Hacker News Reader

Comment by haolez | Hacker News Reader

I think that an underestimated benefit of game streaming is anti-cheat. It would benefit a lot games like Fortnite and Pubg.

View on HN

Competitive games, at least competitive shooter games, are really the worst application for this though because they suffer the most from input-to-display latency. Games like League of legends might be fine though.

But assuming that the latency gets reduced to acceptable levels for competitive games, I see this distribution model as the only one to effectively fight cheaters. I would love to hear alternatives, though.

That's a really big assumption. For a lot of genres the latency is just not going to be low enough.

Based on some rough, early estimates the input lag with Stadia is incredibly impressive. I originally thought it would never work for many types of games like you say but now I'm reserving judgement. It's not impossible at least.

They suffer from uneven or inconsistent input-to-display latency. If Stadia can give everyone the same, (hopefully) modest amount of latency, the games remain as fair as they ever were

It might be fair but that doesn't mean it will be enjoyable...

I'm not entirely sure, but you may have misunderstood my point. It's not about fairness, it's about whether it is physically possible to play the game on the same level of skill.

Talking about shooters, in the traditional multiplayer setup input-to-display latency isn't actually a concern, because input-to-display doesn't go over the network at all. When you move your mouse, the display update loop doesn't contain a network roundtrip. There is only the concern of latency to the server for resolving which hits actually connect, and I agree with you that having this latency be roughly the same for everybody is probably quite a bit more important than how low it is.

With game streaming, the input-to-display loop does include a network roundtrip, which makes it harder to aim.

If you've ever tried to play a shooter with <15 fps, or suffered from the mouse input lag that some games used to have on Linux in the past, you should know what I mean: aiming with the mouse crucially relies on a feedback loop where you correct your hand movement based on what you see on the screen. The longer that feedback loop takes, the harder it is to aim.

That's what the concern is with competitive shooter play on game streaming services. (And it's why I wrote that other types of competitive games may not be affected as much, but it's been a long time since I've played a lot of games, so what do I know...)

If you have <20ms latency (possibly a bit higher), I don't think it will inhibit the skill cap for most shooters.

If it creeps up into the 40's and 50's though, it will for sure.

I don't think so - the reason I don't use VSync for first person shooters is because it introduces 1 frame (8ms ~ 16ms depending on your frame rate) of input lag. It really feels laggy when I try to move my mouse and aim.

And mechanical keyboards make you type faster and gamer glasses make you aim better and a tincture of molybdenum in your star-water will cure your cold

I'm not sure you understand fairness. If the game is harder for everyone (in whatever metric suits you to contrive), it's just..harder

How so? There are very common aimbots based on computer vision that will work just fine on a PC with Chrome running Stadia games.

Sample naive algo:

-Get a color range (RBG) of the enemy model's head.

-Take screenshot of screen and scan for pixel in that range.

-If match, get the x,y pos of that pixel.

-Set mouse(or controller) pos to that x,y pos.

-Click

E.g https://github.com/Dewep/POC-AimBot-Overwatch

The sample code worked well enough for Blizzard to forcibly take it off the internet.

Not every game outlines the enemies in red like Overwatch. Many times, the player model heads don't have their own exclusive pallette.

That's for aimbots. Streaming would remove a ton of other exploits, like wall hacks.

However, for the aimbots, since you are sending your input to Google's servers, they could detect the input patterns of cheaters.

> However, for the aimbots, since you are sending your input to Google's servers, they could detect the input patterns of cheaters.

Isn't that exactly what happens today on a console or a PC?

On individual boxes, with games having their own versions of anticheat detection. Not at Google scale looking for patterns across all games.

Ah, I see what you mean. That's interesting. I feel for people who'll get banned because they are too good given Google supports reputation :)

There are undoubtedly major benefits in the war against cheaters by completely removing user access to the underlying system. However I think there are two fundamental points to keep in mind on why it's not going to be an ultimate solution.

1) Unless cross-platform play is forbidden or the game is exclusively streamed-only, the cheaters will just use a different platform and will still ruin your game the same old way. It's the cheater's platform of choice that matters.

2) Games are some of the least security conscious pieces of software out there [1], primarily written in C++. There are bugs, lots of bugs. Process-takeover enabling bugs. I'm sure Google sandboxes the game to protect their systems, however cheaters only need access to the game process to enable most of their desires. Yes this would raise the bar in how easy it would be to cheat. Average Joe Cheat Engine users would be gone, but more skilled cheat makers will continue business as usual and their released cheats will do the exploits hidden from the actual people doing the cheating.

Bonus: See this cute hack that injects flappy bird into Super Mario World. https://www.youtube.com/watch?v=hB6eY73sLV0

[1] Even AAA developers are clueless about threat models. Games like Tom Clancy's The Division [2] and Fallout 76 [3] are multiplayer games that put extreme trust into the client. Trust that nobody would modify their script files, trust that the client is always telling the honest truth.

[2] http://web.archive.org/web/20170611084112/http://gafferongam...

[3] https://www.reddit.com/r/fo76/comments/9u71m1/get_ready_for_...

Except for the part where it appears you'll be able to play from a computer just using Chrome. This allows for all sorts of botting (artificialaiming.net for example has been making PC aim bots for at least a decade as I tried it out on Battlefield 2142 and it was eerie at what all they could do including the settings to make it even believable so you were unlikely to get reported. Sure that was using a lot of client-side data but that was a long time ago and now you can probably do sprite-recognition type stuff effortlessly on the user side of a streamed game.

Similarly fishing and mining bots and the like for games like WoW would translate to repetitive farming type activity.

Arguably I think being in a browser window would make it easier to do.

Client side input could always be messed with. I think what the anti-cheat feature refers to more so is the direct communication with the server part, where cheating programs can e.g. modify UDP packets in transit to teleport your characters to new locations or give themselves inventory items. If all you're feeding into the system is controller input, there's no way for you to do any of that. Conversely, backend developers for exclusive Stadia games can iterate more quickly knowing that their client input is completely trustworthy, which is an INSANE win.

Imagine building a REST API and not having to worry about request validation, sanitation, authorization, data scoping, etc. because you can 100% trust that the input you're processing is from your own client. Wouldn't that be perfect?

Reply on news.ycombinator.com