Skip to main content

Pretty Good Hat

Tag: destiny

Screenshot displaying rows of armor items and their stat values

Screenshot showing a scatterplot of stat values for a selected armor configuration

I’ve had a nice afternoon working on my hobby R/Shiny project, a loadout finder for the game Destiny 2. These improvements make it a lot more flexible and informative: It can now optionally include armor that would otherwise be filtered out of configurations by the minimum stat threshold, and it will show current mods used in displayed loadouts.

ArmoreR: A Destiny 2 Loadout Finder

Since I’m talking Destiny projects lately, here’s a another open invite to Destiny-curious micro.friends who may want to try it out! The new expansion is very good and it’s a great time to jump in. Happy to be a space magic shooty guide to anybody who needs a friendly way into the game!

A screenshot showing armor statistics for loadouts in Destiny 2

This revision of ArmoreR is really turning into something I’m happy with and proud of. I’ve learned so much since the very first iteration of this about a year ago. It’s really fun and rewarding to come back to it with a bunch of new expertise and make it much better in all ways, including a new approach to dealing with such a large amount of information.

screenshot showing a set of colorful scatterplots displaying a variety of armor types and their stats

I spent a bunch of the weekend beginning a redesign of my ArmoreR project, which aims to be a Destiny 2 armor stats profiler built in R and Shiny. A year on from when I began it, I have a much better understanding of how a Shiny app works, and have also incorporated a proper, working oauth2 workflow into this revision (incorporating the things I figured out for my power level tracker). It’s really, really satisfying to be rebuilding it with all the things I’ve learned. I think the application is going to be so much better and less complicated than the first iteration. I still have a ways to go, and am happy with just how much I have transformed it with a year of learning and practice on other things.

a screenshot of a web application showing a table of statistics for a game character

Last fall I wrote a bit about a Destiny 2 power level tracking tool I built using R. I’ve now converted it to a full-on Shiny app and solved some issues with the oauth2 flow that stumped me in my intermittent tinkering with it. I’m super satisfied to have been able to get this to work! Now that I have the authentication process figured out, I’m eager to also convert my armor profiling tool to use it. Look out!

You can check it out here: traveleR

Today I made improvements to some R code in my Destiny 2 hobby-coding-verse after learning how to much more cleanly deal with nested lists. I had previously used a solution using map() to apply a selector to each item in the list, but this was clunky, hard to remember, and became really hard to read with several levels of a nested list.

The far better solution is the unnest_auto function from {tidyr}, which I came upon when tinkering with the last.fm API data recently. Once I understood how it works, it’s so easy and satisfying! The key is to first make a named tibble.

> tibble(my_tibble = instanced)
# A tibble: 906 × 1
   my_tibble        
   <named list>     
 1 <named list [10]>
 2 <named list [10]>
 3 <named list [10]>
 4 <named list [10]>
 5 <named list [12]>
 6 <named list [10]>
 7 <named list [10]>
 8 <named list [10]>
 9 <named list [9]> 
10 <named list [9]> 
# … with 896 more rows

That nice tibble can be operated on by unnest_auto():

> tibble(my_tibble = instanced) %>% unnest_auto(my_tibble) %>%
select(itemLevel, breakerType)
Using `unnest_wider(my_tibble)`; elements have 8 names in common
# A tibble: 906 × 2
   itemLevel breakerType
       <int>       <int>
 1       132          NA
 2       133          NA
 3       133          NA
 4       132          NA
 5       133           3
 6       132          NA
 7       133          NA
 8       133          NA
 9         0          NA
10         0          NA
# … with 896 more rows

(I selected just a couple of columns for readability there; if you don’t do that, you’ll receive all fields at the current list level, including additional nested lists if they exist). After figuring this out, I realized that I also needed to keep the names of each list element, because they constitute a unique ID for the element returned from the API query, and I banged my head a bit on trying to do that as a part of the unnest operation, before I backed up, recentered on the outcome I wanted to produce, and realized I could do it really cleanly using mutate()! The final code looks like this:

> tibble(my_tibble = instanced) %>% unnest_auto(my_tibble) %>% 
select(itemLevel, breakerType) %>% mutate(id = names(instanced))
Using `unnest_wider(my_tibble)`; elements have 8 names in common
# A tibble: 906 × 3
   itemLevel breakerType id                 
       <int>       <int> <chr>              
 1       132          NA 6917529338105913753
 2       133          NA 6917529550016812142
 3       133          NA 6917529281178546429
 4       132          NA 6917529231127188610
 5       133           3 6917529301504642848
 6       132          NA 6917529234625200021
 7       133          NA 6917529193313832065
 8       133          NA 6917529182955737017
 9         0          NA 6917529417667489181
10         0          NA 6917529490520758715
# … with 896 more rows

The D2 API returns a ton of nested lists, so this simplified, accessible and effective tool is 100% getting a featured spot in my toolbox.

Destiny 2 Power leveling tool in R

The latest piece of my homebrew Destiny 2 toolset is Travelr, a power leveling tracking tool built with R and using a much simpler process than most of my prior nonsense. (Aside: I'm still using ArmoreR, but it has some key limitations that I haven't spent much time with, including a problem incorporating the authentication flow more fully into Shiny so that I could make it an app that others could use in a more friction-free way.)

The power climb is part of the game with each new Destiny season. A lot of this process is luck: A player can only get so many pinnacle drops per week and there's no guarantee that a given drop won't be in a duplicate category. But, some drops can be targeted to broad categories -- armor versus a weapon, for example -- and there is some efficiency across characters if you can strategically raise one category to increase the base level for all characters. 

I was inspired by a fantastic tool written by a member of the small Destiny 2 community I'm in, that identifies categories to target (or, try to target, anyway; a lot of it is still frustrating random luck, sometimes), and decided to try building my own version. It doesn't do anything unique, but it works and it's wonderfully satisfying to step through an R markdown notebook that authorizes me to the D2 API, retrieves all my equipment statistics, calculates my maximum power level and indicates which, if any, categories should be targeted.

For example, here's a screenshot from the beginning of the season, where every single equipment type would increase my total maximum power level:

A screenshot of a table of data showing lots of equipment categories that would increase power level to the next integer

... and later in the season:

Travelr screenshot showing a table of armor and equipment values, highlighting values that would increase the total maximum power level to the next highest integer

It's been great fun to have this small tool with me during the season's play. It's currently dependent on another small tool that grabs the equipment "manifest" file, and I'll be publishing the whole notebook to github when I finish incorporating that into the repository. 

Today I dusted off my latest video game hobby project, ArmoreR, my Destiny 2 armor stats profiler, and finished some improvements that have been on my list. I haven’t been motivated to dig back into it for a while partly because I just … didn’t feel like it … and partly because I had hit a point of complexity with it that made it hard to approach again. But I had a really nice time catching up with just how it works and then gradually identifying and implementing the improvements I wanted to make. It’s a good lesson that is sometimes hard for me to remember, that small but concrete steps add up.

Destiny nerd warning, most people can probably stop reading now: This version adds something really useful to loadout building: class items! I had to think on how to do this for a while, because most class items don’t add much to the stat value of an armor build, and each added class item combinatorially increases the number of armor set combinations that have to be measured. I settled on including only masterworked class items because they add enough stat points to tip one or more categories into the next tier. Still, for characters that have a lot of armor to choose from, doubling (or more) the total number of builds does add to the time that the tool requires. I added some ways to optionally include/exclude more armor to try to make this processing time more bearable. I also found a few improvements to make and fixed a number of bugs along the way, and will plan to return for a round of efficiency-finding later.

The tool is at a neat, workable point right now, where it’s actually useful for finding armor builds and, really critically, the armor pieces that work across multiple builds of interest. I’ve learned a ton working on it, and am pretty proud of it. There’s more documentation over at the application and if you happen to be into Destiny I’d love to hear what you think.