Balancing a tactics game like Mech Armada with lots of parts and monsters is not an easy task. There are some rules of thumb you can use during development to try to get things right but, at the end of the day, the proof is in the pudding. You need to look at how people are playing the game, and whether or not it matches your design and expectations. That’s what we’re doing today.

A couple of notes about gathering data. Yes, Mech Armada collects gameplay data. No, I’m not going to be emailing you spam: I don’t have your email address, your name or anything personal.

Second, whenever you look at a game’s data things are always all over the place. It’s naïve to think that players behave in a consistent manner and as you change the rules the results change accordingly in a clean fashion. Nothing could be further from the truth. Get enough players and you will literally see every possible outcome. However, what’s important is how frequently something happens. Let’s say a weapon is overpowered. That doesn’t mean that players will win 100% of the time with it, but it will mean that they’re significantly more likely to win than with a different weapon. Like 10%-20% more likely.

One final thing. If you haven’t played the game, I’m going to discuss some of the fights (including boss fights) and parts that you need to unlock. So if you don’t want spoilers, come back after playing a bit more.

With that out of the way, let’s look at some data.

Winning Battles

The most important number is how often are players winning battles. Win every battle and you’ll reach the end of the game (beating the Demon). Of course this is tightly connected to the challenge level.

Overall, 4.5% of runs end in victory at Challenge 0. If you look only at people who’ve unlocked all the parts (you’d expect those people have been playing longer and have more experience) the number goes up to 13%.

Now the highest Challenge level is 12. The game is a LOT harder. At the same time, only players who are really good at the game are typically going to try the hardest mode. The result? 10% of Challenge 12 runs end in victory. That’s higher than the base for Challenge 0 but less than the number for experienced players.

So the next question is, where are people actually dying? While there is a small chance of losing every single battle, for the most part the bosses are causing the most trouble. Here are the chances of surviving a boss fight at Challenge 0:

• 76% vs Goliath
• 81% vs Scourge Leader
• 50% vs Spider Queen
• 90% vs Sandworm
• 68% vs Demon

Clearly, the Spider Queen battle is currently a problem. Ideally the win percentage should be closer to the 80% the other battles have. The Demon being a bit lower is ok – he’s the final boss and he’s supposed to be tough. On the other hand, the Sandworm might be a bit too easy. With these numbers I can adjust the game. But first, let’s verify the numbers match what we see at other difficulty levels. Here’s Challenge 12:

• 85% vs Goliath
• 92% vs Scourge Leader
• 58% vs Spider Queen
• 95% vs Sandworm
• 90% vs Demon

Overall slightly better but quite similar percentages. This is to be expected because the global win rate is higher as we saw earlier. Interesting to see that the Demon is not as difficult. It’s possible by that point players really have a good grasp on the game…or I simply don’t have enough data and a few people got lucky.

General Stats

Another piece of data that’s important is how much of “something” are players doing. This helps me make sure the game length is adequate and there is enough content.

A common stat for turn-based games is how many turns it takes to win a battle and, conversely, how long it takes to play. Typically, a battle in Mech Armada is between 3 and 6 turns, while it can last anywhere from 30 seconds up to 5 minutes or more. These numbers fall well within my design expectations.

A key element of playing Mech Armada is buying mech parts to make your team stronger. So, how many and what kind of parts are people buying in a run? Overall, here are the results:

• Transport: 1-3
• Body: 1-3
• Weapon: 1-5

It makes sense more weapons are being bought, as in a mech you can equip 2 or 3 different weapons. At the same time, these numbers are a bit misleading, because they include people who died early in their run, so they didn’t have a chance to get many parts. If we look at people who got to the last environment, the numbers go up:

• Transport: 1-3
• Body: 2-4
• Weapon: 4-8

Weapons stick out even more as what people need the most. Transport is less exciting, you start with some pretty decent options, so there is less pressure to get more. Bodies provide some benefit, so they sit in between.

Pick Rates

Because Mech Armada presents players with a choice when they buy parts or earn expansions, I can track which ones are most popular. I prefer looking at players who’ve unlocked all parts, as they have all the options and know the game better. Here are the most and least popular parts and expansions with the percentage of times they were picked when offered:

Transport

Body

Weapon

Expansion

Ideally, you want the pick rate for every part to be 50%. That would mean they are all perfectly balanced. That’s never going to happen, but I can definitely look at some outliers and figure out why people don’t like them (or like them too much!)

Some quick conclusions that can be drawn from these results:

• Players really like weapons with lots of damage, range and area-of-effect. And they hate anything support.
• They like big legs and they cannot lie.
• Some expansions need some love, but damage, armor and health are always popular.
• The Chain Reaction body is super popular, so a nerf is likely coming.
• Conversely, nobody like the Kamikaze. I get it. It’s hard to use, but can be game-breaking if used well.

Which brings me to another point. There’s another angle to the pick rates. Players pick certain parts because they think they have a better chance of winning with that part but, is this actually true? Let’s find out.

Win Rates

This is the trickiest number to calculate. A lot is happening in every battle and every run, and it’s not obvious how to link victory (or defeat) to a specific part. The way I extracted the data is not perfect, but will have to do for our purposes: I look at all the runs where a player picked a part and figure out how many of those ended in victory. This ignores progress made with a part that didn’t go all the way to the end, but because parts drop equally throughout the run it should be close enough. Also, I’m once again looking at players that have all parts unlocked, otherwise parts that you unlock later have better stats not because they’re better, but because only good players have access to them.

Because I simply don’t have enough data to give exact percentages for win rates, we’ll just look at which parts do better in general. Also, I’m excluding the initial parts from this list.

Transport

Already we see something interesting here. Buggy and Trike are not very popular, but do well. Walker is the opposite. Tracks and Mini Spider are both not popular and not successful, so they could use some tuning.

Body

Again, here some bodies are well understood by players, like Eagle Eye, Ranger, Commando and Maximus (all very popular). Engineer is under-appreciated (pick rate is 35%).

At the bottom, Off-roader, Turbo and Bullseye deserve a redesign, as they don’t seem to be serving players well right now.

Weapon

So much to unpack here. It seems that players want the damage-dealing weapons, but the best predictor of success is picking support parts. I’ll need to think about this. There could be an imbalance in the game where support is sort of required to succeed but it’s not attractive enough; I believe this is pretty common for this type of game. But it could also be that players who like to use support weapons are generally more skilled at the game and therefore more likely to succeed – this is always the challenge with looking at data.

Oh, and the Neutralizer is a disaster, so that needs to get fixed.

Expansion

This one blows my mind a little. Salvage does really well, but I wonder if I just don’t have enough data; honestly, it doesn’t make a lot of sense. Accuracy and Protector perform as advertised, while the health-based expansions like Restore and Pristine tend to disappoint (probably because you need a specific team composition to engage a health economy).

Lastly, Initiative is a surprise. I keep hearing how game-breaking-overpowered it is, but it seems to not perform nearly as well in practice.

Data Informed

There’s a lot of talk of being “data driven” in the games industry. I prefer the term “data informed”. The difference is that all of these numbers are just one consideration. I look at player feedback from Discord and Steam, videos on YouTube and Twitch, conversations on Twitter and my own personal experience playing the game.

And then I decide to make adjustments.

And then, the whole process starts all over again. Expect a number of tweaks based on these results. Hopefully they’ll make the game more balanced and, as a result, more fair and enjoyable.