Movement Prediction

[SuperPsiPower]

I believe the next thing we should work on, is the movement prediction for nonhosts. I feel that this is more or as important as RoG because prediction movement is the main core of pvp and the actual game. I have asked most of the community how they felt about it.

80% Said It needs to be fixed. I strongly agree as well, because of the way it affected the regular game (Dont Starve Singleplayer). The majority of people I asked said that the reason that they think that movement prediction should be improved because of fighting mobs in DST, Moving around without the rubberbanding, and playing pvp without the problem of having issues with moving and attacking the opponent.

[rezecib]

I'm not sure if I understand the differences between the systems correctly, but my impression is that in Dota 2, for example, it uses movement prediction constrained by prediction a certain amount ahead of what the server's information provides-- I think by default, 133 ms. This makes it feel like something a little between a system without prediction and one with-- if there's a spike, it feels like there's no prediction, because it'll only predict 133ms, then stop. If the lag is low, it smooths things nicely so it feels responsive. Perhaps something similar could be done in Don't Starve Together?

 

I think that many of the problems with it could be alleviated by better exposing the option to have it on or off. Personally I would rather play without it entirely, although maybe a constrained prediction like Dota's would convert me.

[DarkXero]

But really, even in Dota and LoL, if you have 200ms ping, you WILL notice it.

Try playing on a server with 50ms, then go play in one with 200ms.

 

With 200ms, it's ok to play characters that don't spam abilities. But as soon as you try to play Lee Sin on LoL or Meepo in Dota, you will see that whenever you try to make very quick follow-ups or very intensive micro, the game will feel delayed.

 

Given how close the players need to be in Don't Starve to actually hit each other, I feel that no matter the prediction, the game will feel weird with 150+ ping.

[rezecib]

@DarkXero, Yes, you notice it in Dota with 200ms because it only predicts as far as 133ms. There's no way to conquer lag, but the thing is that action prediction and trust-the-server both have desirable attributes, and that using action prediction for the short term gets most of the benefit of action prediction (feeling responsive and smooth in low-lag conditions), while giving most of the benefit of trust-the-server (having a very accurate idea of what's actually happening) as well.

[DarkXero]

If I get it correctly, right now, a client can only do on his screen what the server lets him do. When he runs, chops and attacks, he sends the action to the server, the server performs it if correct, then the client receives the info that he did it then sincronizes. This is why there is delay for movement prediction, and why when it is enabled, with extreme lag there is rubberbanding: the client walks, the server cant predict or misses an action, the server doesn't do it, then it sincronizes and displays to the client what's actually happening.

 

My question is, couldn't the client have a separate simulation, where he does things as if he is a host, and he sends a stream of data about what he is doing (example, move right 1 (oldposition, newposition), move right 1, move up 1, chop 1 (positionofchopper, targetposition)). This data is received by the server and the server evaluates it and performs it if it makes sense (example, the distance between oldposition and newposition cant exceed 2 given the current locomotor speed, else there was a cheat), if it doesn't, then it forces a sincronization.

 

For fighting monsters, the client would communicate everything: when he enters a monster range, when the monster starts the attack, when the monster hits, when the client hits, when he runs away of the range. Basically, if the server receives a valid engagement and fight simulation, then it allows it.

[rezecib]

@DarkXero, You're describing a trust-the-client architecture, which would be completely different from the way the game's implemented. The main argument against trust-the-client is that it allows cheating. But really it's pretty moot, as switching to trust-the-client would be basically just restarting DST development.

[DarkXero]

Well, it's not exactly a trust architecture.

 

If somebody wants to cheat, then when the server gets the "I moved right (0, 999)", it will say it's impossible, invalidate the other data received and sincronize.

[averagewalrus]

I believe the next thing we should work on, is the movement prediction for nonhosts. I feel that this is more or as important as RoG because prediction movement is the main core of pvp and the actual game. I have asked most of the community how they felt about it.

80% Said It needs to be fixed. I strongly agree as well, because of the way it affected the regular game (Dont Starve Singleplayer). The majority of people I asked said that the reason that they think that movement prediction should be improved because of fighting mobs in DST, Moving around without the rubberbanding, and playing pvp without the problem of having issues with moving and attacking the opponent.

 There is a mod that I installed removing movement prediction (Fixing rubberbanding, some lag, ect)

the downside to this mod is that using the mouse creates lag (Try killing monsters with out your mouse),

I do agree that this should be "fixed", or an option removing movent prediction because I'm tired of hardly using my mouse.

[rezecib]

@DarkXero, If it's not trust the client, then you're describing what the game does already.

 

@averagewalrus, That mod disables movement prediction, so it's not really relevant to this thread.

[averagewalrus]

@DarkXero, If it's not trust the client, then you're describing what the game does already.

 

@averagewalrus, That mod disables movement prediction, so it's not really relevant to this thread.

 I was stating that the mod disables the movement prediction, and I was stating that I think that there should be a function in game that allows you to "turn off" movement prediction or that it should indeed be fixed.

Sorry if what I said was not relevant.

[ryouryou]

Telling us that the server can validate what the client is saying based on other things the client has said is underestimating the capabilities and dedication of modders.

Oh, so, the server requires the client to tell it your travel speed so it can decide if your movement makes sense? Well, okay. Now this mod's teleport function spends a second feeding the server data that says I'm going really fast before warping me. Maybe I just tell the server that I am traveling to a wormhole at the target location. Does the server then have to scan the entire map for the locations of wormholes and make sure that that is actually one of them? People will find a loophole. Closing up one of them doesn't close them all, and might even open new ones.

The only way for the server to truly validate what the client is telling it would be for the server to perform 100% of the same simulation, which would make the whole thing pointless. Just because it is technically possible to do what the client is saying it does, doesn't mean that that's actually what happened.

For example, it would be relatively easy to make a modded client that ALWAYS tells the server you were not in range of an attack, because the distances involved would be small. How does the server know you didn't move an inch to the left if it isn't performing the simulation? It would also be easy for the client to claim that all close attacks always hit. You barely missed the butterfly? Nah, tell it you hit. It was close enough, it'll believe you. It is not very hard to make a plausible lie. Some people spend their whole lives doing just that.

[rezecib]

@averagewalrus, Oh, yeah, Vito said at one point that that was planned. You can turn it off very easily with the console, though:

ThePlayer:EnableMovementPrediction(false)

Edit: No, it's fine. There's a lot of activity on the forums, I don't know how I manage to read and remember so much of it o_o. And the post from Vito was ages ago (maybe 4 months?).

[averagewalrus]

 

@averagewalrus, Oh, yeah, Vito said at one point that that was planned. You can turn it off very easily with the console, though:

ThePlayer:EnableMovementPrediction(false)

 Thank you, I'll try to stay more relevant when psoing on threads. Going to attempt to get rid of movement prediction.

[DarkXero]

ryouryou, I would never underestimate modders. My point is not to make the server trust the client, as in, "I traveled a 9000m distance, just trust me" then update it. My idea was the server going "this guy just traveled 9000m distance, do I sincronize?".

 

When you have movement prediction enabled you can move up and down in a responsive manner because you see the client stategraph. You can bump your speed to 999, and try to fly. You will move really fast for a second, but the server will sincronize and make you go to the position you really should be. But this isn't because "you said your speed was 999, but it was 6, go back", but because "you gave me an input, this is the result, sincronize".

 

I understand the difficulty between distinguishing a legit from a cautiously crafted situation.

 

rezecib, I don't know if it's rigid one thing or the other, shooters can use movement prediction and also have the server listen to a guy say "I shot this guy in my game, I was here, he was there, this is the map, verify that the shot was possible".

 

But you are right, this conversation is moot. I just can't grasp how will prediction work when finished and how attacks and items in inventory and everything will be handled if handled. What I do know is that I hate it and I can't truly enjoy fighting creatures unless I'm the host, or have less than 100 ping. And even if I have less than 100 ping but any ping at all, I will not be able to sneak a 3rd attack on the deerclops, because every ms counts.