Adds track repair time to the Tankopedia survivability stats, matching the value the game lists.
- Added a required `repair_time` to track definitions, filled from each chassis's hysteresis health divided by its health regen per second
- Applied the toolbox, the repair crew skill and the protective kit as multiplicative repair speed bonuses
- Added a Track repair time row under fire chance; it appears once the definitions are regenerated on the next build
Fixes#308
Shows the chance a shot penetrates in the dynamic armor shot card, right-aligned on the shot's result line.
- Penetration varies by ±5% (the same randomisation the armor highlighter shader uses), so the chance is the share of that range where the shot's final outcome is a penetration
- Re-runs the existing shot calculation at different penetration values, so spaced armor, HEAT gap losses and ricochets are taken into account
- For ricochets the chance is shown on the ricocheted shot's line; HE splashes don't count as penetrations
- Adds the `shot_card.stats.penetration_chance` string
Fixes#342
Lets people on limited data stop the Tankopedia from downloading 3D models automatically.
- Added an Automatically load 3D models switch to the settings page, on by default
- When off, the tank page hero shows a Load 3D model button where the loading bar sits
- Held back every model download on the tank page until requested, including the dynamic armor switcher and the gun flexibility visualizer
- Wrapped the tank sandbox in its own suspense boundary so the loading bar shows real progress instead of the whole page waiting on the model
- Measured loading progress from when each loading bar appears so it always starts empty
- Added the English strings; other languages fall back until translated
Fixes#311
Prepares BlitzKit for tier XI tanks ahead of their release. Tier X was hardcoded as the maximum in the tank search filters, the guesser and the tech tree, and had no roman numeral beyond X.
- Added `XI` to `TIER_ROMAN_NUMERALS`
- Added an awaitable tier list derived from the highest tier in the tank definitions, so nothing changes until tier XI tanks exist
- Used it for the tank search tier filters and the guesser tier picker and defaults
- Made the tech tree check for successors instead of assuming tier X is the last tier
Fixes#302
Tanks sharing the same tier, nation, class and type were shown in an arbitrary order that changed after selecting a tank. The meta sort had no tiebreaker, and the filtered list was sorted and reversed in place, so each re-sort started from the previous reversed order and flipped tied tanks around.
- Added tank ID as the first sort key in `metaSortTank` so tied tanks always come out in the same order
- Sorted a copy of the tank search results, pre-sorted by ID, instead of the cached list
- Added an `exclude` option to tank search and used it in Compare so tanks already being compared can't be added again, including through select all or by swapping another slot
Fixes#313
The ID discovery script exited about a second after starting. Since the switch to `atMaxRate`, neither the discovery loop nor the request loop was awaited, and `atMaxRate` didn't await its callback, so the script went straight to `save()`, committed and exited while overlapping duplicate requests failed in the background.
- Awaited both `atMaxRate` loops and the callback inside `atMaxRate`
- Moved the ID range exhaustion check out of the retry loop so retries can't remove extra regions
- Treated network and JSON errors as failed API calls so a long run retries instead of crashing without saving
Fixes#306
Contact shadows under the tank disappeared when the camera orbited low enough to look up at it. The shadow plane and the ground disc are both transparent and centred on the origin, so three.js sorts them by camera depth. Looking upward flips that order, and because the shadow doesn't write depth, the ground disc was drawn over it.
- Gave `ContactShadows` a `renderOrder` of 1 so it always renders after the ground disc
Fixes#299
On some browsers a thin sliver of the Insanity banner showed through along the rounded edges of the equalizer toggle, most visibly on the right where the gradient is opaque. The banner and the gradient were painted on two separately clipped elements, so their anti-aliased edges didn't line up.
- Moved the gradient onto the outer box as the top layer of its `backgroundImage`, above the banner
- Removed the separate gradient layer from the inner flex
Fixes#297
The fade flag images are 2:1 but the flag only fills the left part, with the rest transparent. That made `cover` + `center` in the tech tree node crop and offset the visible flag.
- Trim the transparent border from the fade flag in the `/flags/fade/[id].webp` route using `sharp`
- Revert the tech tree node back to `cover` + `center`, since the image is now just the flag
The filters on the playlist page stretched into one long strip beside the tank pool, squeezing the pool into the remaining space.
**Changes**
- Place the filters and the generate button in a fixed-width column (`22rem`) on medium screens and up so the filter buttons wrap compactly
- Stretch the generate button to the column width
- Let the tank pool take the remaining width
- Tighten the gap between the two columns
- Keep the existing stacked layout on small screens
Dragging a tank's track scrolled its roughness/metalness texture at twice the speed of the base color and normal textures, so highlights drifted off the track links.
**Cause**
- `translateTexture` added the scroll offset to `roughnessMap` and `metalnessMap` separately
- glTF has a single metallic-roughness texture, and the loader assigns that same texture object to both slots, so it received the offset twice
**Changes**
- Collect the track's maps in a `Set` before scrolling so a texture shared between slots is offset exactly once
Users were no longer sent to their preferred locale. `LocaleRedirector` was emptied in `115fa5aca` after its previous versions caused repeated redirects, and nothing replaced it.
**Cause**
- The earlier logic compared the best locale (a string) against the current locale entry (an object), so the check was always true on non-default locale pages
- That made the redirect fire repeatedly, which led to the logic being removed entirely
**Changes**
- Derive the current locale from the URL, treating the default locale as unprefixed to match how pages are generated
- Redirect only when the best locale differs from the current locale, so it stops after one hop
- Redirect users with a saved `preferred-locale` from any page
- Redirect users without a saved preference from default-locale URLs only, using their browser language, so shared locale links are respected
- Preserve the rest of the path when swapping the locale prefix
- Skip `/api` paths
The mixer's hull/turret/gun row is about 440px wide at its fixed sizes and is centered with `left: 50%` and `translateX(-50%)`, so on phones narrower than that it overflowed past both edges.
Use smaller button sizes on phones and scale the tank images with the viewport (`clamp(2rem, 11vw, 4em)`), with smaller separators and gaps. From 520px up the layout is unchanged.
The static armor panel hid itself with a fixed `right: -6rem`. In languages with longer labels (e.g. Spanish) the panel is wider than that, so part of it stayed visible at the edge.
Anchor the panel at `right: 0` and slide it out with a transform based on its own width (`translateX(calc(100% + var(--space-4)))`) when static armor isn't active, so it always clears the edge regardless of label length.
The "Help translate BlitzKit" link on the home page had only a bottom margin, so it sat flush against the hero above while the tool grid below added extra space.
Add a top margin so the space above matches the space below, leaving the bottom margin unchanged.
Both buttons in the testing and unlocked filter pairs started on, so showing only one kind meant turning the other kind off.
Start all four filters off, treat a pair with neither button on as showing everything, and make the buttons in each pair mutually exclusive: turning one on turns its partner off, and clicking an active one turns it off. The playlist page now sets the unlocked filter on (and locked off) explicitly, since the old default no longer provides that.
The footer showed the game's full version with its build number and labelled the environment as "CDN".
Show only the part of the game version before the underscore (e.g. 11.20), and rename the "CDN" label to "Branch" since it shows the environment/branch the site was built from.
The reload visualizer and the Firepower reload rows showed two decimals.
Add an exported `RELOAD_PRECISION` (1) in the reload visualizer and use it for the visualizer's times and placeholders and for the Firepower reload rows (per-shell reloads, reload and intra-clip).
**Enter key**
- Enter redirected to the previous results' first tank while a search was still buffering, because the key handler captured a stale `searching` value.
- Enter on an empty field redirected to the first tank of the unfiltered list (Sheridan).
- The handler now reads live filter state and only acts when there is a settled search, matching the quick-link button.
**Full names in search**
- Search only knew short tank names, so queries like "super pershing" found nothing when the game stores a short form.
- Add an optional `name_full` to `TankDefinition`, filled from the game's full name when a short name exists, and index it (plus an accent-stripped copy) in `fetchTankNames` / `SEARCH_KEYS`.
- Displayed names are unchanged. Full names become searchable once the definitions are regenerated on the next build.
The hero loading bar was offset below the title by half the font size plus a fixed 16px. The title is centered as a line box, so its descenders and heavy caps reach lower than half the font size, and the fixed gap didn't grow with the title.
Offset the bar by 0.75 of the font size instead, so the gap scales with the title.
The gun flexibility visualizer at the bottom of the Tankopedia Firepower section computed depression/elevation purely from the raw gun `pitch.min`/`pitch.max` values, ignoring `initial_turret_rotation.pitch` (the tank's baked-in turret mount tilt). Every other place that reports gun depression/elevation already accounts for this offset; the visualizer was the outlier.
On tanks with a level turret mount the offset is zero, so it was invisible almost everywhere. Minotauro's turret sits pitched 3° forward, which made the visualizer's min/max/pitch stats and wedge shape disagree with the Firepower stats directly above it (e.g. 7°/-18° instead of 10°/15°).
**Changes:**
- `GunFlexibilityVisualizer/index.tsx`: fold the turret's initial pitch offset into `minPitch`, `maxPitch` and the displayed `pitch` stat, and into the drawn limit shape, the pitch dot and the pointer mapping, so the dashed neutrality circle sits at the displayed 0°.
- `GunFlexibilityVisualizer/components/CanvasContent.tsx`: nest the gun mesh inside a turret group carrying the same static tilt as its base rotation, so the mini 3D preview's live drag-pitch composes on top of it, mirroring `useTankTransform.ts`.
## Summary
- Add a Spall Liner toggle button to the Tankopedia model viewer, placed alongside the existing Calibrated Shells / Enhanced Armor buttons. It's only shown for tanks that can actually equip Spall Liner.
- The toggle reads/writes the same `Duel.protagonist.provisions` array as the Provisions tab, so both stay in sync automatically.
- If a provision slot is free, clicking equips Spall Liner immediately. If all slots are full, a popover shows icons with a shuffle arrow so you can pick which active provision to swap out in one step.
- Buttons use the same medium rounding as the consumable buttons on the tank page.
- When Spall Liner is equipped, the Dynamic Armor 3D viewer's click-damage tooltip now applies the same 20% HE damage reduction Spall Liner grants in-game.
- Refactored the hardcoded `-0.2` HE damage coefficient in `tankCharacteristics.ts` into a shared constant so the Compare tool and the 3D viewer use a single source of truth.
## Scope
This only affects the single-tank Tankopedia page. The Firepower stats table (the viewed tank's own outgoing gun damage) is intentionally untouched, since a defensive damage reduction would be semantically backwards there. The Dynamic Armor viewer's click tooltip is the only place a "damage taken by this tank" number exists, so that's where the reduction is wired in.
PrimaryArmorSceneComponent rebuilt its ShaderMaterial every render with null uniforms that were only set once on mount. Create it once with useState instead.
Replace wordy swap prompt with compact icon row (spall liner icon,
swap icon, current provisions) and drop unused ProvisionButton
tooltip that flashed on popover open.
Adds a **Peekaboo DPM** entry nested under DPM for single-shot guns, estimating damage per minute when only firing after resetting camouflage (or waiting out the reload, whichever is slower), assuming at least 10 seconds between shots.
- Shots per minute are capped at `60/10` for reloads faster than 10s, and fall back to `60/reload` otherwise
- The result is multiplied by the currently selected shell's damage and rounded
- An (i) icon next to the label opens a popover explaining the calculation
- The row is hidden for auto-loader/auto-reloader guns
- No percentile ranking bar is shown, since no dataset exists yet for this new stat
The /session/ page crashed when a tracked player's tank stats included
a tank with no matching tank definition or no average stats (WN8
baseline). The delta memo dereferenced `tank.id` on a possibly missing
definition and forced `average?.mu!` through as non-null, so a missing
average threw inside `compositeStats`/`calculateWN8` uncaught, breaking
the page on every load while the bad state persisted in localStorage.
Skip those entries instead of crashing: entries with no tank
definition or no average `mu` are dropped from the delta via
`flatMap`, so the rest of the session table still renders.
Was rendering the raw lowercase layerTypeNames key instead of resolving it through strings.website.tools.tankopedia.sandbox.dynamic.shot_card.element, same pattern already used in ShotDisplayCard.
There's a few things you'll need to get started. If this is your first time working with BlitzKit, it's completely safe to ignore all optional prerequisites below to get things running as quickly as possible.
## Invite the bot
### Bun
Go to [application directory for BlitzKit](https://discord.com/application-directory/1097673957865443370) and click "Add to Server." Voila! You're ready to go.
[Bun](https://bun.sh/) is the package manager and runtime that BlitzKit uses.
Ideally, you should also uninstall Node.js if you have it installed to avoid any cases where Node.js is used to run code instead of Bun which happens in some cases. But, this is not required and is unlikely to cause issues.
### Protocol Buffer
BlitzKit uses [Protocol Buffer's protoc CLI](https://protobuf.dev/installation/) to generate efficiently packed data structures to serve data to clients in place of JSON.
### World of Tanks Blitz
You will need a local installation of [World of Tanks Blitz](https://wotblitz.com/) for BlitzKit to build off.
### aria2 (Optional)
If you plan to use BlitzKit's client-management CLI for BlitzKit Previews, you will need to [get aria2](https://aria2.github.io/) to speed up your downloads.
### libarchive (Optional)
If you plan to use BlitzKit's client-management CLI for BlitzKit OpenTests, you will need to [get libarchive](https://www.libarchive.org/) to keep your downloads completely in-memory.
### Linux (Optional)
All work on BlitzKit is done on Linux and macOS. It is theoretically possible to work on Windows; please feel free to try and share your discoveries with us!
## Developing: Setting Up
Start by cloning the repository and changing directories.
> There is a closed source submodule under `packages/closed` that you can safely ignore. This sole purpose of this submodule is to provide `dvp.ts`, a proprietary Wargaming game asset tree compression tool. You will not need this at any point.
Then, download the dependencies with Bun.
```bash
bun install
```
Finally, compile the proto files into TypeScript.
```bash
build build-protos
```
## Developing: Environment
Go through the different packages in the `packages` directory and clone their respective `template.env`'s to `.env` files. Follow the instructions in the `.env` files to fill them out as much as you can.
## Developing: Running
Launching the website is as easy as running the `dev` script.
```bash
bun dev
```
If you ever make a change to any proto file, make sure to rebuild them.
It's all technical updates! The HD tanks will be rolled out over the span of few days.
## Technical Changes
- Shadows are disabled if your framerate drops below 30fps
- Pages that redirect to other pages (like the old /tools/tankopedia) now load a few hundred times faster
- Replaced Zustand with a custom state manager called Varuna with the goal to rewrite a significant portion of the website to not require any React for rendering increasing performance
- The transition wireframe in the Tankopedia sandbox now uses a lower quality model for performance reasons
- Tanks with PBR now use the correct maximum resolution textures
- This is a significant update as you can now literally see individual splats of dust on tanks and scratches haha
- Baked the ambient occlusion maps into the textures, hence reducing the overall model load times even with higher resolution textures
- Added a new foggy-wireframe transition between tabs to smooth out the jarring jumps
- Added animations pop-in and out animations on a few elements
- The tank name title now shrinks in size as soon as a disturbance is caused
- The shell selector, antagonist switcher, and thicknesses range now ease into frame
- The Tankopedia search page now loads as a skeleton by default before populating with tanks, allowing for the recently viewed tanks to load without cumulative shift
- Something similar was done to the Gallery tool
- Gave the meta section of the website some zest with nation-based backdrops
- The compare tabs in the Tankopedia are now centered
## Fixes
- The camera no longer snaps to the initial position when switching between tabs
## Technical Changes
- Backported some technologies from BlitzKit Reforged; you will not notice any differences in usage
- The popular tanks and gallery indexings have been moved to the asynchronous cron job
- The tank names in the Tankopedia sandbox now have less margin from the navbar
- Controls in the Tankopedia sandbox are now shown in all modes, except for the advanced highlighting button which only appears in the dynamic armor mode
- The settings page now reacts to input again
- The settings page now has a tab title and description for search engines
- Added support for transparent textures, fixing tanks like the Herdbreaker and the WZ-132-1
- The tank filters in the Performance and Tier-List tools work once again
- Reduced the amount of code that executes for every rendered frame, hopefully improving performance
- Reduced the number of lights by 20% to decrease the amount of shadow calculations
- Decreased the ground size by 80% for an equal reduction in shadow texture sizes
- Added the ability for BlitzKit to dismount the components that handled the intro lighting animation when not animating, reducing checks running every frame
This update massively upgrades BlitzKit's graphical fidelity (while making it faster) and updates the way BlitzKit deploys updates, speeding up future versions.
## Features
- Improved the performance of the Tank Sandbox in Tankopedia
- Reduced the number of lights by 25%
- Reduced the resolution of shadows by 50%
- Improved the graphical fidelity of tanks in the Tankopedia
- Added soft contact shadows below tanks
- Added sky and ground reflected lighting to the top and bottom of tanks
- Removed the fog that used to darken tanks when zoomed far away
- Made BlitzKitLand™ fade out circularly to hide the sharp edges
- Pointed the lights 1.5m above the ground
- This genuinely makes BlitzKit almost photo-realistic in my opinion
- Non-PBR tanks are now lit up brighter to accommodate for their dark textures
## Fixes
- Fixed a bug where the ambient occlusion map (a measure of how dark a part of a tank gets due to low light bounces) was applied to the color of the tank directly, making many tanks appear dark even in direct light
- Tanks with chinese characters now generate urls through transliteration
## Known Issues
- The flags found behind the selected tank in the tech tree section of the Tankopedia is offset incorrectly
## Technical Features
- BlitzKit can now calculate and store performance statistics and account details for up 10,000,000,000 players
- This will help make the BlitzKit Performance tool more accurate, and play well with an upcoming, unannounced major feature
## Technical Changes
- The `production` branch now deploys updates to `blitzkit.app` instead of the `main` branch
-`robots.txt` now disables indexing of BlitzKit variants based on the environment
- The website is now compiled in "chunks" of which the `api` chunk now takes on the task of mangling what was previously stored in `assets3`
-`assets3` will be removed in `1.31.1`, please migrate to the new `api` if you use any BlitzKit assets
This update fixes an issue that occurred with Wargaming chose to declare the class of a tank in the game code with a tag that isn't the first in the list.
Thank you to [Mark Greene](https://github.com/Hitjack007) for your contributions this update.
## Features
### Tankopedia & Armor
- Added a spall liner button to the Tankopedia beside the other equipment [Mark Greene]
- Listed track repair times in the Tankopedia [Mark Greene]
- The penetration chance is now displayed when penetration testing armor [Mark Greene]
- Added a new setting to disable automatically loading tanks in Tankopedia [Mark Greene]
### Tank Performance & Mechanics
- Added Peekaboo DPM [Mark Greene]
- Your Peekaboo DPM is lower than your normal DPM if your reload is faster than 10s as you'll have to wait for the remaining 10s for you to get unspotted before you can poke and shoot
### Tank Filters
- The test and owned tank filters are now more intuitive as clicking on them now turns the clicked option on instead of off [Mark Greene]
### Game Compatibility
- Added compatibility for tier XI tanks [Mark Greene]
## Fixes
### Tankopedia & Armor
- The armor no longer disappears when you turn on Dynamic Armor System [Mark Greene]
- Camouflage values in Tankopedia are now more descriptive when shooting [Mark Greene]
- The penetration chance is now displayed when penetration testing armor [Mark Greene]
- Listed track repair times in the Tankopedia [Mark Greene]
- Added a spall liner button to the Tankopedia beside the other equipment [Mark Greene]
- The flags in the Tankopedia tech tree are no longer offset to the top left [Mark Greene]
- The loading bar below the name of the tanks in Tankopedia is now sufficiently spaced below [Mark Greene]
- Removed the vertical sliver of pixels found to the right of the Equalizer toggle in Tankopedia
### Tank Search & Filters
- The search bar no longer redirects to a seemingly random tank when clicking enter when searching [Mark Greene]
- Tank search now properly handles the full names of tanks like the Super Pershing [Mark Greene]
- The tank search now remains consistent in order even when an added filter does not add/remove tanks from the list [Mark Greene]
- The test and owned tank filters are now more intuitive as clicking on them now turns the clicked option on instead of off [Mark Greene]
### Visuals & UI
- Removed the Beta badge from the Equalizer button [Mark Greene]
- Updated the background image for the Equalizer button to the actual banner of the new game mode
- The contact shadows on the ground no longer grow & darken when switching between tabs [Mark Greene]
- Additionally, the contact shadows stays visible at low angles [Mark Greene]
- The "Help Translate BlitzKit" text on the home page is now properly centered vertically
- The footer now says "Branch" instead of "CDN" [Mark Greene]
- The game version in the footer now displays a shorthand version instead of the entire version number [Mark Greene]
- The reload visualizer now displays just 1 decimal place [Mark Greene]
- Plot controls for the Static armor tab no longer pokes in from the right on non-Static armor tabs in some languages [Mark Greene]
- Mixer's controllers now scale better on smaller screens [Mark Greene]
- The list of available tanks in the Playlist is no longer excessively wide [Mark Greene]
### Tank Mechanics & Models
- The gun depression visualizer now properly handles the inherent tilt of tanks like the Minotauro [Mark Greene]
- When moving the tracks of tanks, the textures no longer move at different rates [Mark Greene]
### Localization
- Users with a preference on a language are once again redirected to the correct localized pages [Mark Greene]
### Session & Game Compatibility
- BlitzKit Session now properly handles tanks removed from the game [Mark Greene]
## Technical Fixes
- BlitzKit now handles the new ID system for game-modes in WoTB
- The directory which preview versions of the game are downloaded to are now created properly
- This was the last bug standing in the way of BlitzKit Previews
- Welcome back, BlitzKit Previews!
- Removed some dead code from the BlitzKit model mangler
- The player discovery loop now persists for longer than 1 iteration [Mark Greene]
- Bun is now forced to run scripts with the `--bun` flag
## Technical Features
- Added documentation to the README on setting up and developing BlitzKit locally
- Added bug report and feature request forms to GitHub
- Switched to the GNU AGPLv3 license
## Technical Fixes
- BlitzKit now handles the new ID system for game-modes in WoTB
- The directory which preview versions of the game are downloaded to are now created properly
- This was the last bug standing in the way of BlitzKit Previews
- Welcome back, BlitzKit Previews!
- Removed some dead code from the BlitzKit model mangler
- The player discovery loop now persists for longer than 1 iteration [Mark Greene]
- Bun is now forced to run scripts with the `--bun` flag
Go to [application directory for BlitzKit](https://discord.com/application-directory/1097673957865443370) and click "Add to Server." Voila! You're ready to go.
What is considered a "day's worth of games" is determined by the reset time. The reset time is generally at midnight everyday after which newer games are counted towards the next day. Reset times vary from region to region:
- North America: `UTC-5:00` (Central North American Time midnight)
- Europe: `UTC+1:00` (Central European Time midnight)
- Asia: `UTC+7:00` (Central Asia Standard Time midnight)
**Reset times are different from Blitz Stars** (Blitz Stars resets at 5:00 am locally relative to the user opening their website). Inconsistencies with Blitz Stars only occur with sleep-deprived players and players who play in the wrong region.
This content was pre-recorded on a test server and the version you may see may not be the same. Please see [the official WoTB portal](https://wotblitz.com/) for the most up-to-date information.
The holiday season has arrived, and this year the celebrations spill directly into your garage, transforming the familiar space into a festive hub filled with activities, rewards, and seasonal flair. Instead of just offering decorations or a visual makeover, this event brings a full set of mechanics that turn everyday battles into a steady stream of progress. Whether you’re aiming for bonuses, collectibles, or new tanks, the garage is designed to make every match feel like part of a larger holiday journey.
## The Music Box
The familiar Snow Globes are absent this year, replaced entirely by **Music Boxes**, which act as the new way to try your luck for seasonal rewards. They behave similarly to the Snow Globes and give you a chance to obtain various items, including a shot at the brand-new **[GSOR the TANK](https://blitzkit.app/tanks/gsor-the-tank/)**. But this time around, **they do not guarantee a tank every single time.** But as a bonus, the Music Box guarantees you will not receive duplicate tanks. Once you have a tank, the Music Box will try its best to give you a different one each time.
## Win Battles, Earn Rewards
Each victory during the event rewards you with **3 Hot Coffees**, a seasonal currency that serves as the backbone of the progression system. These coffees can be exchanged for a variety of Festive Town rewards, giving you reasons to keep pushing through battles even if you’re just casually playing.
With the coffees, you’ll also be able to claim **2 tank certificates**, one for **tier VII** vehicles and another covering **tier IX–X** options.
The tier VII pool includes the [E 25](https://blitzkit.app/tanks/e-25), [CS-52 LIS](https://blitzkit.app/tanks/cs-52-lis), [AMX 13 57](https://blitzkit.app/tanks/amx-13-57), [Steyr WT](https://blitzkit.app/tanks/steyr-wt), [Kunze Panzer](https://blitzkit.app/tanks/kunze-panzer), and [Super Hellcat](https://blitzkit.app/tanks/super-hellcat). Meanwhile, the tier IX–X certificate gives access to higher-tier picks like the [121B](https://blitzkit.app/tanks/121b), [AMX 30 B](https://blitzkit.app/tanks/amx-30-b), [Obj. 260](https://blitzkit.app/tanks/obj-260), [Char Futur 4](https://blitzkit.app/tanks/char-futur-4), [TL-7-120](https://blitzkit.app/tanks/tl-7-120), and [Object 452K](https://blitzkit.app/tanks/object-452k).
## Upgrade Your Christmas Tree
Golden tokens allow you to upgrade your Christmas Tree through **four distinct stages**, each bringing your garage closer to its fully decorated holiday form.
As the tree grows, additional ornament slots open up, eight in total, which appear on the left side of your screen. These slots give you more room to customize the look of your garage and make the tree feel like something you’ve built up piece by piece over the course of the event. Each new ornament adds to the atmosphere, making the progression feel both cosmetic and rewarding.
A fully decorated tree grants up to **+25% silver** per battle, turning your regular gameplay into a more profitable experience without requiring extra effort. Ensure all your ornaments come from the same set to unlock the full potential of the tree (all left, all middle, or all right).
## Upgrade Your Yuletide Stall
Your golden tokens can also be invested into upgrading the **Yuletide Stall**, another element of the garage that evolves as you level it up. Each upgrade unlocks new rewards, giving you incentives to distribute your tokens across different attractions instead of focusing on only one. The stall begins in a simple state, but after upgrades it transforms visually and functionally, becoming a more vibrant part of your garage.
Each upgrade unlocks a tier worth of rewards that you can claim in exchange for Free XP, Gold, Silver, and more Hot Coffees!
## Upgrade Your Shooting Gallery
Another upgradeable area is the **Shooting Gallery**, which contributes its own set of rewards as it levels up. Enhancing the gallery expands your options for exchanging mastery badges, making those badges more useful during the event than they usually are.
KitScore is a performance metric that makes sense. This metric can be applied to any game, virtual or real. Specific pieces of information relating directly to World of Tanks Blitz will be provided within parentheses and with a "WoTB:" prefix.
## Consumption of Observations
KitScore is driven solely by the behavior of other players and finding meaning of a new observation within a pool of condensed data based on previously observed data. Generally, we will be provided with a number of statistics (the $x_i$ of an observation) per player (WoTB: $x_\text{frags}$, $x_\text{xp}$, $x_\text{spots}$, etc.) including a target statistic (the $x_j$ of an observation) that all other statistics try to maximize (WoTB: "winrate" which we will defined later as $x_j$).
If the statistics are cumulative, they will have to be normalized (WoTB: statistics fetched from Wargaming are indeed cumulative even for a single player; $x_\text{frags}=\frac{\Sigma x_\text{frags}}{\Sigma x_\text{battles}}$, $x_\text{spots}=\frac{\Sigma x_\text{spots}}{\Sigma x_\text{battles}}$, $x_j=\frac{\Sigma x_\text{wins}}{\Sigma x_\text{battles}}$, etc.). Each player will have such average statistics, giving us many points of data to work with. If the statistics are not cumulative, then you will be dealing with multiple observations per player which will only help improve accuracy.
## It's Impossible to "Find What Makes the Metric Tick"
WN8, a widely adopted metric in the World of Tanks and World of Tanks Blitz communities is plagued with problems. One of them being that people often find and exploit "what makes the metric tick." In other words, people easily exploit the rudimentary nature of WN8 to score undeservedly high.
In WN8, the gains of some rewarding categories overshadow the losses of others though that may not necessarily be in the best interest in maximizing winrate. KitScore makes this nearly impossible as it punishes a player for scoring well solely high in a certain category. You will never exploit KitScore. Period.
| $A_\text{human}$ | The human readable version of the raw score $A$. |
| $A$ | The raw KitScore. |
| $C$ | An arbitrary coefficient. |
| $I$ | The number of statistics including the target statistic. This means that there are $I-1$ entries for $x_i$ and $1$ entry for $y$ for each observation. |
| $i$ | The statistic index of where there are $I$ entries. |
| $j$ | A discriminator for the target statistic. There may never be more than $1$ entries for $j$. |
| $m$ | The degree of the polynomial regression. |
| $n$ | The number of observations. |
| $a$ | An "atomic" score which is a function of the $i$th statistic. |
| $w$ | The weight of the $i$th atomic statistic. |
## Computing the Score
The final metric $A_\text{human}$ is an arbitrarily scaled value of the raw score $A$ where $C$ is any coefficient of your liking. BlitzKit uses $C=10^3$.
$$
A_\text{human}=CA
$$
The final metric $A$ is the weighted average of individual scores $a$ which are functions of the $i$th statistic excluding the target statistic $j$ itself.
The "atomic" score $a$ is the difference between the predicted target statistic $P_{i \leftrightarrow j}(x_i)$ acquired using a polynomial regression of degree $m$ and the actual newly observed $x_j$.
$$
a_i=x_j-P_{i \leftrightarrow j}(x_i)
$$
The polynomial regression of $m$th degree can be written as follows:
Polynomial regression coefficients $\beta$ can be acquired through basic linear algebra where an $m \times m$ of constants is multiplied with $m \times 1$ of unknowns which equals another $m \times 1$ of constants.
The weight $w$ is a slice of the Pearson corelation matrix. The numerator is the corelation between the $i$th statistic and the target statistic $j$ while the denominator the sum of the corelations between the $i$th statistic and all statistics but the target statistic $j$.
Forming the corelation matrix will create an $I \times I$ matrix where. This matrix is best represented as a table. The example below illustrates a matrix/table for $I=4$ which includes $3$ entries for $i$ statistics and $1$ entry for the $j$ target statistic.
It should not be a surprise to find $r_{i \leftrightarrow i}^2=1$ which forms a neat diagonal across the table. Note that there are many duplicates within the $I^2$ entries. The "bottom-right triangle" can be eliminated to remove redundancies leaving us with only $\frac{I(I+1)}{2}$ fields occupied.
When programming this matrix into an array, $r_{i \leftrightarrow i}^2=1$ will be redundant. I recommend condensing the matrix into an array of size $\frac{I(I+1)}{2}+I=\frac{I(I-1)}{2}$ to avoid all redundant entries. For $I=4$, the array will look something like the one illustrated below (here, $j=4$).
Some may find the weight equation arbitrary; it evidently isn't. The goal of $w$ is to eliminate over-rewarding the same behavior represented in multiple statistics. It makes all atomic statistics concurrent by leveraging the Pearson corelation matrix.
A simplified example will clarify the idea. First, we compute a Pearson corelation matrix for $I=3$ statistics. We have $x_1$ and $x_2$, the statistics that are trying to maximize $x_j$. Reminder: a Pearson corelation matrix finds the $r$ value not only between $x_i$ and $x_j$ but also between all combinations of $x_i$ and themselves. Here, $r^2$ is calculated using polynomial regression corelation analysis to the $m$th degree.
| | $x_1$ | $x_2$ | $x_3$ | $x_j$ |
| ----- | ----- | ----------- | ----- | ----- |
| $x_1$ | $1$ | $\boxed{0}$ | $0$ | $1$ |
| $x_2$ | | $1$ | $0$ | $1$ |
| $x_3$ | | | $1$ | $1$ |
| $x_j$ | | | | $1$ |
In the case above, $r^2_{1 \leftrightarrow 2}=0$ meaning that both statistics have absolutely nothing to do with each other. Hence, the player will be awarded for optimizing both individually. And since $r^2_{1 \leftrightarrow j}=r^2_{2 \leftrightarrow j}$, they should both be rewarded equally; in other words, they should have equal weights.
$$
w_{1}=\frac{1}{1 + 0 + 0}=1
$$
$$
w_{2}=\frac{1}{0 + 1 + 0}=1
$$
$$
w_{3}=\frac{1}{0 + 0 + 1}=1
$$
Now consider a case where $x_1$ and $x_2$ are fully correlated. It would be a bad idea to reward them equally; hence, their weights suffer.
Loaded 100 of 952 files, more files were not shown because too many files have changed in this diff.
Show more
Reference in new issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
tresabhi
Mark Greene
GonnaHetzMeandTrèsAbhi
GonnaHetzMe
dependabot[bot]
Mark Greene
imgbot[bot]andImgBotApp
TresAbhi