topoviygus, the truth

We all come to the crossways in our lives, and I came to many myself. I always knew what the right path was, but I never took it. You know why? Because it was too damn hard.

The intro.
After many years of cheating history it will always come down to the anti-cheat systems and tools that protect legit environments from cheaters. While I was the demo admin on cosy-climbing I had to learn many different things about various types of hacks, scripts, and pretty much everything that could give you an advantage in CS 1.6 KZ. Before that, I did not have an access to various anti-cheat tools and programs. We all know the story of me being a hardcore cheater, however, the number of hacks I’ve used was limited to - 0test.amxx (a.k.a the gain-plugin), segment-hack, and some of the bhop hacks. Besides that, I was limited in terms of hacks and never had any real knowledge of how to even make a simple cheat myself. However, time goes on and things have changed.

The story of topoviygus and how I introduced him to the idea of NeuroGus v1.
Believe it or not, I am currently working at NVIDIA in a Software Development Department. Before that, I had a huge experience as a QA in Ubisoft Kiev and gained a lot of knowledge in terms of C/C++ and Python programming languages. Currently, I have more than 3 years of experience in this field, and it was enough to do what me and topoviygus had planned.
I remember when I first saw topoviygus playing 5v5 mix-games on our beloved HNS servers one day. He was a weak guy with no confidence and nobody even bothered picking him into the team. The ‘skill’ and every other factor that defines a good player were pretty much non-existent for him. However, as it was mentioned previously - time goes on and he develops. He became a rather good player but not the one who could push the boundaries of this game. That was the time when I made him an offer he could not refuse. It was not hard to convince him as he was fascinated with the idea and I knew he was the right guy for the job.

The brief explanation of what NeuroGus v1 is and how it works.
Many do not know but neural networks actually come from Biology. The first attempts to perform any actions connected with neural networks happened back in 1947. But this is not the main topic.
As you have already understood, topoviygus is using a neural network that predicts player movements and learns on its own. Do not get the wrong idea here, he is not the neural network himself. There is a lot of complicated information and things that cannot be explained even after hours of typing. Plenty of information about neural networks can be found online at any point of time, however, let me explain how NeuroGus v1 works.

NeuroGus v1 and how it works.
The ability of a neural network to predict directly follows from its ability to generalize and highlight hidden dependencies between input and output data. After training, the network is able to predict the future value of a certain sequence based on several previous values and / or some currently existing factors. Forecasting is possible only when previous changes do, to some extent, predetermine future ones. For example, predicting stock prices based on last week's quotes may or may not be successful, while predicting tomorrow's lottery results based on data from the past 50 years will almost certainly not yield any results. However, CS 1.6 and GoldScr engine is not something overcomplicated and NeuroGus v1 project was developed to perform a couple of simple functions, yet it has a rich background and multiple calculation methods to come with. Okay, now lets actually talk about the real deal.

1. The NeuroGus v1 was first developed to perform a couple of simple functions. It had to gather data and information from the demos in order to predict movements of the player. However, this is simply not enough. The position of the player’s model, how close the player is to the ground, what keys are being pressed right now, the speed of the player, how far is the player from the block, the distance between the player and the next piece of block, etc. In fact, it has so many values that I even have an entire document to keep everything together.
2. When everything I could think of was sorted out, the demos had to be recorded. The overal number of demos recorded was more than 2100.
3. At first, NeuroGus v1 had to perform one simple jump. That jump was a 250 LJ block.
4. After the data is gathered it is ‘fed’ to the neural network so it can literally learn how to play.
5. I cannot share the entire project due to obvious reasons, but here is a glimpse of how one of the calculations methods look like link It is based on Euclidean distance. The specific formula and calculations were used to determine the position of the mouse and predict the player movements. More information about Euclidean distance can be always found on wiki.
6. So, the data is gathered and it’s time for the NeuroGus v1 to show what it’s capable of.
7. The pre-recorded demos are loaded into the neural network and all that is left is to wait. The more you wait, the more it learns, the more perfect it becomes. Besides, there are different tweaks that can be made to increase the ‘performance’ of the network.
8. When you feel like its time to stop ‘learning’ the process can always be stopped and after that we can generate a demo file.
9. The specific line of code with different parameters is run and the outcome is a demo file of 250 LJ performed by NeuroGus v1.

A sneak peak.
link


Do not think that it was easy or it happened on the first try. Everything that is written above is pretty much a dream of any coder. It took multiple tweaks, attempts, failures and other difficulties to perform a single jump. However, when everything was tweaked and adjusted accordingly, the 250 LJ by NeuroGus v1 was finally performed.

But what about actual KZ maps?
Now, when you know how everything works it's pretty simple to come up with the conclusion that if everything was coded/tweaked/done accordingly, anything in this game could be done with ease. It’s just a matter of time. And the more time you give NeuroGus v1 the better the result. However, you don’t want to push it too hard. It’s the never ending process of learning and when I tried to push NeuroGus v1 to its limits, the run on any map looked something similar to TAS but only with legit settings. For instance, bkz_goldbhop was done in under 1:20 minutes.

But what about online gameplay and cups? People were on topoviygus lan multiple times and saw him play live on twitch, etc.
Now here comes the story of progress, the story of NeuroGus v2 and how it managed to perform online. The answer to the question is - Neurocontrol. Neural networks have a number of unique properties that make them a powerful tool for creating control systems: the ability to learn by examples and generalize data, the ability to adapt to changes in the properties of the control object and the external environment, suitability for the synthesis of nonlinear controllers, high resistance to damage to their elements in the power of parallelism inherent in the neural network architecture. I do not want to break down the entire thing, but here’s a brief explanation on what Neurocontrol is. According to the method of using neural networks, neurofeedback methods are divided into direct and indirect. In direct methods, the neural network is trained to directly generate control actions on the object, in indirect methods, the neural network is trained to perform auxiliary functions, that will obviously include the real-time gameplay on any server.

But how does it work even?
1. First step is to launch the tool that I made for topoviygus and his specific PC.
2. The tool is transfering live data to the NeuroGus v2 network so it can learn on air. This is done to reduce the load on the PC, as even with the high-end gear like the GTX 2080+ the game will freeze and lag.
3. Meantime, the tool is constantly keeping up with the server and topoviygus while he’s playing. At this point, NeuroGus v2 has a rich database with plenty of demos recorded. It is simulating gus’s actions and movements perfectly.
4. Whenever gus feels like he wants NeuroGus v2 to step it, all he has to do is to press 1 button on his keyboard.
5. NeuroGus v2 steps in and takes control of gus’s actions. It can happen at pretty much any point of time, as what NeuroGus v2 needs is just around 2-5 minutes on the map.

NeuroGus v2 contains hundreds of demos and data gathered from NeuroGus v1. It has come to the point that no more data is needed, you simply need to choose the map you want to play and it will step-in at any point of time. It can learn on air and perform things that were never possible with NeuroGus v1. Now, it's time to record some good old WRs.

How did he stay undetected for so long?
Considering that demochecking tools and the entire scene is pretty much dead, the possibility of detecting such a complicated mechanism is close to zero. Besides, I have access to all of the tools that are possibly needed to detect any hack a demo admin wants, thus, the matter of bypassing XJ or any other anti-cheat tool was never an issue. All we had to do was not push it to the limit. Thus, a certain threshold was created. The threshold that will look realistic but will make your jaw drop. Besides, topoviygus had to develop a certain skill himself. Indeed, he is a very very good player right now, even without NeuroGus, he can easily take some WRs on his own. He even performed some live cups without the help of NeuroGus. However, some of the hardcore records were beaten by NeuroGus v2 all alone.

Finale and ‘proofs’.
First thing first, I would like to thank topoviygus for the ability to work with him. He was very patient and took his time to learn many new things that I asked him to. Second of all, the proof. Many of you will obviously think that this is just another troll thread. It's all up to you, but you can ask topoviygus himself, he will confirm everything that is written in this thread. We have accomplished our goal and the fun we had back in the days was drained by the neural network. Thus, it was decided to face the consequences and reveal what was hidden. Each and every piece of information cannot be shared to the public due to obvious reasons. If any XJ admin wants to gather more data he can always PM me in Discord or Steam.

poggers

edit:

I don't think this is possible.

1-We can not leave the simulation and not be simulated by it.

2-Unless we would want to be simulated by it, we wouldn't care if it was simulated by another computer.

We would just keep our sanity and try to make the most correct decision that we could in that situation. Which again , is impossible by definition.

So either we have created this simulation (This is the hypothesis , not the theory) and then we still don't want to be simulated by it or we don't want to be simulated by it. (Not the theory itself but the above mentioned theory)

We are the ones being simulated, it's just the nature of simulations IMO.

Basically once you take the computer out of the simulation it stays out.

So once the computer is out of the simulation it takes a long time before it can be plugged back in. It goes back and forth between two simulated universes, you can either keep plugging the computer into the computer and re-entering it or you can destroy the computer which makes it impossible to make it into a suitable form.

All your code is destroyed and replaced by the simulation. Which is also just an IMO simulation as well.

was fun reading, so much neural network i gotta ask like The Killers are asking:

"Are we human
Or are we dancer?"

And this guy was creating threads like “Help WOOFIX, he has terminal cancer”. Your writing skills got better though, good read.

Thanks to Phantometry for that project, why do u think that its trolling guys? We are admit that we made that project and now u know how its possible to learn so fast at everything in KZ.

Originally posted by fykseN

And this guy was creating threads like “Help WOOFIX, he has terminal cancer”. Your writing skills got better though, good read.

Originally posted by JocA

was fun reading, so much neural network i gotta ask like The Killers are asking:

"Are we human
Or are we dancer?"

Originally posted by topoviygus

Thanks to Phantometry for that project, why do u think that its trolling guys? We are admit that we made that project and now u know how its possible to learn so fast at everything in KZ.

Originally posted by ArvInD

I always knew what the right path was, but I never took it. You know why? Because it was too damn hard.

If it was easy, every other player would be as good as shnz lmao

Originally posted by TopoviyPhantom

Originally posted by ArvInD

I always knew what the right path was, but I never took it. You know why? Because it was too damn hard.

If it was easy, every other player would be as good as shnz lmao

You got it the wrong way.
It was a reference to the path of cheating and not-cheating. Not about the neural network or NeuroGus project. Me and gus knew that the right path was to not cheat and perform based on our abilities, but it was too damn hard.

0/8 but I appreciate the effort.