Skill curves & Flow

The focus of this post are skill curves—what they are; how some of them might look; and how they can help with defining skill floor and skill ceiling. After that we can look at the connection between skill curves and the flow channel. All this might give us a better idea what's the issue with Co-op difficulties and why some players call Zeratul overpowered.

Skill curves

I will define a skill curve as a curve representing power (or effectiveness) as function of "skills". In this context skills will mean a combination of mechanical skills, knowledge, progression and effort.

Skill curves are useful for visualization and explaining ideas. I won't define Skills or Power in Co-op setting. But even if I did, commanders will benefit differently from different types of skill. Skill curves are a simplification but a useful one.

I also have to note that I don't have actual data for StarCraft Co-op. So don't try to read too much into those diagrams. If you want to see some real data, Riot's Greg Street showed in his GDC talk[1] how they visualize champion winrates as a function of player skill or as a function of the number of games played with each champion (~28:40).

We could see skill curves as a subclass of learning curves. Learning curves will frequently use time on x-axis, and so I felt it's necessary to separate them. In a game like StarCraft, increased number of games played without deliberate practice will often lead to only a minimal improvement. The relation between learning curves and Co-op is better described in my older post about Progression and Strategies.

(Note: Skill curves can be different between normal gameplay, speedrunning and various mutations. Here the focus is on normal gameplay.)

(Note: As has been pointed out to me, slopes of curves should be more different from each other. Raynor scales better with skill, meaning his curve should be significantly steeper than Zagara's. )

Here we have skill curves for Zagara and Raynor. Raynor has relatively high skill floor—meaning you need higher skill to play him at the very basic level. This corresponds on the diagram with Raynor's skill curve intersecting the line of minimal effectiveness more to the right than Zagara's skill curve. In Co-op, we can think of this minimal effectiveness as being able to beat casual difficulty. It's relatively easy for Zagara, and so she has a low skill floor. Having low skill floor in any game is a good thing as more players can enjoy the game, and it's easier to start playing.

skill floor –

the minimum amount of “skill” required to achieve minimal effectiveness

(Note: Sometimes a skill floor is defined the other way around, but this is more commonly used definition.)

If we look at higher skill values, we can see Zagara doesn't scale with skill nearly as well as Raynor. At certain point Zagara's skill curve plateaus. We will call this point a skill ceiling. Zagara reaches this point at lower skill, and so we will say she has a low skill ceiling. On the other side, Raynor seems to scale with skill very well. We will say his skill ceiling is either very high or that he doesn't have one. A high skill ceiling is generally a positive thing.

skill ceiling –

the amount of “skill” beyond which power increases only negligibly

Now, if we add Zeratul's skill curve, it might look something like the diagram below. Zeratul starts very strong and has a low skill floor. This is thanks to all his units performing at least decently, limited macro and minimal unit control necessary. He scales better with skill than Zagara (his skill curve is above hers), and at higher skill ranges he falls inline with other commanders.

I marked a high skill ceiling for Zeratul in the diagram, but I can only speculate about its exact position between Zagara's and Raynor's skill ceilings.

Flow channel

"Flow state" is a type of focused state of mind that accurately describes the experience of playing a game like StarCraft. There are few key components necessary for a player to stay in this flow state[2]:

• Clear goals
• No distractions
• Direct feedback
• Continuous challenge

The last key component will be interesting for us. It can be described as staying in a "flow channel" – a space of challenges that are not too hard or too easy.

Flow channel –

narrow margin of challenge that lies between boredom and frustration

All diagrams we have seen so far are similar. They all have Skills on x-axis, while Power and Challenges on y-axis are closely related. Let's combine diagrams.

In Co-op, we have 4 difficulties (or challenges) players can choose from. These difficulty options can mostly keep players in the flow channel between anxiety and boredom. However, we can see the issue with Brutal difficulty. It tries to cover too wide range of skills. Even though commander skill curves start to plateau, the level of challenge Brutal difficulty provides will go deep into "boredom" territory as commanders' power scales well beyond provided challenge.

Additional difficulty could cover those high skill ranges, but it would also make matchmaking harder. Alternatively, trying to spread difficulties to cover wider skill ranges would mean that gaps between difficulties would widen, and players could stray into anxiety or boredom regions in those gaps. Brutal mutations cover high skill ranges, but they are not a complete solution in their current form of custom and weekly mutations.

This last diagram also lets us see better the reason for concerns about Zeratul. While he is not overpowered at high skill ranges, his minimal skill (and with that minimal effort) required to play on Brutal difficulty is possibly lower than for other commanders.

main points

• Skill floor is the minimum amount of “skill” required to achieve minimal effectiveness
• Skill ceiling is the amount of “skill” beyond which power increases only negligibly
• In this context skill means a combination of mechanical skills, knowledge, progression and effort
• Low skill floor and high skill ceiling are generally good as they make the game interesting for the widest range of players
• Commanders have different skill curves making them strong at different skill ranges
• Difficulties should keep players in the flow channel if possible

Interesting reading materials:

• [1] Balancing 'League of Legends' for Every Player, from Bronze to Bengi (Greg Street / GDC 2017)
• [2] The Art of Game Design: A Book of Lenses (Jesse Schell)
  

Achievements

Achievements can have various functions—challenging players, providing creative limitations or showing various parts of the game a player could otherwise miss. I will focus on commander specific achievements since other achievements in Co-op (levels, difficulty, missions, ascension, and mutations) are fairly boring incremental achievements.

Commander achievements

Commander specific achievements will often serve as an extended tutorial helping you master the commander. They will direct you to the main strengths of each commander, encourage you to take advantage of those strengths and optimize your play around them. Achievements can also point out strategies that might not be readily apparent.

The majority of commander achievements are of two simple types. The first type encourages you to play to the strengths of the commander by requiring you to complete a certain task over several missions. This gives you enough time to get familiar with those commander strengths. Raynor's Feels Good, Man achievement is a good example of this. The lesson is simple: medics are good, and they can also heal allies.

The second achievement type rewards players for getting better at using tools that the commander has available (e.g. calldowns or heroic units). In this case it makes sense to limit the task to one mission or one calldown usage. For example Raynor's Hyperactive achievement encourages you to learn how to get the most out of the Hyperion calldown.

Swann's achievements work in very much the same way. We are taught that we should use Vespene Harvesters and encouraged to learn how to use Laser Drill effectively.

Combining both of these achievement types is important. Having only the first type that spans across several missions can feel like a grind without actually challenging yourself and learning anything. An example of this is Han and Horner's Did It Work? achievement. Magnetic Mines are free, and players don't need to detonate 5000 of them to realize they are useful. The achievement is not even for 5000 kills which would encourage using them better. This achievement did not work.

Buy Buy Buy (Buy Buy) is an excellent achievement because it shows you the strength of heavily investing into gear and makes you optimize Tychus' buildorder.

WHAT COULD BE BETTER

Devolution Retribution is a strange one. Five hundred kills in a single game are too many for any normal mission. Both cumulative 500 or 50 in a single game could work. But it's unfortunate that the more interesting Avatar of Essence gets this achievement while Avatar of From is less flashy, underperforms and gets no achievement. You also have to kill devolved units yourself which doesn't promote cooperation.

Zagara's Queen of Destruction is too generic. What does it teach you exactly? That you shouldn't get carried every mission?

Alarak would benefit from an achievement encouraging using Ascendants as it's not obvious how good they are. Trying out new strategies is not easy, but achievements can help to motivate players to do so (more on it here). Something along the lines of "Power Overwhelming: Deal one million damage with Ascendants" might work and could be completed in just few games on Oblivion Express.

Stukov could use an achievement for Bunkers and how to use them effectively. Perhaps something like "Free Army: Spawn 100 Infested Troopers at once".

Abathur would benefit from an achievement focusing on getting an Ultimate Evolution under 4 minutes. It would show one of Abathur's biggest strengths. Many players are still not aware of it.

Raynor could use an achievement encouraging using multiple Orbital Commands.  New players often won't realize how strong his macro can be.

CONCLUSION

I think achievements in Co-op are in a good shape. Most of the time they direct players into using commander's kit and getting better at it. There could be fewer grindy achievements, but both types have their place. I would like to see more achievements that direct you at exploring true optimal strategies. I mentioned few commanders, but even Zeratul would benefit from an achievement related to projecting his static defenses.

Interesting reading materials:

• The Cake Is Not a Lie: How to Design Effective Achievements (Lucas Blair) (full dissertation)
• The Zeigarnik Effect and Quest Logs (Jamie Madigan)
• Make Them Work for It: Designing Achievements for Your Games (Darran Jamieson)
  
• Achievements Considered Harmful? (Chris Hecker)

Leaderboards

Co-op leaderboards have been under work for a long time, but we know very little how they will work. In my opinion they have a potential but also their share of problems.

I see the main difficulty in making them interesting for players of all skill levels. If rewards are tied to reaching the top of leaderboards, then majority of playerbase won't achieve that, and there is very little reason for them to care about leaderboards. The interest in leaderboards would go down with time except for a few players competing for top spots.

weekly mission focus

I think there are two things that could make leaderboards interesting for a wider audience. First, there could be rewards for set challenges—for example finish a certain mission under 12/14/16/18 minutes. These timings are just an example, leaderboards will likely use points combining various performance metrics and difficulty settings.

Each of these challenges would provide in-game experience similarly to weekly mutations. The goal is to engage players in leaderboard ranking. Once players try to complete those challenges, they can compare their own score with that of the top records. And while they are at it, they can look at mastery allocation and replays to improve their own game.

The second part is to focus on one map at the time, changing it weekly and resetting its leaderboard. This would bring a new challenge each week—letting community to focus on sharing tips and learning maps one by one.

Concept – User interface of Co-op leaderboards

Combined leaderboard

Apart from weekly leaderboards there would be a leaderboard that aggregates and combines scores from each map including weekly leaderboards. Resets would happen only after big gameplay patches and for specific commanders. This combined leaderboard is likely what has been in development. It should contain both combined ranks and records for individual maps—letting more dedicated players track their progress.

Other notes

• Placing in top X% in either leaderboard could be rewarded with a special portrait or banner. However, at least some reward should be achievable for majority of players each week.
• Support for solo play and leaderboards would be nice, but I don't think it's likely to happen.
• Limiting valid records to arranged teams could potentially reduce toxicity.
• The calculated score for missions should be fair and transparent—making it clear what can be improved.
• Should there be set challenges for only certain commanders? It would add more variety. But since many commanders have to be purchased, it might not work as well.

What are your thoughts on leaderboards, and how would you make them work?

Pushing Forward & Rewards

Let's examine mechanics that provide rewards from fallen enemies, their connection to aggressive gameplay and how they can lead to snowballing. Then we look at few other games that use them to encourage aggressive gameplay.

Biomass and Essence

Abathur and Dehaka have been doing very well in both speedruns and mutations. The power they gain through killing enemy units lets them kill enemy units even faster. In the hands of experienced players, these commanders can snowball and finish missions very quickly or stay one step ahead of dangerous mutations. What things are required for a commander to snowball?

Abathur and Dehaka gain a significant part of their power through Biomass and Essence which drop from killed enemy units. This is the main prerequisite that lets them snowball. Another important factor is that the resource investment required to take advantage of gained Biomass and Essence is minimal. Dehaka himself becomes available for free, and Abathur only has to build three Roaches, tech to Spire and get another three Mutalisks. This independence from economy assures that snowballing isn't bottle-necked. In case of Dehaka, Primal Pack Leaders also require only a minimal investment. The last thing necessary is the ability to efficiently trade units in the early game. This was the main issue with old Abathur before his masteries and Ultimate Evolutions got buffed. At this moment both Dehaka and Abathur have tools to be aggressive from the very start.

All this is very useful in both speedrunning and mutations, where not having to rely on economy and being one step ahead of the enemy is crucial. To sum up, there are three things required for a commander to snowball:

• Rewards from killed enemies that provide a significant increase in power
• Minimal economy investment
• Ability to trade efficiently from the start

Direct Kills Speaking of Biomass and Essence, I have to appreciate that these mechanics work significantly better than rewards for direct kills which were common in older RTS and RPG games. Rewards for direct kills lead to players, units or characters competing for kills. Such zero-sum mechanics usually don't fit well into cooperative modes. Toxic Nests might be the only mechanic in Co-op where a player can hurt his ally by doing what he essentially should be doing—killing enemies. This is one of reasons why reducing bonus biomass from Toxic Nests in the upcoming Abathur's revamp is a good change.

DOOM & DIABLO

There are two games I want to mention that use such rewards to encourage aggressive gameplay. In Doom 2016 players are rewarded by additional health and ammo from enemies if they kill them with a melee attack. This is one of many mechanics—reduced need for reloading, no health recovery, better projectile dodging, etc—that supports their "push forward combat". In linked video from GDC 2018, id Software's Kurt Loudy & Jake Campbell detail the philosophy behind their combat design.

Diablo 3 health drops were chosen as the most fitting health recovery mechanic for constant action. In the video from GDC 2013, Blizzard's Wyatt Cheng talks about many iterations of heath recovery they went through. The following part about Command Input is extremely interesting as well.

Conclusion

Abathur and Dehaka benefit greatly from these mechanics—placing them in top spots in both speedruns and mutations. I don't think snowballing was Blizzard's design intention for these commanders. I find it more likely that these mechanics just fit well to their fantasy. Abathur's early units and Toxic Nests hint towards intended gameplay that is more passive. On the other side, aggressive gameplay was definitely a focus for Dehaka which spawns earlier than any other commander.

Snowballing can lead to very enjoyable gameplay. However, in case of Abathur and Dehaka, economy and supporting army might have become little too inconsequential. This removes some depth from the commander gameplay. Upcoming revamp update for Abathur seems to address this too with some nerfs to Ultimate Evolutions.

Wheel of Misfortune

This week we had about a day of Wheel of Misfortune, making this its 9th appearance in 119 weeks of weekly mutations. And while it was promptly replaced by a brand-new mutation, I will take a closer look at it.

The issue I see with this mutation is the pairing of high difficulty and high randomness. A small amount of randomness adds variety and leads to surprises that make the game more interesting. However, the goal should be to provide challenges that can be overcome. Being completely at the mercy of randomness is neither fun nor does it encourage continuing playing. In Self Determination Theory, one of the intrinsic needs is competence that can be described as "seeking to control the outcome and experience mastery". This is crucial for players' intrinsic motivation and thus for their desire to play the game. I finished every one of Wheel of Misfortune weekly mutations, but I can't say it was because of my skill or good decision-making; it was only luck. I see this lack of player agency and perceived competence as the main problem.

We can feel unfair and unavoidable effects of randomness in the real world, but their absence in games is partly what makes games great. No sane game-designer would let a random meteorite or lightning hit you without any chance of preventing it if it meant losing twenty years of progress. The cost is nowhere near as high in a match of Co-op, however, players will get invested in it very quickly.

What could be changed about this mutation? The goal is to either reduce overall difficulty level so players can consistently overcome its challenge; or reduce the randomness to the level that it won't be the main cause of players' failure. To lower difficulty, decreasing the number of active mutators to two could work. Challenge would be reduced to the levels of easy weekly mutations, but each game would still be unique. As for lowering randomness, removing certain mutators from the pool would help. Ideally, each mutator would have a difficulty score, and the total score of all active mutators wouldn't rise above a certain threshold. Notifying players about upcoming mutators in advance could give players more control over the situation. However, intervals between mutator changes would have to be longer so players wouldn't get overloaded by messages and had time to prepare.

Conclusion

High randomness doesn't pair well with high difficulty. It's important that outcomes are kept in the hands of players and not randomness. Mutations and mutators that take control from players should be handled extremely carefully. A similar example are Fatal Attraction or Micro Transactions mutators that go against what many players enjoy about StarCraft 2—controlling units. This again will naturally turn many players away.

Update: relevant video on the difference between challenging and punishing games

In Defense of Defense

There are many players that enjoy defensive missions and would rather focus on base-building than rushing against mission timers; players that will tell you "I defend" at the start of Dead of Night map and will enjoy doing that. Ideally, there would be a wide variety of maps, each enjoyable in its own way but playable by everyone.

Some players don't like auto-scrollers, missions on rails that you cannot speed up in any way (e.g., Temple of the Past, Mist Opportunities or Oblivion Express). It might seem that survival maps would suffer from this, however, they don't need to. Players could be given an option to speed up enemy spawns, making the mission faster and upping the difficulty. Miner Evacuation works a bit like that, but it's not a survival map. Alternatively, the part featuring infinite survival could be optional and at the end of mission. Dead of Night could have worked that way. Unfortunately, performance issues and the lack of scaling for attack waves meant it couldn't fulfill that role.

They Are Billions (single-player survival strategy game)

There are many maps, modes or games that could serve as an inspiration for full-fledged survival maps (several campaign missions, Blizzard's own Co-op survival map Left 2 Die; games like They Are Billions, plethora of arcade maps and even horde modes from other genres). The original Last Stand mission in Legacy of the Void campaign was thematically very fitting for infinite survival-type mission. You had to destroy Zenith Stones and lure enemy on Shakuras. Both offense and defense were required. And more importantly, staying alive longer made sense with respect to that mission.

Left 2 Die (StarCraft II Co-op survival map)

Let's focus on the other variant where the survival is optional and at the end of mission. One option is to have a soft timer for the final objective. With that I mean players have to finish before they get overwhelmed or lose because of other condition over which they have control. Preceding objectives could have standard hard timers – players are required to accomplish objectives before some arbitrary timer runs out. Hard timers might be useful as it's better to let players lose than let them play without a chance of winning. However, Dead of Night gets away without any hard timers; players can only lose by getting overwhelmed and losing their base.

Bonus objectives could be used to put emphasis on the defensive part of the game. One bonus objective could open more enemy spawns and increase pressure on defenses. Additional bonus objective could give experience periodically and encourage staying alive longer.

The downside of this approach is that both players need to stay in-game for the survival part. Another option, which require a special action from each player to win, is less elegant but lets one player to leave (and win), while the other can stay in-game. For example each player has to order a transport ship to lift off, or deliver his own cargo to win.

Few things required from the map or such that would make the survival aspect better:

• Soft timer on the final objective
• Interesting defensive positions and choke-points; decisions whether to defend natural expansion or not
• Attack waves sizes and frequency increase with the mission time
• If there are any special units attacking, increase their frequency too (e.g., Nydus Worms or Stank)
• Small periodic experience bonus for staying alive longer
• Personal best time shown on the character profile
• If best times were meant to be compared, the main difficulty cannot come from basic Co-op attack waves as they vary in strength too much

To sum up, challenging (infinite) survival mission would be a good addition to the Co-op map pool. And it doesn't have to be an autoscroller or interfere with the standard playthrough of the mission.

Gameplay Variety and Modularity

Intro

Let's look at gameplay variety. The number of possible gameplay scenarios a player can experience in Co-op is proportional to [#commanders] x [#commanders -1] x [#maps] x [#enemy races]; where "#" means the number of. Plus there are other less significant factors such as player and enemy unit compositions or map specific patterns which increase the number further. The cooperative nature and progression system help too, and the potential of mutations shouldn't be underestimated either.

This relatively high number of possible gameplay scenarios leads to a greater replayability than is usual for campaign missions. Those are often limited to one scenario per map with small variations provided by global upgrades and unlocks. However, even with this, Co-op can't measure to the competitive Versus mode, which has almost unlimited replayability through player contest and self-improvement.

Because of its non-competitive nature, having enough content is particularly important for a game mode like Co-op. In this post I will argue that having high modularity is a good way to provide more gameplay variety. However, it might not be the optimal thing to do in Co-op now. It would be easier to take advantage of tools that are already available. And by that I mean weekly mutations, which were abandoned for some time, and custom mutators, as there is little incentive to play them. The first step could be going back to making new weekly mutations, creating a lobby system for custom mutators, adding a matchmaking queue with random mutators or something completely different.

Modularity

A system is modular when it can be subdivided into smaller parts "modules" that will work in a variety of configurations. Modules are connected in a standardized manner, and their main strength is that they can be easily changed and reused. While the term modular is usually found in connection to games in modular level design, it might be helpful to think of certain parts of a game as modules. Then we could look at gameplay variety as a number of all possible module combinations. Modularity might be reminiscent of systemic game design, but modules usually lack the level of awareness and interactivity systems have.

Co-op maps are already very modular. Commanders are created separately and are not restricted to certain maps. Enemy compositions and most of the functionality are shared between maps. Mutators are a great example of a module that affects all maps and commanders and yet functions almost independently. If I put Just Die! mutator on a newly released map and play it with a new commander, it will mostly likely just work.

It's generally better to build a system from the ground up with set modularity. Increasing it later takes a lot more time, and it's likely not a realistic change for Co-op now. However, it's interesting to try to learn from the current game, and see what could be improved. Let's look for places that would be suited for the modular approach and see what their hooks would be. I will use the term "hook" as a place or an event where modules communicate between each other or with the main system. Hooks have to be universal enough to fit all possible configurations. For example, many mutators come into effect when a unit is created, takes damage or dies.

Bonus Objectives

Bonus objectives are prime targets for becoming modular. They stand alone and do not significantly affect the difficulty of the main mission. Mixing them up would greatly increase the gameplay variety. They could be even completely replaced by other events.

To make them modular, all bonus objectives' data and triggers would be moved in the Allied Commanders mod which is shared between all maps. Each separate map would then call for bonus objectives through a function distributing them. This way, every map would have access to all bonus objectives and their assets. The advantage of this approach would be that a simple change to the distributing function could switch a bonus objective for another bonus objective or event. Also, new bonus objectives could be easily introduced into older maps.

Apart from randomizing and introducing new bonus objectives, there could be more interactions with mutators, and bonus objectives could be temporarily altered or replaced by special events such as in-game or cross-game promotions. Players could hunt pirate ships because of a new World of Warcraft or Hearthstone expansion, or they could collect parts from bonus objectives in a time-limited event, which would reward them with a new skin, mount or a companion in any Blizzard game. Daily or weekly quests could alter bonus objectives. Overall, modular bonus objectives would open a lot of options.

Problems with randomizing bonus objectives

• Hooks have to be implemented carefully to ensure good compatibility. The most important part is choosing the best location. Even then, some bonus objectives, such as destroying trains, could only appear on some maps.
• Voice lines for map specific announcers have to be flexible to accommodate for various bonus objectives. Optionally, there could be announcers just for bonus objectives.
• Bonus objectives are often tightly integrated in a map so they fit the theme and gameflow. While it should be possible to maintain gameflow, not every bonus objective's theme is compatible with all maps.
• Randomized bonus objectives have to be communicated early enough so players can adjust their gameplay (managing input randomness).

Even if these issues had proven to be too difficult to overcome or randomized bonus objectives didn't fit the game direction, having modular bonus objectives would still be an advantage.

Other hooks

Mutators are a special in that they have almost zero hooks in co-op maps themselves. Many mutators react to units, structures and resources, but that's automatic and map-independent. "Blizzard" mutator is one of few mutators that is affected by maps – namely the direction in which storms travel is different and hardcoded. Introducing more hooks in Co-op maps would make it possible to produce content that is more tightly woven into missions. The more hooks maps have, the more options there are.

Some possible hooks

• Attack wave spawns. Currently, attack waves are called in several different ways. "Heroes from the Storm" mutator which is tied to them should "just work" on all maps, but unfortunately it doesn't.
• Checkpoints marking players progress through the main mission (e.g., 25%, 50% and 75%).
• Areas of main bases and expansions (useful for example for environmental mutators).
• Areas of enemy bases.

Overall, this post is neither a critique nor a request for changes. Co-op has evolved spontaneously and is exploring relatively unknown sub-genre. It's interesting to learn what works, what doesn't and what could be improved. And while having modular bonus objectives would bring interesting options, it probably wouldn't be worth it now and development time would be better spent on something else.

Dehaka

There are many things to love about this fire breathing monster, and he deserves a post similar to which I did on Nova. I will go through his strengths and aspects I like about his gameplay and design. There are few things that I don't enjoy about Dehaka, and I will look at them afterwards. This is not a guide, but it might help readers understand his strengths better. The gameplay guide will be released on Teamliquid.net after the Fenix guide. Now, let's look at things that make him such great and unique commander.

Hero unit

• Strong fantasy
• In campaigns, Dehaka is small and fast. However, in co-op missions, Blizzard has fully embraced the Godzilla theme. Dehaka is big, slow, breathes fire, takes hits from anti-air weapons, and can devour even the biggest hybrids.
• Unique hero
• The leveling system for Dehaka is unique among co-op hero units, and draws inspiration from Warcraft 3. Further similarities to Warcraft 3 can be found in buildings that can uproot and fight, which is a feature of Night Elves. The ability of Primal Zerglings to mine resources resembles Undead Ghouls. However, this ability was removed from the game.
• Early in the action
• Dehaka appears on the map earlier than any other hero unit. This presents players with more options and choices. Will they slap rocks or wander onto the map to get that sweet, sweet essence? Finding good places to get essence early encourages players to learn maps in a way they never did before. I find this one of the biggest accomplishments of Dehaka's design. Here is a great example of Dehaka's early game with commentary.
• Devour 
• Devour rewards precision, speed and decision-making. Lower cooldown when targeting low health enemies is an elegant solution and makes the ability even more interesting. Casual players will be satisfied with devouring big hybrids, while better players will try to maximize the effectiveness of this ability by devouring targets with low health to keep the cooldown short. Different bonuses for various unit tags further increase the complexity of this ability.
• Psionic explosion might be the most rewarding effect of any ability in the co-op. It rewards positioning, mouse precision and decision-making. For the maximum effect, Dehaka has to put himself in danger in the center of enemy units. But the reward is very high, often resulting in clearing the whole wave. The fact that various enemy compositions have different amount of psionic units forces Dehaka to engage in different ways. This makes enemy unit compositions feel distinct, and Dehaka can't rely on Psionic explosion completely.
• Good respawn mechanic
• Losing Dehaka isn't as punishing as losing other hero units. Players don't have to wait to get back into the action, and instead the cost is more to the economy as a result of devouring Primal Drones. Nova's respawn mechanic is still a bit more forgiving, but Dehaka is also a lot more powerful.

High skill ceiling
• Challenging micro and multitasking
• Dehaka has plenty opportunities for micro-management and multitasking. In the early game, players will quickly gain control of Dehaka, and while they are microing him around the map, they have to manage their economy, expand and build dens for Primal Pack Leaders. 
• Later in the game, dens are built and economy is set up, but the nature of Dehaka, Primal Pack Leaders and his army often encourages players to split them apart. This is because of their individual strengths, Dehaka's Deep Tunnel and ability to summon Primal Pack Leaders anywhere on the map.
• Note that even (Greater) Primal Wurms have two active abilities. This is very similar to Nova, who has more abilities than almost any player will be able to use fully.
• Despite all micro-management and multitasking, Dehaka is accessible to players of any skill level. And all these options for an improvement make Dehaka very rewarding to learn and play.

• Powerful calldowns
• Dehaka's calldows enable him to push heavily fortified defenses extremely quickly. This lets him collect essence faster, which again enables him to be even more aggressive.
• Glevig is the first and the best calldown Dehaka has. His Incendiary Acid lets him demolish enemy defenses and attack waves. Glevig can attack several places on the map when using his Deep Tunnel ability. This opens a lot of potential and makes controlling Glevig very fun.
• Murvar, the Hollow Mother, isn't nearly as fun to control as Glevig, but deciding where to use her for the maximum effect is rewarding in its own way.

Army and economy

• Several viable unit compositions
• Most of his power lies in Dehaka himself and Primal Pack Leaders, and so players are free to support them with various unit compositions that fit their playstyle and current mission.
• While mass Primal Mutalisk is the most all around composition; Primal Guardians, Impalers and Primal Swarm Hosts are all good in many situations.
• Combined upgrades
• Combined upgrades, limited research for specific unit types, and cheap and universal unit production structures make mixing different unit types very easy. Also, switching between unit compositions is a lot more affordable.
• Simplified economy
• Dehaka doesn't need to make overlords; doesn't lose drones when building Extractors; has one production structure for all combat units; and upgrades are researched directly from Primal Hives. Even if his macro is not as reduced as for Nova, it lets players focus more on microing Dehaka.
• Accessible detection on the hero unit
• As with Nova, this makes playing Dehaka a lot more enjoyable. In comparison, getting detection for commanders like Kerrigan, Zagara or Vorazun can be awkward and at times frustrating.
• Good ability to recover from losses
• While Nova can struggle to recover from heavy losses, Dehaka does not have that issue. This comes again to the fact that most of his power lies in Dehaka himself and Primal Pack Leaders. All of them are locked only behind cooldowns, and are powerful enough to win brutal missions without the help of an army.

Issues

Now let's look at things I don't like as much. However, I understand design decisions behind some of them, and I can see why they were made. After all, weaknesses define a commander as much as his strengths.

• Static defenses cost supply.
• Units evolved through primal combat have lower supply than combined base units. While it's good for balance, it makes it little annoying to reach the maximum supply. You will max out, start primal combat, max out again and so on.
• Dakrun is neither very exciting nor very powerful calldown.
• I wish Dehaka had more options for his evolutions than available upgrade points. This way your fully leveled Dehaka would be different from the Dehaka of other players'. You could make your Dehaka unique to fit your playstyle, the map and enemy compositions. I hope this will be explored with a future commander.
• I don't usually complain about having too many things. But considering how his army plays only a minor role, a lot of attention has been paid to it. This is partly because primal zergs have been a part of campaigns. However, Dehaka has almost double combat units compared to Kerrigan, who is more focused on her army and without any calldowns. Then there is the evolve mechanic and random bonuses added through it. The ideas behind it are good, but I wish these mechanics were fully developed and used with another commander, where they could have a greater gameplay impact. Here is a nice video about design by subtraction.
• It's shame that the ability of Primal Zerglings to mine resources didn't make it through. They already have very little to do apart from breaking rocks, and as I mentioned, there are a lot of army units already. Then again, perhaps we will see this mechanic with another commander when it's more fleshed out.

Wrap up

As with Nova, Dehaka strikes a great balance between having low skill floor and high skill ceiling. More casual players can lean on Dehaka himself and strong Primal Pack Leaders, while more skilled players will find plenty of opportunities that reward micro and multitasking. The best parts about Dehaka are his theme and how differently he plays and encourages players to learn maps better.

Progression and Strategies

After the post about The popularity of Nova Terra, it has been pointed out to me that the idea of First Order Optimal Strategies can be helpful when looking at her design and the ease of transitioning between strategies. I think this is an interesting topic, even if not extremely important to the Co-op mode. First, let's see what these words even mean.

First Order Optimal Strategy (FOOS) is a strategy with the best power/effort ratio. These strategies require the lowest effort and skill, but provide relatively high power or other reward. They are often discovered early on.

True Optimal Strategy (TOS) is a strategy with the highest power regardless of the effort and skill required to execute it.

A simple example would be the usage of shotguns and sniper rifles in many first-person shooters. A shotgun doesn't require precise controls and is effective on low skill levels (using it is FOOS). On the other side, sniper rifle requires a lot of skill and new players will struggle with it. However, it can become the deadliest weapon in good hands (TOS). In older RPGs, warrior class was often FOOS, while mage TOS. But with the addition of complex abilities to warriors, the skill differences between classes were erased for the most part.

These concepts are not limited to only games, but they are applicable to any process or strategy optimization. The post on Reality Refracted describes well these ideas in relation to tabletop games, and is well worth reading. Also, Extra Credits did a very good video on this.

FOO strategies provide a good way to introduce new players into a game. Because they require low skill and effort to execute, players can feel powerful and useful early on. As players get more accustomed to the game, they will start to improve FOOS and look for different ways to play. Eventually, they might even discover TOS. The path between FOOS and TOS can be anything from trivial to very difficult.

In games, FOOS and TOS can relate to various parts of gameplay - deciding between weapons, classes, skills, playstyles or ways to micro and macro. Let's look at Alarak and apply this to his unit compositions (Figure 1). Alarak is particularly interesting because of his clearly defined FOOS and TOS, and how difficult is it for players to transition between them.
In Figure 1, the Baseline represents random unit composition and players experimenting. However, players will quickly find a style that works - FOOS - in this case a solid opening followed by Supplicants mixed together with Wrathwalker and other units. This style works relatively well, players keeps improving it, and it gets stronger as players progress (even Baseline is increasing). After some time, another strategy becomes available. In this case it's the Supplicant/Ascendant composition (TOS). There is a Power drop when players switch from their optimized FOOS to this different and unoptimized build. After an adjustment period, the power of TOS overtakes that of FOOS. On a side note, it's possible that actual TOS include "1qe2c" or "No cooldown Alarak", but I won't go into that.

Figure 1: Simplified progression - Alarak

There are several reasons why players have hard time transitioning into the Ascendant playstyle (transitioning from FOOS to TOS). In the Alarak's case, TOS becomes available late at level 12 (→ wider Gap ). At that point, FOOS is already optimized and players are reluctant to switch to a significantly more difficult and uncertain strategy (⟷ significant Power drop). That is accentuated by how different those compositions and their playstyles are. Another issue is that when Ascendants become available at level 8, they are rather underwhelming and players might try and discard them as under-performing units. Only after level 12, when Alarak gets the Power Overwhelming upgrade, Ascendants become truly spectacular (→ even wider Gap).

Even though this might look as a design issue, it's mainly a balance problem. If mech builds were a viable alternative to Ascendant TOS, players wouldn't get trapped in FOOS. Furthermore, Alarak's playstyle diversity would be improved.
For other commanders, TOS can be the same as FOOS, or the differences between them are very minimal - for example for Zagara and Nova. However, even with Nova, players can get stuck on slightly sub-optimal strategies, if they don't like using certain units. Also, certain TOS are possible since the start, but they can get progressively more viable. An example of this is building multiple Orbital Commands with Raynor, where higher upgrades and masteries let players to be even more greedy early on.

Did I stay on a sub-optimal strategy for too long?

Some TOS can be unintuitive, for example opening with Siege Tank/Hercules, massing Orbital Commands and Omega Networks, or using Toxic Nests offensively. And as with Alarak, several FOOS are without a clear path to TOS due to balance issues - Fenix's ground compositions, Artanis' Robotic compositions, etc.

The difficulty of reaching TOS has to be set right. Having no good way to transition between FOOS and TOS can mean that players will never progress from FOOS. On the other side, if the path between FOOS and TOS is too straightforward or they are the same, the gameplay could feel more boring and shallow. A good middle ground can be a FOOS that easily transitions to the Second Order Optimal Strategy with strenght close to that of TOS. TOS then requires further experimenting and discovery.

How players can be led to TOS: Intuitive use of units Preview videos and graphics Good progression unlocks Achievements Quests (daily) Players learn from watching allies

What to avoid: Unintuitive use of units Too high Power drop Too wide Gaps between strategies Introducing units without their crucial unlocks Having objectively bad units available Completely different FOOS and TOS without an intermediary strategy

Wrap up

I don't think there is a big problem in transitioning between FOOS and TOS in the Co-op mode. In this mode, players can progress to higher difficulties at their own pace, as opposed to some other games, where the inability to adapt new strategies and skills can leave players in very poor situations, wasting their time or requiring backtracking. I prefer if the game doesn't always hold your hand, and there is a place for exploration.

In Co-op, dead ends created by underpowered compositions pose as a bigger problem. However, the idea of FOOS and TOS should be kept in mind when deciding on viable unit compositions for a new commanders, designing abilities, progression unlocks, achievements and creating preview screenshots and videos.