Hello, this is Team Samoyed! Teamfight Manager 2 aims to consistently share our development progress through dev logs every 2–3 weeks. We appreciate your continued interest and support!
Demo Version Feedback
Through the Steam Next Fest demo version, we received a tremendous amount of feedback from all of you. Among these, we’ve summarized the most common opinions and would like to share how we plan to develop the game moving forward.
Match Engine Feedback
We received especially a lot of feedback related to the match engine in the demo build. We fully understand your concerns, and this demo has been a valuable opportunity to clarify how we want to improve the match engine going forward.
By reviewing your feedback on the match engine, we want to explain in more detail the direction in which we’re improving the system and the goals we’re pursuing internally.
Champions just stay in the bush doing nothing, moving back and forth endlessly.
During a player’s roaming phase, the waiting time is too long and nothing happens during that period, creating a frustrating experience. Internally, several different factors overlap to cause this issue.
Cases where the behavior is actually unintelligent
These occur when there are bugs or missing considerations in the AI’s logic or reasoning process, leading to genuinely poor decisions.
Cases where the AI’s behavior is reasonable, but the reason is unclear to the player
In these cases, the AI’s action may appear foolish even though it has logical reasoning behind it—the player simply can’t see that reasoning clearly.
Cases where the AI’s decision differs from the player’s personal judgment
Even in identical situations, different people may have different opinions. Especially in a MOBA, where situational judgment is highly complex, there is rarely a single “correct” answer. As such, an AI taking a different approach from your own can sometimes feel unintelligent.
Let’s look at each of these cases in more detail. As we mentioned in a previous dev log, our match engine’s player AI is primarily improved through statistical evaluation. The AI operates independently within the engine, allowing us to test an improved version against a previous one and measure which achieves a higher win rate. Naturally, the AI that makes better decisions achieves better results.
Based on the feedback received after the demo release, we have continued to improve the player AI. Under controlled conditions (equal skill levels, balanced drafts, etc.), the latest version of the AI currently achieves about a 70% win rate against the AI from the first demo build. Below is how the newest version behaves under similar circumstances.
Using your feedback, we changed roaming and ganking behavior so that players now decide more quickly—either returning to lane or attempting a gank—rather than waiting too long in bushes.
However, even with these improvements, opinions may still differ. Some might think, “Shouldn’t the Fighter have attacked the Jiangshi instead of backing off to lane?”
In fact, if we force the AI to dive the tower, the Jiangshi can be defeated safely—but had the Jiangshi’s skill activated, the combined damage from the tower and the Jiangshi’s skill could have killed the Cavalry. Of course, if the Fighter had properly taken the tower aggro, it would’ve been a clean dive.
Of the three clips, the third one shows the best play. However, it’s impossible for AI to always make perfect judgments—and that isn’t our goal either. As we’ve mentioned before, while we aim to minimize cases of obviously wrong or bug-induced decisions, achieving professional-level perfect decision-making is not our target.
Our goal for improving the match engine is to make the AI’s intentions clear—so players can understand *why* it acted that way—and to provide indirect ways for players to influence those decisions.
For example, in the second clip, if the player could clearly see why the Fighter chose not to dive, and also had tools to make players behave more aggressively next time, then those decisions would feel far less frustrating.
Aside from these logical cases, there were also many actual bugs in the demo build—particularly due to ability and tactical system revisions. Some of these truly led to unintelligent play. Based on your feedback, we’ve fixed many of these issues and are now focusing not only on smarter behavior, but on *clearer and more understandable* AI reasoning.
Here are some other frequently mentioned feedback points:
During fights, they hesitate and neither attack nor retreat.
This was due to AI performance issues. We found both hidden bugs and deficiencies in the logic that evaluates surrounding situations.
In the initial demo, you can see champions in mid-lane hesitating, taking odd paths, or entering bad fights, showing poor engagement logic.
In the improved match engine, the same scenario shows players committing more decisively to fights.
They rarely recall when low and keep canceling recalls repeatedly.
This issue stemmed from both bugs and logical reasoning—sometimes changes in vision or local situation would make the AI decide that pushing one more wave was better than recalling. We’re working on making these decision changes more intuitive to the player.
Morguard steals happen far too often.
Deciding between “contesting an enemy steal attempt” and “focusing on Morguard” is a complex late-game problem involving multiple interacting players. Based on feedback, we have fine-tuned these behaviors to avoid both overcommitting to Morguard and completely ignoring contest opportunities.
In the demo, players barely contested incoming enemies while fighting Morguard, making the result feel heavily luck-based.
In the updated version, both sides are more proactive about contesting and fighting during Morguard encounters.
Players build strange items.
Item builds were temporarily implemented in the demo. In the future, they will be based on statistics, tactics, and each player’s tendencies, allowing for more rational itemization.
Players don’t follow assigned tactics.
Some tactics were poorly implemented or had unclear intent, and other cases resulted from complex interactions between player traits and situations. Again, we’re focusing on making it clear *why* a player deviated from their assigned tactic.
Champions can play too many roles regardless of position. I wish roles were more distinct.
Teamfight Manager 2 does not aim to force champions into fixed roles. Some champions should be flexible by design. However, if too many are universally viable, the ban/pick phase loses meaning. We plan to strengthen role identity and remake certain champions or skills if needed. Please also look forward to unrevealed champions that will deepen the strategic variety of drafts.
I wish champions had ultimate skills.
Originally, ultimates were planned to be added after Early Access launch, but since that schedule was delayed, we’ve decided to include the ultimate system at the start of Early Access.
Champion HP is too low, making ranged or utility champions weak and fights hard to follow.
This issue connects to the missing ultimate system. Without ultimates, lowering damage would drag out fights excessively. As for match length, we received many comments that matches felt short and fast, and very few saying they were too long and boring. Accordingly, along with introducing the ultimate system, we plan to re-balance by slightly slowing the overall tempo of the game.
I don’t understand what each stat does. The difference between stats feels minimal.
In our internal tests, when a team of players with all stats set to 1 faces a team with all stats set to 100, the win rate comes out to about 25:75. This represents a meaningful impact on match outcomes, but it still falls short of the real-world gap between the weakest and strongest players. We plan to continue improving so that the effects and characteristics of stat differences can be perceived more clearly and intuitively.
UI Feedback
It’s difficult to recognize when it’s my turn to ban or pick, and the animations for bans and picks feel weak.
Since the demo version did not go through a sufficient polishing process, the presentation was lacking in several areas. By the Early Access launch, we will refine the details to make the ban/pick process feel smoother and more immersive.
The indicators for preferred and non-preferred champions look cluttered and unclear.
This issue occurred because the feature was added without enough polishing. We’ll explore various ways to deliver this information in a cleaner and more intuitive format.
I wish the pause screen didn’t cover the gameplay.
An in-game time control UI was originally planned but not included in the demo version. The pause menu in the demo was meant only for settings and quitting, not for pausing the match. Because implementing full time control required significant optimization work, it was excluded from the demo, but that feature has now been completed — you will be able to freely pause, move 10 seconds backward or forward, and jump between highlights.
The text-heavy statistics and information screens feel hard to read.
We’ll continue improving the UI to maintain clarity and readability without making it overly complex.
There are translation errors.
Including a professional translation team during Early Access development would slow progress considerably, so we plan to use AI translation and fix errors quickly with community feedback. Given the current and future quality of AI tools, we believe this approach is viable. We’ll also share efficient and accurate ways for players to participate in this process soon.
Content-Related FAQ
Will the Workshop be available at the start of Early Access?
A basic version of the Workshop will be available at the Early Access launch, and more advanced and detailed modding features will be added through updates.
Will multiplayer be available at the start of Early Access?
Yes. Multiplayer will be included from launch. Unlike the previous game, players will also be able to spectate other players’ matches when their own games are over or while they wait for their turn.
Will controller support be available?
Controller support will not be available at the start of Early Access but will be added in a future update.
Are console versions planned?
Nintendo Switch and Steam Deck versions will not be released during Early Access, but we plan to launch them alongside the full release. PlayStation and Xbox versions are not currently being considered.
Will there be a mobile version?
No. Due to hardware limitations and the heavy UI/UX workload required, we’ve decided not to develop a mobile version at this time.
Early Access Content Goals
For Early Access, our goal is to include all essential features even if some areas are still incomplete. The features we currently aim to include are as follows:
Additional Leagues and Promotion/Relegation System
We plan to expand the leagues from the demo and add second divisions to all regions, introducing promotion and relegation. At Early Access launch, there will be six regions — Korea, China, Japan, Europe, North America, and South America — each with a first and second division, totaling 120 teams.
Transfer Market
Players will move between teams through scouting, transfer offers, contract renewals, and free agency. Since the players you’ve carefully nurtured may end up performing for other teams, you’ll need to pay close attention to financial management.
Coaching Staff
We want great coaches to feel genuinely valuable to the team. Coaches will offer advice in tactics, drafts, and recruitment, influencing your decisions as a manager.
Management System
Beyond tournament prize money, teams will be able to generate income through streaming, facilities, and merchandise. The goal is to create the fun of managing your organization through efficient spending and smart business strategies.
Multiplayer
Players will be able to communicate through in-game chat and spectate their friends’ matches. Our goal is to provide stable connections and enjoyable interaction.
Workshop
Even during Early Access, players will be able to create basic mods, such as adding new champions or adjusting stats. We plan to keep expanding this feature so players can create more flexible and diverse mods over time.
Locker Room Feedback
We’re testing various ways to make the process of giving feedback to players engaging. Our goal is to make it feel like your feedback truly affects player performance. If this feature feels too formal or meaningless, we’ll remove it and reintroduce it later in an improved form.
Custom Item Builds
Since item builds can significantly change champion performance, we’ll give managers some control over them. We’re still determining how deep this control should go, but our goal is to make it intuitive and straightforward while maintaining strategic depth.
Manager Traits
Manager traits allow you to distribute points earned from experience gained throughout seasons. These can grant bonuses in areas such as player training efficiency or revenue generation.
League and Team Customization
We’re developing tools to easily edit, save, and share details of leagues, teams, and players — including logos, names, and stats. This will let you freely modify entire leagues to match your own concepts, even without using the Workshop.
We’re taking all of your valuable feedback into careful consideration and doing our utmost to create a game that truly delivers the experience we intend.
That’s all for this dev log. Thank you for reading!