The games are getting better and better as the tournament goes on. Today’s OG vs. LGD game 1 is an absolute must-watch.
I agree, I just watched that one. What a match!
I thought this short movie was also really well done:
All of Liquid vs VP is must watch dota.
Yes great series, especially the first match!
Newbee vs. LFY today is good too.
Lower bracket finals were just amazing. Really hyped up for the grand finals.
Yeah, the third game of the LB finals had me on the edge of my seat. Hopefully we get 4 or 5 good games out of the GF.
Yeah the third game especially was so good.
Man, Liquid up 2-0.
Those three games were simply unbelievable. Liquid just kept pushing, ganking, split pushing, and then the awesome team fights. Those team fights were usually ambushes and some poor opponent was dead in the first second. Lots of fun to watch.
Liquid was just playing on a whole other level in the finals. Crazy combos. That rosh pit in game 3 where jugg got the Omni off in the last second was legendary.
The finals weren’t as fun as the games with LFY, but amazing play from Liquid nonetheless.
I personally think this is pretty silly hype, promotion for DOTA and Musk. I see it akin to “man has just created a machine that can hit a baseball better than the best human”.
If it’s true that the bot learned to play DOTA2 by playing games against itself (as opposed to being scripted to play that particular game), and it got good enough to beat a skilled human player, I think that’s pretty cool.
It’s pretty cool, but at the same time a solo mid matchup is fairly amenable to AI and nothing like a full 5-on-5 matchup. I’ll be interested to see how this AI develops in that direction, assuming they bother to do so. If only so I can have better bots to play against.
Yes, I think they may be a little coy with the term learned. That could merely be a cover for calculating efficient pathing. Also, the AI would never have to “rest”, even for a millisecond, and I believe he does not have to use a mouse which gives him a distinct advantage physically.
I am sure it is possible to beat it. But it has no room for even slight mistakes
Makes me think the human realized that rest and efficiency were deciding factors. That’s where I draw the batting analogy - it is really nothing new and very plausible.
EDIT: and wanted to mention one of the reasons Deep Blue beat Kasparov was because in a game the day before, everyone knew it would be a draw, but the bot insisted on playing it out, and that pissed off Kasparov and mentally drained him for the next round.
It wasn’t just about things like pathing, the bot learned things like creep blocking on its own. That’s a pretty subtle interaction: you try to prevent your own creeps from moving forward by body-blocking them with your character, just to move the point where the battle happens closer to your own tower.
But when the public was allowed to play, a lot of them beat the bot by doing really silly strategies that it wasn’t at all prepared for. Like skipping behind the opponent’s tower, luring away the creep wave, running around the map with the creeps following you for just the right amount of time until you could intercept the next wave of creeps and lure them away too. And have your own creeps just slowly chip away the tower.
Arthur Gies finally reviews Dota 2.
On the flip side, when Dota 2 is bad, it can be soul crushing. I’ve never played a game so capable of making me feel like shit, like so much of a failure, or that every decision I made was, ultimately, futile. I’ve made dozens of friendships through Dota 2 — I also wonder how many it’s ended. I’ve been stuck in matches I knew I couldn’t win for more than an hour. My longest match was two hours and six minutes (at least I won that one, and in an exhilarating and improbable comeback).
But this is part of Dota 2’s biggest and most consistent problem, a problem that’s arguably only gotten worse over time. Dota 2 is an intimidating, complicated game to new players and even existing players, and the manner in which Valve onboards new players is, frankly, inadequate (though Valve has acknowledged this before, and fairly recently even). Dota 2’s biggest obstacle is the at times overwhelming amount of bullshit and opacity that stands between less experienced players and understanding the game.
More info on the 1v1 bot. Interesting notes was that creep blocking was taught separately (makes sense, as it learning that its an advantage naturally would probably require multiple orders of magnitude more time/games) and item builds were pre-populated but allowed to be changed.
Also interesting is the narrow line of winning/losing - 1 day of training (Aug 9 to Aug 10) was the difference between losing to Sumail to winning.
I wish they discussed their training infrastructure - just how many amazon instances where they running?