Project Neural Cloud: CBT Afterthoughts

Welcome back, Professor! Now that the first closed beta test of Project: Neural Cloud has concluded, our writers team decided to give some of their opinions and observations. If you haven't already seen the Project Neural Cloud - Everything We Know articles, please check them out below! This article is a companion piece to these so having a look at them will give more context into what we discuss here.

If you're mostly interested in the deveoper's official statement, see the Dev Response section.

The first CBT has revealed a lot of info about the game, and what we've seen so far is pretty promising! First of all, I have to applaud Micateam for giving such a high level of polish to the game’s visuals. Even in this stage of development the game’s UI and 2D character illustrations are exceptionally good. There are also some great news about the gameplay too!

The most common complaint about the it being boring was resolved in this beta test. While the main battle part still plays like Autochess, the game revealed itself to have surprising mechanical depth with the inclusion of Algorithms and Function Cards. 

There still are, of course, some flaws within the gameplay. One of them would be the randomness of each run. As a fan of roguelikes, my favourite part about this type of game is how each run is different from the last one. Players will naturally settle for an efficient strategy, which often results in “optimizing the fun out of a game”.

Some games employ a high amount of randomness to counteract this and incentivize players to try out new things. This however, is not the case in Project Neural Cloud. The abundance of Cache Coins and Function Cards available in shop nodes as well as some Overclocks all make it easier for players to create the perfect loadout.

While some players might prefer this type of gameplay, I think it will quickly devolve into a simple execution of most efficient tactics. As of now, players who don’t experiment with their team composition and card choices, will most likely quickly become bored of the game.

Something I think the game would greatly benefit from is the introduction of micro-goals within each level. What first comes to mind is something like the weapon Aspects and Bounties in the game Hades. It would be interesting to see cards with special activation conditions or in-level mini missions that reward the player for completing special goals.

On the other hand, the game still has some truly superb gameplay mechanics. The Function Cards, while still very early in development have already showcased their power. One example being Allergic Reaction, a card that made medics deal true damage to enemies around a healed target. While nerfed later on in the beta, this card turned healers like Gin or Persica into the team’s strongest DPS characters.

What remains a mystery about the game is the gacha. As it hasn’t been implemented yet in the first closed beta. Because of this, we still don’t know how the character acquisition and progression elements will mesh with the roguelike gameplay. Earlier this year we saw a creative marriage of these 2 game aspects in Arknights’ special gamemode - Integrated Strategies, where the units' levels were gated behind special items. This resulted in a smooth progression of difficulty from the first to last stages of a run, making the new gamemode the playerbase's favourite. PNC most likely won't have such a system but, it’d be interesting to see different types of stages with different rules regarding character acquisition and progression.

From what i’ve seen on streams, the game itself looks good visually, with a wide variety of characters to choose from. The function card system means that no unit is objectively better or worse, so this game lets you focus on your waifus. The Roguelike factor means that replayability is high, and forces you to use your brain as each stage has a unique combination of function cards. The story itself is pretty good, and the worldbuilding was done well. The story also isn’t mandatory, and even skipping the story gives rewards.

Unfortunately, the game itself seems quite short, although that might be due to it being Beta. People were already at zone 3 of 5 at the end of the week, so unless difficulty ramps immensely the main storyline might be doable within a month. The ‘endless’ mode is also poorly named, as there is a limit of 400m (40 stages).

Another issue is that the entire meta revolves around function cards, so if RNGesus decides a particular function card will never drop, the composition you brought might not function at all. Some of the UI decisions leave much to be desired, such as the placements of certain buttons, or some tabs not being as obvious as they should be. Finally, the tutorial is shorter than I would like, and could use some more work. Hopefully much of these issues stems from the game being in beta.

The stamina system is quite generous, and should never cap if you log in just twice a day. Unfortunately, most of your daily keys are spent just doing the resources collection/fragment sim stages, so not much is left to spend on the main story unless you buy more keys.

Below is a list of suggestions I have for Project Neural Cloud. Keep in mind this is purely my opinion.

• Enemies get function cards of their own instead of just raw stats at higher levels
• More than 5 enemies so we can have more enemy composition variety like in GFL
• Special enemies that take up more than 1 tile
• Impassable terrain tiles
• ‘Tall’ tiles that block line of sight/ranged attacks
• True endless mode that doesn’t offer rewards unlike the current ‘endless’ mode
• PvP in some form, such as theatre
• Daily map with fixed cards like a puzzle map
• Sandbox mode with all the function cards and enemies to play around with
• Changes to resource collection and fragment sims so daily gameplay won’t be 90% farming
• Forced manual for skills like in GFL

Stamina/Keys

I was a bit worried when I first found out during the livestream that Project Neural Cloud would have a stamina system. I hated it in other games - even the ones where you got plenty of refills - so it was a pleasant surprise when I saw that the system gave a very generous amount of stamina. 

To recap the numbers from Part 2, a maxed Key-Collect Port gives 240 max keys and a key generation rate of 0.27 keys/minute, so 3min42s for 1 key and 14.8 hours for 0 to full.

For story dungeons or endless maps, you could get anywhere between 3-10 minutes of gameplay per 10 keys depending on what you're doing. If you run out of keys, you can even attempt a higher depth of Endless Mode with 0 key cost, so you'll always be able to play even if you run out of stamina.

When you don't have much time to play, spending all your keys quickly on the daily Bitcoin/Fragment Collection stages is possible, but they do have a daily clear limit. Players who don't want to waste stamina therefore needed to spend a decent amount of time each day re-playing endless modes or story stages to effectively convert their stamina into drops, at least in the first Closed Beta Test. 

It's fairly likely that the game will add more ways to spend your stamina later (and hopefully more quickly). 

Are there items equivalent to Stamina Potions?

In the first Closed Beta Test, key rewards did exist, but they're instantly added to your key total - you couldn't save them for later. It's possible that this will change, and I personally would like to have some way of saving up stamina; I hated the timed Sanity Potions in Arknights and if MICA does implement a "stamina potion" item, I hope it doesn't expire

Core Gameplay Loop

First things first, the initial Closed Beta Test is something that only showed us the very early game. Think about the first 3 chapters of Girls' Frontline or any other gacha and it'll be clear that this isn't something we can rely on to accurately represent the difficulty or complexity of content. What it does show, however, are the core gameplay aspects, which I'll comment on below: 

PNC's autochess gameplay have been known since the initial public demo and its exact mechanics are described in the prior articles, so I'll skip directly to my thoughts on specifics. 

Map Variety

The autochess premise has a lot of promise (and I like the hexagon grid a lot), but take a look at these two maps: Rotham and Cyclopes: 

Can you tell the difference between those two maps? Well, neither could I, because as far as I can see, there isn't any.

Want different strategies? A simple answer is changing the map layout, for example: 

• A dual-lane map
• A map where you must deploy your units on two flanks
• A map where your units start in the middle, surrounded
• A map with a river in the middle

Of course, adding variety to the tiles on the map can also provide improvements that heighten the strategic depth; for example, any of the following ideas can change the gameplay up a ton even when staying on the same rectangular board: 

• Obstacle tiles like rocks that can't be moved onto
• Defensive walls that your ranged units can shoot over
• Destructible terrain such as trees
• Bushes or other terrain that hide your units from enemy ranged fire
• Hills that offer defensive bonuses
• Quicksand that slows unit movement
• Lava that damages units over time

At this point, there's no official response from the devs suggesting that they will do any of this, but I think it's an intuitive next step and hope the devs will implement them into the game.

Unit Balance

Because the fights are not timed and there are no important DPS checks, players progressing in the first CBT mostly gravitated towards the stall meta, wherein they deployed a team of 2 Defender, 2 Healer, 1 DPS and relied on outsustaining enemy damage to clear stages. Furthermore, the most-picked Professor Skills and most-used Ultimates were almost always crowd control oriented, since they act as a "heal button" by stopping enemy DPS while letting healer skills pop off. 

This, of course, really hurts the viability of units who didn't fit a stall meta - especially since non-sustain teams trying to clear a high depth could lose early due to not snowballing their DPS hard enough. Hopefully, this will not be the case in future rounds of Closed Beta. 

Enemy Balance

While most story enemies are easy to defeat, they scale extremely hard in endless mode and the most noticeable increase is their defense value. Just like how armor is OP in Girls' Frontline if you can stack enough of it, enemies soon become so tanky in PNC that your units deal almost no damage and have to rely on True Damage sources to progress. 

I don't really like this sort of progression gating as it just further encourages stalling rather than compositions that can pull off well-synergized burst/DPS rotations, and hope the formula can be changed to become some sort of percent reduction rather than flat. 

Ultimates

They look great. Too bad you can't skip them. *cough*

One other thing I noticed is that some ultimates are just so good, you never have a reason to use anything else. For instance, consider the above clip of Sol activating her ultimate - it's lined up perfectly and hit every enemy, something that doesn't happen very often and requires the player to pay close attention to Sol's relative position to other enemies and her facing direction. Now consider Hubble's ultimate, which:

• Is basically guaranteed to hit every enemy even if activated randomly
• Stuns them for the entire duration
• Pulls them together and lets the player deal more damage with AoEs
• Deals true damage rather than normal damage

There's no contest. A minmaxing player will be activating Hubble's ultimate over Sol's almost every single time. I've thought of some possible ways to encourage ult diversity: 

1. After using an ultimate, buff the power of ultimates that weren't recently used. 
2. Decrease the amount of meter needed to activate ultimates that weren't recently used. 
3. Separating the ultimate gauge for each character

Unfortunately, since I was not eligible for the first Closed Beta, I didn't have a chance to use the feedback form and have to hope the devs saw my CN social media post. Hopefully some other players thought of the same thing and voiced their concerns. 

The Endgame

As I mentioned previously, it's somewhat pointless to speculate about the endgame since so few story chapters are out. (Imagine trying to predict GFL endgame with just the tutorial and the first three chapters...)

Nevertheless, I can at least comment on some of the long-term goals I foresee players having: 

Maxing every character

Remember how high rarity goes? 6 stars!

With current Neural Fragment income rates (10 clears of frag sim per day for just a few pieces each), fully maxing every character is likely not possible for F2P players because it'll take about 1.5 months per 6-star character. However, since fragments may come from the gacha as well as various one-time rewards or event rewards, the actual pace may be quicker (and players can most likely raise their roster to 4 or 5 stars with relative ease). 

Speculation about the Gacha

Project Neural Cloud having a character gacha is all but confirmed; based on what was datamined from the game files, the following was present in the CBT1 client (though this may not be indicative of the final product): 

• There are two types of lotteries: LotteryNormal and LotteryAdvanced. It is unknown what this means (perhaps there's a friend point gacha?)
• Both 1x and 10x roll options are present. 
• Gacha banners have an unlock date and a duration.
• The gacha can produce both GameObject_hero and GameObject_item.
• The presence of Obj_ConvertToFragment indicates that duplicate units may be turned into fragments.
• There's a variable called dailyLotteryLimtNum which appears to be a max number of times the gacha can be rolled each day, tracked against todayLotteryNums.
• There are apparently plans for a regularly scheduled free roll, due to the presence of clientside UI rendering code that sets the content of text_drawOnce (label of the 1x Draw button) based on whether currentTime is before or after lotteryCfg.NextFreeTime. It is unknown if this resets daily, weekly, or on some other timer. 

Of course, there's no guarantee that any of this will survive into future betas and the mechanics may change, so please take this with a grain of salt. 

Function Card Variety

There's an amazing number of function cards, and there will probably be more to come in the future! There's a compilation of them here: 

As far as I can tell, most function cards fall into one of three categories: 

1. A generally useful but mild buff for Dolls in its class
2. A buff that's fairly strong, mostly usable by a subset of Dolls in the class
3. An extremely broken buff that requires 2+ specific dolls to pull off

Generally, players have gravitated towards building around the third type of function card and using the Overclock Tower to pick an overclock that gives them the ability to guarantee having a copy of that card at map start. 

While this is a pretty interesting way to force a build, I feel that this somewhat defeats the purpose of Neural Cloud being a roguelike - players just pick their OP combo and pilot it the same way, upgrading their bread-and-butter card while ignoring the rest. Hopefully the game will evolve in a way that makes other playstyles viable. 

Miscellaneous Suggestions

For the sake of brevity, I'll limit the remainder of my thoughts to a bulleted list of suggestions:

• Allow endless mode to scale past 400m
• Provide a system to reroll Algorithms
• Some form of auto-battle to let players spend stamina when they cannot play the game
• Non-expiring stamina potions, or a way to "condense" stamina for later use
• Higher bitcoin and hashrate caps
• More diversity in Professor Skills

Here are some highlights from CN players who participated in the first Closed Beta! Notable parts are roughly translated to the best of my ability. The full comments are too long, but you can try machine translating them if you are interested in reading more. 

Sijun (似君)

Original page: https://www.bilibili.com/read/cv8063310

Combat

Function Card tier list: Medic = Defender > Sniper = Special > Warrior

This isn't based on any specific card, but on the overall number of cards needed to make a build viable. Medics and Defenders only really need 3 cards to shine, while Snipers and Special class units need 3-6 cards and you have to give your Medics/Defenders some cards as well. For warriors, even if you feed them 6+ cards, it's hard to get them to really show their value. 

In the first beta test, there aren't enough Function Cards to make too many builds. They're also not balanced very well yet, since some orange cards don't feel as strong as purple cards despite being a tier higher. 

Character Raising

There's no gacha in the first CBT, so I'll comment only on the character leveling and rarity increases. I don't think it's necessary to spend much money to enjoy this game, and here's why: 

• If you have a character, you can max their rarity just from fragments
• Fragment missions give 30 guaranteed fragments per character, and you can get 1-2 fragments per clear, up to 6 times per day
• Some characters start at 3 stars, making it easy to raise them to 4 stars
• Neural Fragments can be bought in the shop as well
• Bitcoins and what looks like gacha currency can buy said fragments

Therefore, as long as you are patient, it should be easy to raise most characters to 4 stars and unlock Skill Level 7. 

As far as leveling goes, the 1-week beta test wasn't enough to max anybody out. I think characters will max out at Level 50 because based on my observations starting at Lv40, EXP books that you can earn each day only gives enough XP to raise one character 2 levels, at Lv43 one level/day, and at Lv45+ you probably need several days of EXP to increase level once, so it's probable that the devs don't intend for players to go past Lv50. Skill training is a grind too and probably even harder than EXP since the skill books are harder to get. 

Algorithms on the other hand, provide the real minmaxing. Because there are so many algorithm types and random stats, players likely have to spend a long time optimizing their boards over and over as they earn more during gameplay. Algorithm multiplier is straightforward to raise but super expensive. 

By the way, there's enough stamina to finish all the limited-clear stages and daily missions each day. 

Stage Design

Main story stages are procedurally generated and they're pretty random. In-stage encounters are basic/medium/hard, but the only benefit of hard encounters is increased coin drop rate and a chance to obtain orange Function Cards. Therefore there's not much need to challenge the hard stages. 

[Sijun goes on to describe his impressions of the game's story as well as provide a lot of feedback to the developers. I have omitted most of them due to length and the fact that it repeats a lot of what is mentioned in this article, but I'll include some of the unique ones]

Suggestions

• There needs to be a better way to spend stamina, like an auto-battle. Players don't really want to stay in the same stage farming it over and over, since the Roguelike genre is about constantly growing your characters and not about starting over in the same stage again. As a phone game, the design should avoid repetitive grinding. 
• Adding drawbacks to function cards (so that there are both pros and cons) can make card stacking more interesting. How about "special equipment" function cards that only show up when a certain character is present in the existing team? 
• The in-stage encounters don't meaningfully differentiate easy/medium/hard, there's no reason for players to challenge hard right now. Maybe buff the rewards for harder encounters a little bit?
• Daily challenge maps could be interesting. As a player, I want to experience the roguelike aspect of Project Neural Cloud maybe 1-3 times a day, and I don't want to spend my limited time grinding easy maps. Maybe keep a ranking board and have player score be determined by relative performance, and let players ban function cards to increase their own score (the higher the pick rate, the more points you get for banning it in your own run). 
• Give us a real endless mode that doesn't end at 40 levels. 
• Hashrate seems quite useless right now, what else is it going to be used for? 

Liu大官人

Original page: https://www.bilibili.com/read/cv8077654

[Portions of review explaining game mechanics omitted. That's actually the bulk of his review - infographics showing off the game screens.]

I've thought this game looked great ever since I saw the PV, and the actual gameplay mix between roguelike/autochess is intriguing as well. Let's dive in!

It seems that there isn't any feature that lets you place Dolls into oasis buildings, devs should add something like that~

Archetypes are straightforward and match their roles: Defender (meat shield), Medic (healer), Warrior (AD), Special (debuff), and Sniper (ADC), I wonder if there will be more classes added later. 

In term of algorithms, I wasn't able to get a single 3-set bonus for orange algorithms even after a week. (There's also no "view sets" button, I have to check my algorithms one by one...) I think Algorithms are important to the Dolls in Neural Cloud, just as HOC chips are to the HOC units in Girls' Frontline! 

For people who are worried about story quality, remember that MICA Team is writing it, so don't worry about it~

In terms of game balance, it's favoring Defender and Medic right now due to their better scaling with Function Cards. Some of the harder stages, you're actually forced to restart if you don't get the right function cards...

Ultimates will tend to charge to full after 2-3 fights. They look pretty cool~

Endless mode ends at 400m, and surprisingly some dalaos already beat the whole thing. I'm stuck at 300m (heh). 

I'm not used to writing reviews, so please forgive me for the poor formatting. In the first Closed Beta Test, we didn't really get a chance to see the player-to-player interaction aspect of this social game, so I hope we'll see more in the next round of CBT. Do your best, devs!

ylohC

Original page: https://www.bilibili.com/read/cv8004214

I think Girls' Frontline: Project Neural Cloud is a mobile game that deserves a perfect score. 

Why do I say that? From a holistic point of view, it's about the graphical design aspect of Neural Cloud. It blends pixel art and 2D/3D extremely well. It uses a lot of geometric shapes that are thematically consistent and easy for players to get used to. I'll try to avoid being too technical about UI design here, the technical aspects don't matter - the player's impressions of the game matters more here. 

How did MICA achieve this level of integration? They used Motion Graphic (MG) artfully. Most people think MG is limited to web design and advertising, but it's actually integral to game design as well. Take a look at these clips from the live stream:

Not only can good motion graphic usage make everything appear smooth, but it's also another means for the game designer to express their art. 

Neural Cloud is also a partially 3D game and camera control can be integrated into the game for great effect. If the camera work is done well enough, player engagement and retention will increase dramatically, which will definitely keep people interested the game long-term.

There's actually a lot going on in the clip above. Pay attention to:

• The camera's originating point and movement
• The moment that characters are loaded into the field
• When the UI appears 
• Deployment sequence
• Where the roguelike map UI element flies into the scene
• ...and the order by which everything appears in the viewport

This responsive and immersive design shows up a lot in what we've seen so far of Neural Cloud. 

Here, pay attention to:

• How the UI flies out of the scene
• The ultimate cut-in
• The transition between the 2D animation back to gameplay

Now you should understand why I rate Neural Cloud with a perfect score! I could go into more depth, but I'm too lazy to cut out more GIFs...

All in all, the only thing I want to say at this point in time is that doing all of this is not an easy feat! MICA Team is going to make a fortune.

Omake

PNC CBT has ended. We have collected suggestions from many professors on forums, and thank you all for responding! Due to the huge amount of content, we cannot reply to all of them, so let our dev team reply to the most repeated suggestions now~~

Thank you all for your support. Later CBTs will have larger capacities, so more professors can join.

PNC’s first CBT has ended well, thank you all for your support! The amount of responses received exceeded our expectations, and we will keep polishing, working to present a better game to everyone.

At the same time, we also received many criticisms and suggestions. From the majority of replies, people approve the base gameplay and design of PNC. Therefore we will bring PNC further to completion with the shop, optimisation, tempo adjustments, etc.

The “Function Card” system is the core of the roguelike. Professors can quickly build their desired combination from rewards. Currently effective function card combinations are somewhat limited, has rather low performance, and is often repeated. But, this system lets us conveniently expand the foundations of gameplay, by letting us be able to quickly design and deliver more interesting and balanced function card combinations.

At the same time, in each “Segment” and stage, we will gradually input function card combinations, and do our best to show the “Anticipation” of a roguelike game, such that each stage remains fresh for professors. In addition to that, more terrain tiles such as special tiles and roadblocks, as well as new enemy designs are already in development. We believe they will be ready to meet everyone soon~~(?)

We did our best to make PNC into a game light on strategy, putting the emphasis on route selection and function card combination, as well as reducing input required in battles. This is to reduce fatigue on the player from long sessions.

For professors that wish to have more influence on the battle, we designed the “tactical support” system.

Each skill tree of the support system has its strengths and use in gameplay. Such as the “Armed Assault” skill in the beta test, is for light influence: Auto executes, crowd control, vulnerability etc.

There are still unimplemented skill trees, such as directed skill release, concentrate/cease fire, auto skill release, ultimates etc. that are already in development and testing. Professors will be able to choose their preferred Tactical Skills according to their style, stay tuned~~

This beta test’s objective was to verify the core gameplay. Due to concerns about development efficiency, we did the bare minimum development and adjustment for the tutorials. This led to the gameplay experience of players not meeting our expectations (such as base building, character leveling, skill leveling, stamina production and consumption, resource stages etc.).

The issues mentioned above will be adjusted before the next beta test.

The project team has a lot of experienced players. Optimising user experience, creating designs that fit instinct, always has been our long term goal. Clearer function card descriptions, a class icon on each card, optimisation of reconnect times, planning mode, LIMIT skill cut-in skipping, one-tap reward collection, base leveling costs and timing optimisation, a restart option when failing a battle, stage selection suggestions etc. The above have all been entered into our to-do list, and we are working on it, we really are.

Anyway, thank you all for your enthusiastic responses. We will work hard to bring everybody a better version of the game. Future tests will have more testing qualification slots, and we are looking forward to meeting with the Professors!

If you paid attention to the front page Community Forum Post, you probably registered for Round 2 of the Closed Beta already. If not, you'll have to wait until the next round of Closed Beta or Open Beta!