I have some questions and observations about the undeniable & obvious AI and bots in the game that have real user accounts but my first two are unrelated to the bots.
Clarification on Post-Match Polls: Could an official clarify if player reports are the reason for those post-match surveys? Understanding this would help us better interpret the purpose and impact of these polls.
Player Reports and Security Checks: Are player reports linked to incidents where users are logged out of the authentication server or Battle.net app, requiring re-authentication? If so, what measures are in place to ensure this process is not misused?
Surge of AI/Bot Accounts: Is Blizzard responsible for the noticeable increase in AI or bots playing through seemingly real user accounts?
Possible reasons for Blizzard using AI accounts could include:
Enhancing queue times by increasing the player base.
Training AI in real-time, fast-paced scenarios to refine systems requiring visual recognition and logical interactions.
Potential long-term applications, such as deploying AI models for robotics, drones, military, or other technological advancements.
Blizzard’s Role in AI Usage: If Blizzard is not directly involved with these bots, is there any tacit approval of their presence, given the lack of intervention? Additionally, the sudden surge in voice chat silences, once rare, appears correlated with matches involving bots. Could these silences signify an AI identifier, such as bots functioning as squad leaders to train other bots?
Unethical AI Behavior:
In some instances, bots appear capable of altering game mechanics, such as match scores. Is this related to AI activity, player manipulation, or server-side interventions?
There are concerns that server-side player profiling may lead to unfair matchmaking, potentially forcing players to endure matches with disruptive bots as an implicit punishment.
Bot Capabilities and Ethical Concerns:
Bots exhibit near-zero latency, perfect aim, and sophisticated coordination, surpassing even skilled players. Their ability to execute precise strategies and manipulate player movements raises questions about balance and fairness.
Proposed Rules for AI Users:
Introduce server ping for bots, ensuring it mirrors human players’ average latency.
Implement delay mechanisms for actions per minute (APM) and mouse movements to align with human standards.
Restrict bot coordination to match typical player capabilities, avoiding excessive synergy, especially across opposing teams.
Final Thoughts Transparency and communication with players are essential. If Blizzard is actively developing AI systems, I would be very interested in contributing to these efforts. My background includes consistent gameplay, strategic insights, and a deep interest in AI development. I am particularly passionate about creating unique bots and would welcome the opportunity to test them against the best the game has to offer.
Thank you for considering these points. I look forward to engaging in a meaningful dialogue and potentially contributing to the evolution of Heroes of the Storm.
I did read the code of conduct; addressing the seeming contradiction is the point of this thread; it could point to either gaps in Blizzard’s ability to detect and act on violations or a deliberate allowance to address other challenges, such as balancing queue times.
If Blizzard’s guidelines explicitly prohibit the use of bots and AI for gaining unfair advantages, yet there seems to be a widespread presence of such accounts, it does raise questions about enforcement—or potentially, a tacit acceptance. There is an obvious and undeniable surge in AI activity and bots any reasonable person who plays will see.
If bots are indeed being tolerated to improve player engagement or train AI, Blizzard needs to communicate their stance transparently to avoid alienating their community especially for players who value fair competition which is why i added some restriction suggestions for the bots like matching latency, coordination and aiming to the range of an average player.
What I am claiming here; which is alarming; is that over the past 3 months I personally have witnessed so many bots in matches that I believe that bots might constitute more than half of all currently active accounts. Blizzard should respond to this immediate clarification and action from Blizzard to preserve trust and game integrity
ask how any of this stuff works instead of fear-mongering yourself over ignorance.
When you’re resorting to hyperbole and bad conclusions, it should indicate to you how much you do not know, not that you’ve magically found “50%” of the playerbase.
Ofc Hots has long used bots so you have someone to play with. The real peak online on us server in the evening is about 500-1000 players only. The rest are bots.
They have already completely disabled any balance in games, now you’re put just with those who are at least online, no balance in ranked has long been out of the question, this game has long been removed away from the official page Blizzard cause of this.
Honestly, I don’t even know who all these people are that keep playing with bots in Hots. They must be very specific in real life…
This is most likely caused by a disconnection occurring. Your login session and the game session are separate connections so it is entirely possible to be logged out while still continuing to play. Once the match is over, you get booted to the login screen as you are no longer connected to the service. Seen this happen occasionally when playing Vs AI with human allies, especially if my internet connection dropped during the match.
Extremely unlikely. If they were, the AI would likely play the same as the AI used in the vs AI mode. It would not make sense for them to create a worse playing AI and try to pretend they are normal accounts over just subtly replacing in game players with in game AI.
When such AI is let into the wild it is usually good enough that it will beat low tier human players. In which case no one will notice it being an AI until higher skill levels or it fails spectacularly.
Such AI is rather rare as they often require quite expensive hardware and are very difficult to create and train. There could very well be research institutes using HotS to train and develop AI, similar to how there are ones doing LoL and other games. The unethical ones might even let these AI into real matches, but the number of such players is extremely unlikely to be exceeding single digits at any given time.
Military tech needs to be developed in military conditions. They are unlikely to accept or even permit creations made under civilian conditions as it would be too easy for such creations to leak to enemy entities.
Likely it is due to a reduction in moderation due to resources being pulled from Heroes of the Storm. Most immediate and severe account action requires manual review. Likely not as many people doing that as there used to be.
As far as I am aware it is only possible to be voice chat silenced if you actually use voice chat. Practically no one uses voice chat on the EU server.
Maybe this is a US only issue?
Players cannot alter match scores or game mechanics. You would need to provide evidence of this actually happening, and chances are you are just misunderstanding something.
Due to the lockstep synchronisation model used, every action and outcome in HotS is verified by up to 10 separate PCs. Any that produce deviating results are instantly dropped for desyncing. In theory if 6+ bad actors (which would require actors on both teams) decide to break from usual logic in a coordinated way they might get their outcomes to pass, but even then this would instantly desync all other players not involved in trying to cheat, prevent them from re-joining by repeated desync loop. It would also be extremely easy to detect as the replay would agree with the dropped players rather than those that remained, and be extremely hard to pull-off as you require 6 or more bad actors that are not detected by anti-cheat.
This is not entirely true. Such player-side bots are still subject to latencies.
Server to client latency. Since that is how their client receives the lock-step style instructions from the coordination server.
Processing latency. For a bot to be any good it requires running complicated logic. This takes time to execute. This may be very fast, but it is still a latency non the less.
Client to server latency. As like other players, it has to send the orders back to the server before they are accepted and run by all players.
In-game AI are not subject to these since they are executed as part of the lock-step engine. As such AI such as in versus-AI mode do have near 0 latency, with the only latency being due to “next tick” limitations on responding.
As mentioned above. This would already exist for players even if they are controlled by an AI. The only time it will not is if the actual in-game AI is used, which can be easily detected by lack of inputs in replay files.
This is not a concern for HotS other than some edge cases like TLV. The player only controls 1 hero which is only capable of doing 1-2 things per tick and there are only 16 ticks per second. As such APM is already capped implicitly.
HotS likely has fewer people handling reports than it did at its prime. Additionally most of these “bot” accounts are either real people you are incorrectly calling bots, or are bots created by power levellers/farmers to create accounts to sell. Such bots mostly show up in Vs AI, but I guess they might be using ranked mode now to generate more value.
In general.
If the player repeatedly attack move suicides against the enemy down the middle lane, repeatedly tries and fails to merc, or repeatedly does just the bare minimum to avoid being AFK kicked then it is likely a third party bot of sorts. There is no real incentive to generate more sophisticated bots than this as that represents huge cost increases for no gains. These are commonly seen in lower difficulty vs AI mode grinding experience and gold.
If the player is doing things more complex than the previous point, they are likely a real human. Even if they play very poorly, that does not make them an AI.
If the player is coordinating with other party or random players, they are human. Some players do this very well, especially if 2 or more people involved have significant experience using their heroes and know interactions with other heroes.
If the player is a high rank (gold+) and then suddenly starts suiciding/playing like a bot mentioned in the first point they are likely a real human but working for an organisation that is trying to fix matches to unfairly boost accounts or individuals.
You are claiming but have not provided any proof. I suggest showing replays of suspected bots and letting people judge for themselves.
I do agree that vs AI at adept and lower is plagued by bots. There is also a chance Bronze 5 bottom has these as well, but those are the only places you will see them as effectively not playing will only get you so far.
Here’s how a user might create a bot to play Heroes of the Storm on a user account detailed breakdown hit me up if you need help:
Hardware Requirements
Powerful GPUs: renting GPU cloud services is a cost-effective solution. Platforms like RunPod or TensorDock offer affordable options for high-performance GPUs like NVIDIA RTX 4090.
Cloud Space: You’ll need sufficient storage and computational resources to host your bot and process game data in real-time.
AI Model Development Custom vs. Template
Custom AI Model: You can code the AI using python, develop your own AI model tailored to Heroes of the Storm gameplay. This involves training the model on game data, including player actions, map layouts, and hero abilities basically have it workshopping constantly for a while watching replays. if you don’t have your own replays, Hotslogs website used to have a huge database of replays there’s more around just need to check.
Template - Adapting Existing SDKs: NVIDIA provides tools like the RTX Kit and ACE toolkit for AI-powered game characters. These can be adapted for real-time decision-making and gameplay. TensorFlow is an open-source machine learning framework developed by Google. It provides tools and libraries for building and training machine learning models, including deep learning. TensorFlow is widely used for tasks like image recognition, natural language processing, and predictive analytics. It supports multiple programming languages, such as Python and JavaScript, and can run on various platforms, including CPUs, GPUs, and TPUs. Its flexibility and scalability make it popular for both research and production environments.
Real-Time Visual Processing
Object Detection: Tensor flow would be a decent template to start with for image recognition. Integrate libraries like OpenCV to detect in-game objects such as heroes, minions, and structures3. This enables the bot to make informed decisions based on the game state.
Parallel Processing: Implement threads and processes to analyze the game screen efficiently.
Logic and Decision-Making
Game Strategy: The bot needs to understand game mechanics, such as hero abilities, map objectives, and team coordination. This requires programming logic for decision-making based on real-time data.
Machine Learning**: Train the bot using reinforcement learning to improve its performance over time starting with AI easy game mode.
AI Text Chat for Player Interaction
Integration with AI Chat Systems: Platforms like Perchance offer advanced AI text generation features that can be linked to your bot for player interaction. This adds a layer of realism and engagement and its free with unlimited personality profiles and its pretty top of the line Chat AI you would never know if it was real or not.
Code of Conduct Considerations
Game Policies: Ensure compliance with Heroes of the Storm’s terms of service. Using bots on user accounts as of now violates Blizzard’s guidelines and lead to account bans but if that rule isn’t being enforced then the rules are ineffective anyway and if you have an account that’s 18 months old then you also have some legal protection from bans similar to squatter rights (that would be an interesting court case).
Finally, Testing and Optimization
Simulated Matches: This can be done using the tutorial or training mode to Test the bot in controlled environments to refine its performance. Actually, I would use the StarCraft 2 or warcraft 3 engines with a DOTA map you can get from the arcade for this part. Have the bot testing itself 24/7 with each character integrating strategies from replay database. The replays would be most effective with commentary to explain strategies and mechanics.
Performance Metrics: When you try a live test in vs AI make sure to Monitor latency, accuracy, and decision-making speed to ensure the bot operates effectively. If you get a working bot capable of defeating top skilled players without “cheating” then yes I do believe there would be a strong market for such an integrated AI if only for the real-time-strategy & decision-making development. It would not be hard to adapt that into an overlay avatar with capabilities outside of HOTS game to learn many other games and many other apps its a good start on making a true AI. Preformatted text
Y’all really underestimate how far AI has come the past years… I did some reinforced learning AI models for fun, inspired by those YouTube videos of people making track mania bots that drive perfect and the StarCraft AI arena stuff.
With just the on-screen information you can teach a AI to learn basic gameplay. I haven’t done this for hots, but for StarCraft.
The bot was able to build a expansion, make units and rally them into the enemy base with attack move. I did it differently than the usual scripts though, I used only on screen information and learning by repetition instead of fixed scripts.
This AI wasn’t perfect, but I regularly have seen players on either team (watched replays) that press commands/move exactly like such an AI would… My PC has 32gb of ram and a 10th Gen Intel.
You don’t need a NASA computer to run these anymore
P.S.: I never used this bot I made in any multiplayer situation, only vs the in-game ai and myself on a second computer/account for training purposes.
My goal was to learn reinforced machine learning for a future job prospective (I work in tech, if you hadn’t guessed)
