That doesn’t really sound like it understands the game... is this not training by playing against itself? How is it not learning the flawed reasoning behind this strategy?
The reasoning behind worker rushing is sound. StarCraft II is a game of hidden information and there's no time where that's more relevant than the beginning of the game.
The problem is that there's a sort of rock-paper-scissors game of (conservative)-(greedy)-(all in). That is, a player who build defenses early on will crush a worker rush but they'll be far behind a player who ignores defenses and focuses on growing their economy. The worker rush is designed to punish this greedy build order.
There's nothing that can be done, really. If you decide to send a worker to scout super early on then your economy suffers so much that you'll be behind no matter what you do.
Edit: to those of you drawing a distinction between an immediate worker rush and something like an early pool or a cannon rush, that's splitting hairs. These are all classified as all-in strategies that almost automatically lose if they encounter a conservative build order.
The rock paper scissors element is there, but not with worker rushing. Worker rushing is a huge disadvantage.
The term means you take your initial workers and charge the enemy. If you know where the enemy is you'll arrive about the time they are up one or two workers on you, then you just have to fight when you're down one or two workers and you can't back off (they'll return to mine).
A human can win like this if they are so much better at micro that they can overcome the material disadvantage (unlikely) or if their opponent panicks and does something dumb (e.g. pulling only half their workers to fight).
Because worker rushes arrive so early it doesn't matter what plan you have for the match, you'll have at least as many workers (if you are also worker rushing) or you'll have more. Maybe if you're going six pool you'll have fewer but you'll also be able to get lings and win with them.
The computer can lose to worker rushing only because it can get caught in loops of poor decisions (e.g. you lure away a couple workers and kill them).
Possible, but I highly doubt the differences are significant enough to make up the deficit. The other side has one or two more workers, which is just an insurmountable material advantage. If you back off to recharge your shields, I return to mining and get the resources for another worker or two (plus, my guys can recharge, regen, or repair too).
I'd argue that with AI level micro, the Protoss probe shield regen advantage would definitely come in play. Shields regen only when out of combat for 10 seconds and they regen at 2/second after that so it just takes 20 seconds to regen 20 shields.[1] This could definitely come into play enough, if micro'd enough, to make a difference, maybe... It'll be exciting to find out.
20 seconds is pretty long. Your opponent is up 2 workers on you, so assuming perfect micro from both sides, the defender can use one worker to continue to mine minerals and pump out more workers, while microing with the other 11 vs the attacker's 10 workers.
Workers can fade through units (unless they are sieged) when they collecting resources (minerals, vespene). So Idra commanded his drones to mine minerals and canceled the command shortly after.
This is common technique, mostly used to get scouting drone back through a walloff (own unit stands on "hold" in a small choke) or when a mineral line is under siege (e.g. Siege Tank, Liberator), the player can select all his workers to mine a single mineral patch (has to be spammed, otherwise workers redistribute) outside of the siege range until they siege is cleared up.
I think the chess equivalent would be choosing to play with only your King and your pawns under the logic that your opponent won't have practiced playing against only pawns and won't be expecting you to do it. If you're up against a novice they might panic or fail to react and you win! Otherwise, not.
Most cheese in SC2 is a micro check; you're testing to see if you can control your units better than they control theirs, even if they might have an economic advantage. A worker rush isn't nearly as stupid as what you're describing in chess -- especially if it's a ZvZ match, because in that case you can start planting spine crawlers on their creep.
The surprise for a human player won't gain the deepmind AI much advantage. A human can recognize the concept "I am being worker rushed" and immediately counter it, necessarily having more workers at their starting site than the attacking player who sent only their starting workers.
As the grandparent stated, a worker rush only works against the built in starcraft computer player, because the computer AI is too dumb to recognize "I am being worker rushed" and may not commit all of its workers to defending.
I disagree. The part that everyone seems to forget about Starcraft is the micro; humans are imperfect at controlling their units, and even relatively decent players (say Diamond-ish) will sometimes lose to something like a worker rush.
> Edit: to those of you drawing a distinction between an immediate worker rush and something like an early pool or a cannon rush, that's splitting hairs. These are all classified as all-in strategies that almost automatically lose if they encounter a conservative build order.
Things like early pools and cannon rushes are (depending on the details of the meta at the time) a rock papers scissors game. Cheese (early pool/cannon rush) beats aggressive play beats conservative play beats cheese (roughly speaking).
Worker rushes are just bad. Apart from playing against 6 pools on really old versions of SC2 on maps with ridiculously short rush distances they just lose. Even then I don't think I'd call them good.
Other commenters have covered my issue with worker rush. My larger point was that it seems like the AI has learned that this works against the standard Ai, because the standard Ai cant properly adapt to such a strat.
That might solve its task, but it’s a far cry from learning to play Starcraft well. I was surprised they seemed to have trained it against the standard Ai as that’s not useful.
>How is it not learning the flawed reasoning behind this strategy
If it's playing against the Starcraft 2 blizzard AI, it will learn a lot of "bad habits" because outright bad strategies work against the AI. Thus the worker rushing.
As a moderately skilled human player myself, I developed a foolproof cheese strategy for defeating insane AI. They go crazy if you attack their worker line and recall every attacking unit that's not currently engaged. I played terran and built a single supply depot in the middle of the map to give me enough advance warning when they were coming to attack me, and then dropped 2-3 marines behind their minerals to stall them.
If you do this long enough you can build a virtually undefeatable siege tank army that the AI would simply smash its face into, and then corral them to 2 or 3 bases. They would always run out of minerals while I expanded everywhere and slowly inched tanks forward. Doing this same strategy against humans got me to about gold tier in the early days of sc2.
Even a below average one: it's normal to belong in bronze league (~5th percentile) if you beat the AI on every level and then start playing other humans. The AI is not a serious opponent.
"Cheesing" as it's called, is classic SC2 strategy. People absolutely hate it, but it occasionally works, even though it feels cheap. It's kinda like Blitzkreig in chess...
I used to be bitter and hate it with a passion. I realized it has a very important part in the psychology of the game, especially in series. it creates very exciting moments for spectators when used well and helps keep people honest in their openings. If you as a pro never cheese, you will get taken advantage of.
Who hates it? 2017 and 2018 saw a resurgence of aggressive play that resulted in more varied games than ever. Some strategies that were previously considered all-in plays are now standard openers.
Like someone else mentioned, early game sc2 is a bit like rock, paper, scissor - scout too early and you'll be behind in economy against someone who's super greedy, but scout early against cheese and you've got a free win. You can take the "conservative" approach which is relatively safe against early cheese while not falling "too far behind" against a greedy opponent, but you'll still be behind in most of the games. If you're the better player though I'd argue it's still better to play conservative, be slightly behind, but win in the late game, while not falling victim of early cheeses.
People keep saying this but it's really not true in current starcraft. You start with 12 probes now, and terran and protoss probe scout 90% of the time at least with their first probe.
Another name for the "Scholar's mate". Quoting wikipedia [1]:
> Scholar's Mate has sometimes also been given other names in English, such as Schoolboy's Mate (which in modern English perhaps better connotes the sense of 'novice' intended by the word Scholar's) and Blitzkrieg (German for "lightning war"), meaning a quick and short engagement (Kidder 1960).
> After feeding the agent replays from real players, it started to execute standard macro-focused strategies, as well as defend against aggressive tactics such as cannon rushes.
Not a worker rush. Worker rushes only really work against lower ranked players that panic and do not know how to respond.
The reason is that while the attacking player sends over their workers to attack, the defending player is still constantly making workers. By the time that attackers get there the defending player should have more workers. He simply has to command all of them to fight the opponent's workers.
It's plausible that a self playing AI would still learn to worker rush in some race combos though, at least for awhile.
If the opponent doesn't react, you win because you kill all their workers.
If the opponent overreacts, you win because you are mining more than them (IIRC this is a problem with the insane ai).
If the opponent reacts poorly in other ways, you might be able to win by out microing them. E.g. SCVs (terran workers) consistently beat drones/probe (the other races workers) 1-1 without micro. Repairing can be an advantage. The offensive AI being better at stacking damage on a few units. Etc.
The defensive AI should learn to fix these problems and stop losing to worker rushes, but it could take awhile. In other words the defensive AI starts out as a lower ranked player.
PS. I'm not sure how serious a tournament this was, but here's an example of a real worker rush (not an early pool with workers attached): https://www.youtube.com/watch?v=_AtqgaJzJP4
Isn't the idea that because your opponent spent resources on things like their first barracks and gas collector that your workers should outnumber them for the attack?
Of course you lose all of that time it takes for your workers to drive across the map, so it's a pretty dicey strategy. A 50% win rate against the AI sounds about right.
You shouldn't cut worker production at the beginning of the game - it's critical to keep pumping them uninterruptedly. You only build when you have enough resources to ensure that.
Worker rushes do not work at pro level. Cannon rushes and proxy buildings sometimes work if caught off guard. The openings in sc2 is more akin to the mind games of poker than chess. Information is hidden so there is nothing to keep you honest.
What does "understand" mean? Do you understand the game if you know what all the units do, their hitpoints and attack values, how to build them etc.? Do you understand the game if you build a model in your head of the current state of the game and use it to try and predict what moves give you the best chance of winning? If so then it's probably fair to say that any decent AI "understands" the game on some level.
Or is "understanding" an attribute which is only afforded to humans who do these things, much like a machine can process the same inputs in the same ways as a human but will probably never be considered to have 'qualia'?
I think you bring up a good point. People imprint their own perception of things to this AI game. The bot is probably working like other deep learning machine learning things. It runs some models or whatever and tries to minimize error. Not surprising really that the computer thinks differently than humans. A worker rush sounds hilarious to us but hey it can work.
It doesn't sound hilarious to a Starcraft player. It's a very common strategy in the lower ranks that's very effective against players who panic or don't have good micromanaging skills. Everybody has been beaten by a worker rush at least once.
For a complicated game that lacks a dominant strategy like Starcraft does, the two are the same. You must be able to infer the state of the game given your limited factbase of what’s going on. You must be able to choose the strategy that is most likely to Result in victory.
Using this strategy is objectively bad in that it can be repelled by an opponent almost regardless of their initial strategy.
If deepmind’s goal is to beat the standard Ai, then sure I guess it understands that. But that’s a very different problem from understanding Starcraft, and a much less difficult one at that.
They are still learning. So at some point their AI should explore worker rushes, but also explore how to defend against them.
And once the defense gets good enough, the attack portion of the AI should probably do the worker rushes much less often.
Of course, given that the AI has different constraints in micro-management and attention, and that the game is balanced given human skills, some things that would be totally bonkers to do as a human might work for the AI.
