Like many competitive online games, Prismata has a variety of unlockable in-game collectible cosmetic items: unit skins, equippable emotes, and more. However, Prismata differs from many other games in that these items are not obtained randomly through pack-opening. Instead, players complete minigames in the Armory and Black Lab, where they have a choice of a number of different prizes. A consequence of this design is that by examining the choices that players make in these minigames, we can actually determine the most and least desirable cosmetic items.
Want to know what the most popular Prismata skins are? Read on…
Collectibles in Prismata
In Prismata, players earn collectibles by first unlocking access to special minigames. By completing daily quests or collecting wins on a Ranked Play ticket, players can obtain special items known as Power Cores, Omnipower Cores, and Black Cards.
By spending these special items, players can access two card-flipping minigames known as the Armory and the Black Lab. Succeeding in these minigames will reward lucky players with fresh new cosmetics.
In the armory situation above, the player has scored highly by flipping over a number of combos. After the conclusion of the card-flipping phase, the player must now spend the points on cosmetics, and can only choose from the available selection at the top of the screen (which is randomly generated for each Armory session). Specifically, the player above has an additional 120 points to spend, and there are four “level 2” rarity skins available, each at a cost of 100 points. The player can thus choose only one of these four rewards (unless they wish to “Flip Another Card” and see if they can increase their winnings).
Choice Situations
In the situation above, the choice of one reward over the other three exhibits a player preference for that specific cosmetic item. By examining choice situations like these and performing some analysis on them, we can rank all of the Prismata collectibles by how often they are chosen. Of course, we don’t get any data when the player loses in the Armory, or when they win sufficiently many points to unlock everything. For our experiments, we only considered situations where a player selected something of a given rarity whilst not selecting another item of the same rarity. (Unfortunately, this implies that we don’t have this type of preference data for Level 4 items, because they only spawn one-at-a-time in the Armory).
A similar choice situation shows up in the early stages of Prismata’s Black Lab minigame:
In the high-stakes Black Lab minigame, players actually choose which prize they might win before beginning the game itself. This results in more datapoints than we obtain for Armory sessions, since a player gets to express their preference every time they visit the Black Lab, win or lose.
Crunching the Data
Our exploration of this data was motivated by a number of things. After introducing Infusions earlier this summer, we wanted to look at the Armory and Black Lab data to determine whether or not players liked receiving them as prizes. We also wanted to determine whether Black Fragments—a special unlockable that can be chosen in the Armory to accelerate the acquisition of additional Black Cards—were favoured over Infusions.
As it turns out, obtaining the desirability of every cosmetic in Prismata wasn’t really any more difficult. So we did it!
Initially, we took a naive approach and simply counted the fraction of the time that each cosmetic item was chosen. However, there are a number of limitations to this approach. We add new items all the time, so newer items could be underrepresented, or might find themselves picked more often because they compete with the leftover items that are less desired by players who have built up a large collection.
Instead, we computed the Elo rating of each item, treating every situation where Item 1 is picked over Item 2 as a win for item 1. The most-picked items thus ended up with the highest ratings. We tried a number of different methods and many trials using randomized orderings, varying k-factors, and other algorithmic details. However, the results were pretty consistent: some skins were clearly preferred to others.
The Results…
We initialized all the Elo ratings at 2000 before obtaining the results below.
The top 10 level 1 emotes (and their Elo ratings)
10) Where did all my stuff go? 2157
9) Thank you for a very enjoyable game. 2162
8) Excellent game. 2167
7) !!!! 2172
6) That was a colossal mistake. 2180
5) Mistakes were made. 2184
4) You can never have too many Drones. 2185
3) Math is hard. 2191
2) We meet again. 2200
1) If only I had a more appropriate emote for this situation… 2206
The top skins at each level
Level 2) Cerulean Barrier
Level 3) Ship’s Wheel Iso Kronus
Level 4) UNKNOWN (no preference data for level 4 collectibles)
Level 5) Dark Matter Mahar Rectifier
Level 6) Space Hotel Zemora Voidbringer
Level 7) Eviscerator Xeno Guardian
Full data is here.
Black Fragments vs Infusions
As it turns out, there isn’t a huge amount of difference between Black Fragments and Infusions in terms of their selection frequencies in the Armory. Here are their respective Elo ratings:
Level 1) 1 Infusion 2043
Level 1) 1/30 Black Card 2074
Level 2) 5 Infusions 1839
Level 2) 1/6 Black Card 1817
Level 3) 10 Infusions 1897
Level 3) 1/3 Black Card 1920
Somewhat interestingly, Infusions win at Level 2 but lose at Levels 1 and 3. We’re not really sure why. The overall rating of both Infusions and Black Fragments is higher at Level 1 because some players just don’t like emotes very much, and skins are not available as Level 1 prizes.
To double-check our results above, we also looked at the head-to-head matches involving only Black Fragments and Infusions: when players can choose exactly one of the two, which one do they pick? The answers are pretty close to 50-50 and consistent with the Elo differences we see above (for example, a 30-point Elo difference represents around a 55-45 split in “win rate”, which is approximately what we see at Level 1.)
Isn’t Data Amazing?
We’re not sure if this is the first time that Elo ratings have been used to evaluate the desirability of collectibles (it probably isn’t, but we’ve never heard of anyone else doing it). That said, these ratings don’t really paint a full picture of player desires. Some choices are highly polarized, like Infusions vs Black Fragments (most players are relatively consistent in their choice if they make it multiple times). Some players still tell us they don’t like Infusions at all.
Nevertheless, we’re very happy to give players a bit of choice in their rewards, rather than simply give them a random pack of collectibles every now and then. After all, it results in a fantastic dataset.