Update 1.2.0

Since the last update four months ago, we've been working on a wide range of features and improvements for CubeCobra. In the last update I talked about opening up avenues for new projects to benefit from CubeCobra with new integration points, and today I'm really excited to announce two new integrations, in addition to a couple new large features.

The first integration is with a new project, still in open beta, called Snapdrafter. Snapdrafter is currently only available as an android app and if you'd like to see it an iOS app, I'd encourage you to support the app here. Snapdrafter is an app that lets you take photos of your drafted deck, and save them to your phone as a text decklist file. From there, you can click the CubeCobra icon to import scanned decks into CubeCobra. I believe that creating an easy way to collect data for in-person drafts will make it feasible to do data analysis for in-person drafts of a cube. These imported decks won't just be a playtest draft object you're familiar with, we have added a new cube section called "Records" to track actual drafts of your cubes. This section is meant to be a data store, a place where data can be input or imported, in order to perform analysis on the performance of cards or strategies. It's still very basic, so if there is something you'd like to see in this section, please let us know.

The second integration is with a project you may have heard of before, Commander Spellbook. Commander Spellbook is the most comprehensive database for combos within magic. We've added a new analysis tab to cubes that allows you to see the combos that are available in your cube, with detailed information about each combo. If you feel a combo is missing, you can submit new combos here.

If you regularly follow development of the site, you may know that I retrain the ML system roughly once a year. I was unable to retrain this year due to our dataset outscaling the training infrastructure we had set up. This has been a pain point for users as newly printed cards that were outside of the training data were effectively invisible to the model at the time. We implemented some mitigation measures, so cards outside the training dataset are matched to similar cards. This has been an effective stopgap for now, and will help keep the system working well in between re-trainings. Fortunately we have recently had help from a contributor who refactored our training scripts, to be able to handle the scale of our dataset for many years going forward. I'm currently working on rebuilding this dataset and retraining the model, and I hope to have that done in the next month or two. There have been some new features that interact with the ML system, such as being able to view draftbot scores during playtest drafts, and also being able to view these scores in the pick-by-pick breakdown.

New Features

• Added records tab, for tracking of practical draft data

• Added combos analysis tab, using commander spellbook data

• Added bot scores for pick-by-pick analysis and during playtest draft

• Performance optimizations on draftbot calls, making playtest drafts smoother

• Allow users to set default preferences in card searches

• Added word count sort

• Updated custom draft formats to fill restrictive slots before non-restrictive slots

• Added mana breakdown for deck stats

• Added filter for card notes

• Added Mana Pool purchase links

• Added purchase links for entire cubes and selected group of cards in a cube

• Added user preferences for grid tightness

• Add User auto blog preference, which determines whether "Create blog post" is checked or not by default (and defaults to false)

• Async load cube or user followers when the modal is opened, page results for page load performance on popular cubes

• Added spam mitigation measure warnings for improved UX

• Allow custom back of card images to make the Transform button appear

• Added power/toughness comparison filters

• Support tag exact match searching

• Added inner scrolling for large modals, have some buttons always visible

• Return draft notifications to the solo, within CubeCobra bot drafts

• Add notifications for Draftmancer draft imports

• Performance improvements with fetching cube recommendations

Bug Fixes

• Increased card tag color contrast to improve visibility

• Fixed deck deletion

• Fix duplicate card removals that also remove unintended cards

• Fixed collector number sort

• Allow card search with apostrophes

• Card prices for 'Etched' finishes now show the correct price

• Fixed failing drafts when basic lands are not lands (you know who you are)

• Fix not using distinct card names for searches correctly

• Fixed saving of new packages to correct format

• Use short id if it exists in "Your Cubes" dropdown

• Fix card name sorting on playtest data

• Fix invalid dates in date added sorting

• Fix vote package count being stored in the wrong attribute, not aligned with the vote count index.

• Fix error where frontend causes a 404 on redirect because the short id changed.

• Fix Playtest data CSV containing card ID instead of name, so different than the UI

• Updated Tile component to use anchor instead of onClick so browser's normal ctrl/cmd click behavior opens the cube in a new tab.

• Fix delete deck modal actions redirecting to the deck

• Fix bug that resulted in Art Cards appearing before the actual cards

• Prevent last step of a pack in custom draft format from being a "pass", which stalls the draft

• Fixed trashing cards during playtest drafts

• Fix matched cards in CSV upload not tracked when user is asked to correct misidentified cards

• Fix blog post redirection on edit/delete if the cube is deleted

• Fixed searching while not logged in

• Fix color identity filter case sensitivity

• Default card notes to empty string instead of null to fix filtering error

• Fix drafts not finishing or ending in a weird state, related to pending picks and failing bot predictions

Technical Changes

• Added spam mitigation measures

• Split webpack into common and page-specific chunks, reducing page sizes

• Started back-end typescript migration, migrated many files

• Response time logging to help diagnose performance issues

• Significantly increased unit test coverage