EPT
OpenAI's Dota 2 bots vs. 5 top professionals in TI - Page 8
| Forum Index > Dota 2 General |
|
Laserist
Turkey4269 Posts
| ||
|
WolfintheSheep
Canada14127 Posts
Creep pulls should screw them over heavily, they didn't seem to react at all to creeps being anywhere except lane. Observer wards in lane instead of jungles should spot the early rotations, then the players should just hang back. | ||
|
nRoot
Germany928 Posts
On August 06 2018 06:18 cecek wrote: Very interesting to see how many consumables the bots use. But it's kind of lame that they mostly win just by perfectly executing teamfights and having superior reaction time. 5 couriers makes it very hard to play a "normal" dota game vs OpenAI, I would love to see that restriction removed | ||
|
polgas
Canada1770 Posts
![[image loading]](https://i.imgur.com/r75tuRy.png) Disrespects SF. | ||
|
spudde123
4814 Posts
| ||
|
WolfintheSheep
Canada14127 Posts
| ||
|
spudde123
4814 Posts
| ||
Firebolt145
Lalalaland34503 Posts
What the fuck | ||
|
spudde123
4814 Posts
On August 06 2018 06:56 Firebolt145 wrote: I think Lion just dusted because his movement was obstructed by the invis riki. What the fuck I think they just saw Fogged mid with their sentry and they saw him run towards top rune | ||
|
FreakyDroid
Macedonia2616 Posts
| ||
|
WolfintheSheep
Canada14127 Posts
| ||
|
Thermia
United States866 Posts
| ||
|
polgas
Canada1770 Posts
![[image loading]](https://i.imgur.com/PGDFCFc.jpg) Much better for humans but still <30 min game win by OpenAI. | ||
|
Deleted User 173346
16169 Posts
| ||
|
Aerisky
United States12129 Posts
| ||
|
rabidch
United States20289 Posts
On August 06 2018 06:39 WolfintheSheep wrote: Yeah, pretty clear the bots don't understand lane control or movement blocking yet. Which is weird considering they learned to creep block at the start. they learned to creep block given an incentive to do so. openai team probably decided there wasnt a strong reason to keep it around. | ||
|
polgas
Canada1770 Posts
| ||
|
ASoo
2865 Posts
On August 06 2018 06:36 spudde123 wrote: I'm not really following how they can pick Lich in that case. They think Lich gives them 43.8% chance to win, surely there is a hero that they think gives them more than 50% chance to win as 1st pick? Maybe the bot just thinks that picking second is inherently better? On August 06 2018 07:03 Aerisky wrote: This is best of 5, right? Damn though this is pretty crazy, even if it's kind of cheesy with the couriers and stuff. Bo3, but they're playing out the third game. | ||
|
polgas
Canada1770 Posts
Win probabilities after draft. ![[image loading]](https://i.imgur.com/f23QMEO.png) | ||
|
spudde123
4814 Posts
On August 06 2018 07:23 ASoo wrote: Maybe the bot just thinks that picking second is inherently better? Ah right, maybe that's the case | ||
| ||
