpenAI @OGesports @Blitz_DotA
Mhhh why would it be more difficult? Do you expect the bots to totally break the meta in terms of picks and thus for pro to struggle more to adapt?