fix/mathc_collector: fix splitting of boots and first backs

frontend/server/utils/cdragon-cache.ts

@@ -9,14 +9,19 @@ const readFileAsync = promisify(readFile)
 const CDRAGON_BASE = 'https://raw.communitydragon.org/'
 
 // Cache directory - can be configured via environment variable
-// Default to dev/cdragon for development
+// In development, use dev/data/cdragon relative to project root
+// In production, use /cdragon (shared volume)
 const getCacheDir = () => {
   if (process.env.CDRAGON_CACHE_DIR) {
     return process.env.CDRAGON_CACHE_DIR
   }
-  // Default to dev/cdragon relative to project root
+  // Check if we're in development mode (explicitly set)
+  if (process.env.NODE_ENV === 'development') {
     return join(process.cwd(), '..', 'dev', 'data', 'cdragon')
   }
+  // Default to /cdragon for production (Docker)
+  return '/cdragon'
+}
 
 /**
  * Get the current patch from the patch.txt file or fallback to 'latest'

match_collector/src/cdragon_cache.ts

@@ -46,7 +46,8 @@ async function downloadCDragonAssets(patch: string) {
   console.log(`\n=== Downloading CDragon assets for patch ${cdragonPatch} ===`)
 
   // Get cache directory from environment or use default
-  // In development, use a local directory relative to project root; in production (Docker), use /cdragon
+  // In development, use a local directory relative to project root
+  // In production (Docker), use /cdragon (shared volume with frontend)
   const defaultCacheDir =
     process.env.NODE_ENV === 'development'
       ? resolve(__dirname, '../../dev/data/cdragon')

match_collector/src/champion_stat.ts

@@ -177,7 +177,7 @@ function handleMatchBuilds(
   participantIndex: number,
   builds: Builds,
   platform?: string
-): Build {
+): { build: Build; startItemId: number | undefined } {
   const timeline: Timeline = match.timeline
 
   // Find or create the build for this participant's rune configuration
@@ -185,6 +185,7 @@ function handleMatchBuilds(
   build.count += 1
 
   const items: Array<{ itemId: number; goldAdvantage: GoldAdvantageTag; platform?: string }> = []
+  let startItemId: number | undefined = undefined
   for (const frame of timeline.info.frames) {
     for (const event of frame.events) {
       if (event.participantId != participantIndex) continue
@@ -267,9 +268,11 @@ function handleMatchBuilds(
   // This tree includes start item as the root, then branching paths
   if (items.length > 0) {
     treeMerge(build.items, items)
+    // The first item is the starter item
+    startItemId = items[0].itemId
   }
 
-  return build
+  return { build, startItemId }
 }
 
 function handleMatch(match: Match, champions: Map<number, ChampionData>, platform?: string) {
@@ -365,14 +368,22 @@ function handleMatch(match: Match, champions: Map<number, ChampionData>, platfor
     }
 
     // Items and runes (builds)
-    const build = handleMatchBuilds(match, participant, participantIndex, lane.builds, platform)
+    const { build, startItemId } = handleMatchBuilds(
+      match,
+      participant,
+      participantIndex,
+      lane.builds,
+      platform
+    )
 
-    // First back data - store at build level
+    // First back data - store at build level with start item tracking
     const firstBackData = extractFirstBackFromMatch(match, participantIndex)
     if (firstBackData) {
       if (!build.firstBacksRaw) {
         build.firstBacksRaw = []
       }
+      // Include the starter item ID for proper filtering when splitting builds
+      firstBackData.startItemId = startItemId
       build.firstBacksRaw.push(firstBackData)
     }
   }
@@ -445,16 +456,44 @@ function splitMergeOnStarterItem(build: Build, championName: string): BuildWithS
     console.log(`Warning: for champion ${championName}, start item splits build variant.`)
     const builds = []
     for (const c of build.items.children) {
+      // Calculate the ratio for proportional distribution
+      const ratio = c.count / build.count
+
+      // Proportionally distribute boots counts
+      const scaledBoots = build.boots.map(b => ({
+        data: b.data,
+        count: Math.round(b.count * ratio)
+      }))
+
+      // Proportionally distribute suppItems counts
+      const scaledSuppItems = build.suppItems.map(s => ({
+        data: s.data,
+        count: Math.round(s.count * ratio)
+      }))
+
+      // Proportionally distribute bootsFirstCount
+      const scaledBootsFirstCount = Math.round(build.bootsFirstCount * ratio)
+
+      // Filter firstBacksRaw by starter item
+      let filteredFirstBacksRaw: FirstBackData[] | undefined
+      if (build.firstBacksRaw && build.firstBacksRaw.length > 0) {
+        // Filter by the starter item ID that was tracked when storing firstBacksRaw
+        filteredFirstBacksRaw = build.firstBacksRaw.filter(fb => fb.startItemId === c.data)
+        if (filteredFirstBacksRaw.length === 0) {
+          filteredFirstBacksRaw = undefined
+        }
+      }
+
       builds.push({
         runeKeystone: build.runeKeystone,
         runes: build.runes,
         items: c,
-        bootsFirstCount: build.bootsFirstCount,
+        bootsFirstCount: scaledBootsFirstCount,
         count: c.count,
         startItems: [{ data: c.data!, count: c.count }],
-        suppItems: build.suppItems,
-        boots: build.boots,
-        firstBacksRaw: build.firstBacksRaw
+        suppItems: scaledSuppItems,
+        boots: scaledBoots,
+        firstBacksRaw: filteredFirstBacksRaw
       })
       c.data = undefined
     }

match_collector/src/item_tree.ts

@@ -264,12 +264,8 @@ function deriveTags(node: ItemTree, expectedRegionDistribution?: PlatformCounts)
     const totalExpected = REGION_KEYS.reduce((sum, key) => sum + expectedRegionDistribution[key], 0)
 
     if (totalExpected > 0) {
-      // Tag if the item is significantly more popular in a region (>= 1.5x expected rate)
-      // and has a minimum absolute percentage (>= 10%)
-      const SIGNIFICANCE_THRESHOLD = 1.5
-      const MINIMUM_PCT = 0.1
-
-      // Loop through all regions to derive tags
+      // Tag if one region accounts for >= 60% of the normalized distribution
+      // Normalized value = actual percentage / expected percentage ratio
       const regionTags: Array<{ key: keyof PlatformCounts; tag: ItemTag }> = [
         { key: 'euw', tag: 'region_euw' },
         { key: 'eun', tag: 'region_eun' },
@@ -277,12 +273,23 @@ function deriveTags(node: ItemTree, expectedRegionDistribution?: PlatformCounts)
         { key: 'kr', tag: 'region_kr' }
       ]
 
-      for (const { key, tag } of regionTags) {
+      // Calculate normalized values (actual/expected ratio) for each region
+      const normalizedValues = regionTags.map(({ key, tag }) => {
         const expectedPct = expectedRegionDistribution[key] / totalExpected
         const actualPct = node.platformCount[key] / totalRegionCount
+        const normalizedValue = expectedPct > 0 ? actualPct / expectedPct : 0
+        return { tag, value: normalizedValue }
+      })
 
-        if (actualPct >= expectedPct * SIGNIFICANCE_THRESHOLD && actualPct >= MINIMUM_PCT) {
+      const totalNormalized = normalizedValues.reduce((sum, { value }) => sum + value, 0)
+
+      // Tag the region if it accounts for >= 60% of the normalized distribution
+      if (totalNormalized > 0) {
+        for (const { tag, value } of normalizedValues) {
+          if (value / totalNormalized >= 0.6) {
             tags.push(tag)
+            break // Only tag the most dominant region
+          }
         }
       }
     }

match_collector/src/types.ts

@@ -82,6 +82,7 @@ export interface ItemSet {
 export interface FirstBackData {
   timestamp: number
   itemSet: ItemSet
+  startItemId?: number
 }
 
 /**