What’s the problem? I’ve watched all the “evidence” videos and haven’t seen anything that I haven’t seen a thousand times on popular streams. Yes, they always “know” where the enemy is hiding. Yes, they headshot everything in sight. But so do Shroud, DrDisrispect and all the famous “pros”.
I haven’t seen any WGQun123456 level cheating, like killing someone through the wall or shooting into the blind distance and get a kill or headshotting vehicle drivers or speedhacking. All I saw is inhumanly good reflexes and intuition. But isn’t “inhumanly good reflexes and intuition” the definition of “pro play”?
This reminds me to the Geguri incident when a player was accused for cheating because she maintained perfect aim during movement like the aimbot users usually do, but she could clear her name by playing on a clean computer front of audience.
The problem is fundamentally that “git gud” is equal to “get as good as a $5 commercial cheat program”. Players celebrate “skillz”, things that can be easily emulated by some primitive program. Then they get surprised by being beaten by someone using said program. Besides cheats that shouldn’t exist at the first place (the server shouldn’t allow you run faster than possible or pass walls), there is no way to defeat cheats. Proper client-server setup prevents you from see through walls, but nothing helps against a program that simply “plays” the game “fairly”, I mean taking only input that legit players get and performing actions that legit players can do.
Sure, commercial cheats can be analyzed and scanned down in the memory. But custom made or exclusive cheats cannot.
“Git good” necessarily spawns cheating. The only way to prevent it if the game is designed to test intelligence and not hand-eye coordination. For example if in PUBG the player rotation speed would be limited to what a human can do and movement would ruin aim, just like it does in reality, the winner would be the one who plans his actions better instead of the one who rushes in and just headshots everything by spinning and jumping and aiming while doing all these. As long as players value “skillz” that can be simulated by a $5 program, there will be $5 programs.
“to test intelligence and not hand-eye coordination.” “As long as players value “skillz” that can be simulated by a $5 program, there will be $5 programs.”
To be honest, I don’t really see the difference. If ‘intelligence’ rewarding games became main stream, I don’t see a reason why people wouldn’t buy cheat programs that do the planning for them (like a chess bot or a poker bot). Just like there are farm bots to win the ‘test of time investment’, or like there are Hearthstone arena draft tools that are legal (or at least not banned/punished till now) and help you with the decision making.
So the phrase “‘Git good” necessarily spawns cheating.'” isn’t wrong, but I think every pvp activity or competition spawns cheating, followed by regulations and controls and the never-ending cycle between those two. And if it’s DDOS-ing a pro player during an online tournament, or manipulating the market in some way.
LikeLike
What would be a cost of program that is able to find the optimal move in that limited PUBG?
LikeLike
The only way to prevent it if the game is designed to test intelligence and not hand-eye coordination.
AI has already beaten the best human players in the world at Chess and Go. On a more mundane level, there are also apps that feed the correct plays to people in games like Words with Friends and such. “Testing for intelligence” is going to fall by the wayside pretty soon too. And there will be no cheat detection that can impede that.
LikeLike
@Skeddar, @Azuriel: there is a huge difference between a serious AI and a wallhack. While The chess “bots” run on supercomputer and developed over years. It’s no doubt that a decision making bot could be written, but I doubt it will be written considering the costs. Please note that every commercial bot is caught at the end, so it must be not expensive to develop. The bot cycle is “develop cheap, sell to idiots, make profit, get banned”.
LikeLike
Well, it is a sad truth that whenever there is a competition and winning will give some rewards/advantages, there will be cheating. The same happens in RL too, only in a much larger scale and often with much more damaging effects.
The most successful business tycoon in America became rich by cheating, by eliminating competition and leveraging customers by then aquired market dominance.
Rockefeller did it, Gates did it and many others… The result is everlasting damage to “free capitalist markets” that comes with less supply for a higher price…
If these alleged cheaters were to play at an e-sports event on clean systems in front of a public audience, we would soon have proof of one way or another.
As long as we cannot check for sure if cheating is involved, we always assume that a better player must be a cheater, don’t we? 🙂
LikeLike
people don’t know skill anymore and how fucking long it takes to master. It took me a whole year in quake to pretty much nail rails (huge dmg, direct shot, no delay, no physics.). and another several months of dedicated practice to nail reactionshots 360° with rail. not to mention learning meta and feeling and predicting your enemy and disrupting their play.
If you dedicate time into “getting gud” you need /record or other capturing and logs. nothing has changed since the early multiplayer FPS days. because M&S can’t possibly comprehend how to master – to their mind – insignificant motions, for them it is like watching some old china kung-fu training vid. So for them it is just logic to buy software because they don’t know how to improve and project their fail onto others.
cheat whining exists since play exists, competitive and recreational.
LikeLike
Regarding complexity of PUBG, it is actually very interesting. For the longest time, Go was considered the most complex game. However, people who developed AI for both Go and DOTA said that DOTA was actually more complex. Go figure. With that in mind, PUBG is probably even more compelx than DOTA.
I actually have a great deal of hope for humanity’s ability to develop games that computers can’t solve.
As for skillz being replicateable by a bot: don’t confuse skillz for popularity. Here is one of “best moments” lists from recent AGDQ: all require skill, but in none of them skill itself is the point. https://www.youtube.com/watch?v=6HU3H_N4C14
LikeLike
@Gevlon:
“there is a huge difference between a serious AI and a wallhack. While The chess “bots” run on supercomputer and developed over years.”
The bots running on supercomputers are made to defeat the world’s best players. Any semi serious comercial chess AI (eg.: most modern chess games on hardest difficulty) can easily defeat most chess players on a simple PC. For a gaming bot you don’t have to defeat the best ones, it more than enough if you defeat the average one. Which is not a high bar.
LikeLike
“the server shouldn’t allow you run faster than possible or pass walls.. ”
It would be easy to implement but it’s expensive in cpu power to run the checks for all players. You skip it cause the damage ain’t worth the cost to afford those checks. All about the money.
“but I doubt it will be written considering the costs. Please note that every commercial bot is caught at the end, so it must be not expensive to develop. The bot cycle is “develop cheap, sell to idiots, make profit, get banned”
Only the idiots get caught. Wotlk-Mop botting got in a group of ~20 people only one warning. Did leveling, fishing, bg botting.
First glider then honorbuddy. Devs catching everyone is just a lie that keeps the gullible masses from cheating. The ban cycle is also just so that the professional idiots buy back in and don’t give up. If you banned the idiots weekly they would stop. Ban them every quartal and you get additional income. Lastly wow arena got botted to 2.2k. It’s not impossible to write some decision-making at all. You just build it up from a basic bot over the years that provides the initial income.
tldr: The only real protection a game can have is being so insignificant nobody can earn enough from the small playerbase.
LikeLike
@Maxim, cathfaern: this is because chess and go has few possible steps while PUBG has infinite amount of possible steps (you can go any direction). Sure, most of the steps are irrelevant or stupid (for example to step left in the middle of nowhere), but you must code for it. Also, you must code against getting stuck or friendly fire and other content-related crap. For chess you must code for a few dozen possible steps. Sure, if the program wants to plan the game 15 steps ahead, that’s lot of CPU, but not programmer time. For reference: it’s easier to write a program that can find you a million prime numbers (probably less than 50 lines), than a program that “looks” at a photo and tells if there is a dog on it or not.
@Anon: if they don’t even bother to do it on players who got reported, that’s not excusable. You don’t have to introduce corrupted devs for me who don’t ban cheaters.
LikeLike
Go has enough “steps” to be practically infinite. The latest Go AI doesn’t work by predicting billions of moves, it can evaluate the “value” of the board with high precision. Nobody really knows how it works, humans can’t follow the “logic” behind the neural networks.
There is a bot playing LoL in 1v1 duels and it’s unbeatable. https://www.youtube.com/watch?v=XbDmxEOj9OY
Humans aren’t all that smart, we are only very flexible. We know how to do very man different things badly while a computer excels at only one thing.
The latest Google Go Ai taught itself. It didn’t receive any a priori strategies from the coders, which means it is probably just as good at learning any other game (with most games being significantly simpler than Go).
LikeLike
@stawek
The only real way to understand a neural network at this point is to look at what situations it trained itself with to what results and identify the pivotal ones, which became the foundation of the network’s emergent learning structure and significantly altered its emergent behaviour patterns. The problem is that with AI’s teaching themselves automatically, you need to look through millions if not billions of such situations that is hard for a non-assisted human. The good news is that this is basically an image recognition problem, we only need to figure out how to feed it behaviour patterns instead of pixels. Then we’ll get other AIs helping us understand neural networks 🙂
Here is a video of the unbeatable bot being beaten twice in a row by a human. It doesn’t look easy, but it does look stable. https://www.youtube.com/watch?v=YYtajHJoFso
Google Go AI didn’t come up with its own methotodology of describing in-game states. That was put in by human programmers. The ability to generate such methodology for any domain field remains (and will remain in forseeable future) the purview of the humans.
@Gevlon you seem to be saying that this complexity is useless because it contains a lot of reduntant moves. As it turns out, all stable biological systems feature a lot of apparently useless redundancy that you don’t really use in normal situations but that become key for survival in the occasional crisis. This is the big difference from current computer systems that function much better than humans without the redundancies, until you get a serious enough unplanned-for event and your device becomes little more than a brick.
LikeLike