- #Slay the spire tier list update
- #Slay the spire tier list plus
So the format used for Boss Relics didn't work.This was affecting other pages as well but since only a small fraction of runs for the other 3 classes were processed by the new parser these errors were not noticable for them. Fixed a bug with Watcher Card list - "Upgrade" and some other binary properties were recorded as 0/1 by the old parser but I changed that over to NULL/1 in the new version (that saves some disk space when you're dealing with millions of zeroes) and those NULLs got skipped over completely when sorting data into a table to display.how often data is recalculated) from 24 to 48 hours for card list.
#Slay the spire tier list plus
Improved uncached loading of Card List and Relic List (for now only boss relics) and added a fallback-to-cached plus an error message in case retrieving fresh data fails, which has been happening from time to time. So I might revisit them eventually but for now the pages are up so it's at least not a step back from Spirelogs v1. #Slay the spire tier list update
I don't plan on redoing them anytime soon because identifying all the different choices already recorded and converting them into numeric IDs would be a lot of tedious work and then I'd have to update the parser too to deal with the new format. So event pages were rebuilt, but not exactly improved except maybe in terms of load time. Events pages: These are kind of a mess because event choices are the only type of data that has always been recorded verbatim from run files and not converted into a numeric ID and I didn't standardize them for Spirelogs 2.0.
This is probably the final iteration of "Boss Relic" pages until the official launch.
Boss Relics pages: Added more detail - percentage of runs which reach the next act, displaying rank by popularity and floor reached on relic detail pages. : Boss Relics improvements, Event List and some backend stuff Added Encounter Types by Floor - just a random page with some information I wanted to look at. It looked misleading to see a bunch of green text on a bad relic (because all other relics were doing better than it) and vice versa with a bunch of red text on a good relic. Removed color-coding of "Floor Reached" on Boss Relic pages. But to get an accurate "expected number of Cloak and Daggers" I need to somehow adjust expected values by floor because an average run with Cloak and Dagger and ANY RELIC will reach a higher average floor and thus contain more Cloak and Daggers than just an average run with Cloak and Dagger. If I try to measure number of Cloak and Daggers in runs with Shuriken it should obviously be higher than average. 
That didn't really work well because it needs some kind of baseline and calculating it seems trickier than expected.
Non-boss relic pages: I started adding some data here and tried measuring performance and popularity of other relics and cards relative to the target relic. There are some interesting mismatches between Distance and Purchases - for example Mercury Hourglass appears to be criminally underrated. Distance shows the relative strength of the relic (yes, some relics are obtained earlier/later than others so "Floor Reached" doesn't work) and percentage of relics from shop is a way to measure how each relic is valued by players as Shop is the only non-random source for relics. All the new metrics seem to make sense here. Added Relic List pages for non-boss relics. This is just a step towards calculating baseline values which I was missing in the previous update, but I don't have a precise idea of where it will go yet. Each data point is "How did decks that pick Go for the Eyes on Floor X do compared to all decks that reached Floor X". Added a "Floor Reached" graph to class main pages (these will be filled out some day) and graphs to card detail pages comparing the performance of each card to that class's average on a floor-by-floor basis. The plan is still to complete everything on the Roadmap and replace the old site. But we are probably looking at weeks here, not months. Lastly a note on the overall progress of Spirelogs 2.0 - I slowed down the pace of updates recently because I've been spending time on other things and it will likely stay slow for the near future. Overall this shouldn't impact the data much since only a small percentage of the total run pool will be incorrectly classified as "NO BONUS". I added a check specifically to only identify and update the missing Neow data so any reuploaded runs should process much faster than normal. Unfortunately the only way to fix it is to reupload runs, which I'm planning to do. 
While doing that I discovered a bug with the new parser not recording Neow's Blessing data since the start of Spirelogs 2.0. This is pretty much the same page as in Spirelogs 1.0 except reorganized to fit new ascension brackets.
0 kommentar(er)
