Dub Top Market Beaters (Core Creators)

Purpose

Return the current Top Market Beaters leaderboard from Dub's Core Creators on web.dubapp.com — the curated list of Core portfolios that are outperforming the market. For each ranked portfolio: name, ticker, creator display name, today's % return, all-time % return, short strategy description, and canonical profile URL. Read-only; never copies, follows, or invests.

When to Use

• Daily / hourly tracking of which Core creator portfolios are leading the market on Dub.
• Building dashboards or alerts on top-performing copy-trading portfolios.
• Comparing Core (free-to-copy, curated) vs. Premium (paid) leaderboards on Dub.
• Any agent task that says "top creators / top portfolios / market beaters on Dub" — that exact phrase is a literal h2 section heading on /explore.

Workflow

The full leaderboard is fully server-side rendered in the HTML returned by a single anonymous HTTP GET to a stable, public URL. No JavaScript execution, no login, no proxy, no captcha, no ?type= toggle parsing — just GET → parse HTML. Sub-second, ~$0.0001 per fetch via Browserbase Fetch. Lead with the HTTP path; the browser flow is an unnecessary 100× cost wrapper around the same SSR'd payload.

1. Fetch the page (Core Creators):

   GET https://web.dubapp.com/explore/market-beaters
   

   For Premium creators use GET https://web.dubapp.com/explore/premium/market-beaters (returns the analogous 20-card grid with title Top Market Beaters - Premium). The query-string form ?type=premium is silently ignored on this URL — the Core vs Premium toggle is path-based, not param-based.

   Headers: none required. No Referer, no User-Agent, no Cookie. The page returns 200 OK with the full grid SSR'd into the HTML.

2. Confirm the page rendered. Sanity-check by looking for the canonical h1 in the response body:

   <h1 class="... font-serif text-display-lg">Top Market Beaters</h1>
   

   and the subtitle <p>Portfolios outperforming the market</p>. If the h1 is missing, you were probably redirected (e.g. AppsFlyer smart-banner injecting JS); refetch.

3. Split the grid into card chunks. Each portfolio card begins with the literal anchor open-tag:

   <a class="flex h-full w-full" href="/portfolios/{HANDLE}">
   

   Splitting the HTML on <a class="flex h-full w-full" href="/portfolios/ yields one chunk per card. The page renders exactly 20 cards on a single page (no pagination, no infinite scroll, no "See more" — /explore/market-beaters is the canonical full list as of the trace). Order is by rank: index 0 = rank 1.

4. Extract fields per card chunk (regex against each chunk):

   Field	Source
   handle (portfolio ticker, uppercase)	The leading capture ([A-Za-z0-9_]+) immediately after the split delimiter
   name (portfolio display name)	<p class="truncate text-md font-semibold text-primary">([^<]+)</p>
   ticker	<span class="font-mono ... text-action">\$<!-- -->([A-Z0-9_]+)</span> — note the literal <!-- --> comment node Next.js injects between $ and the value
   creator (display name)	<span class="truncate text-secondary">([^<]+)</span> — the first such span after the · separator. May be a first-name + last-initial (e.g. John P.), a single name (Jeremy, Seb), or an org/handle (APMIA, Hugh Mungus)
   today_return_pct	>Today<.*?<span ... text-(gain|loss|primary)[^"]*>([+-]?\d+\.\d+%)</span> (DOTALL)
   all_time_return_pct	>All-Time<.*?<span ... text-(gain|loss|primary)[^"]*>([+-]?\d+\.\d+%)</span> (DOTALL)
   description	<p class="line-clamp-\d ... text-secondary ...">([^<]+)</p> — HTML-decode &quot;, &amp;, &#x27;, &#39;, Unicode entities (’ → ')
   profile_url	https://web.dubapp.com/portfolios/{handle}

5. Sort sanity-check: the page is ranked by all-time return, with rank 1 being the highest. Rank-1's all_time_return_pct should be the largest positive value among the 20 cards. The marketing copy ("Portfolios outperforming the market") means all 20 entries are alpha-positive over the lifetime window — but today_return_pct can be deeply negative on a market-wide red day and that's expected (the page does not re-rank intraday).

6. Stamp as_of with the current UTC date — the page itself does not include a server-side "last updated" timestamp. Cache-Control on the response is private, no-cache, no-store, max-age=0, must-revalidate, so each fetch is freshly regenerated server-side.

Browser fallback

Only justified if Dub introduces a future anti-bot wall on the HTTP path (none observed). If needed:

browse open https://web.dubapp.com/explore/market-beaters --remote
browse wait load
browse get html body          # returns the same SSR'd HTML as the HTTP fetch
# parse with the same regexes from step 4 — the browser does not unlock any
# additional fields, it just adds cost and turn count

Do not click any portfolio card during fallback — the portfolio detail page is not needed for the leaderboard task, and a stray click on the rendered "Copy" or "Invest" button is a real-money action surface (dub is a SEC-registered broker-dealer; this is a real trading app, not a paper demo).

Site-Specific Gotchas

• No anti-bot wall observed. Verified by 2 successive fetches with proxies: true and 1 fetch with proxies: false — all 3 returned 200 OK with the identical 20-card payload in the same rank order. proxies are not required for this endpoint. Default to omitting --proxies to save ~$0.0008 per fetch.
• The phrase "Top Market Beaters" is a literal h2 section heading on multiple pages. Disambiguate by URL: /explore shows a 3-card carousel preview of Top Market Beaters with a "See all" link pointing to /explore/market-beaters; the latter is the full 20-card grid. The carousel on /explore only contains 3 cards in SSR — don't scrape it directly, follow the "See all" link.
• Core vs Premium routing is path-based, not param-based. /explore/market-beaters is Core (free copy-trading creators). /explore/premium/market-beaters is Premium (paid creators — Pelosi, Buffett, Infinity, etc.). The query-string ?type=premium is silently ignored on /explore/market-beaters (page title stays "Top Market Beaters", not "...- Premium"). The ?type=core|premium toggle only works on /leaderboard, not on /explore.
• Don't confuse /leaderboard with /explore/market-beaters. They are different surfaces with different data:
  • /leaderboard?type=core has three tables (Highest Returns, Most Copiers, Most Copying Capital), 9 rows each, with three time-window tabs (All Time / Monthly / Today). Only the active tab (All Time by default) is SSR'd — the Monthly and Today panels exist in the DOM but their <div> body is empty until a client-side fetch fires on tab click. There is no observed URL param that selects a different default tab — tested ?window=, ?period=, ?tab=, ?time= (all monthly/today empty).
  • /explore/market-beaters is the dedicated "Top Market Beaters" page — 20 cards, no tabs, no time-window switcher. Today's % and All-Time % are both shown on every card.
• Next.js <!-- --> comment nodes inside spans break naïve regex. The ticker is rendered as <span ...>$<!-- -->JOHNDPHAN</span>, not <span ...>$JOHNDPHAN</span>. Always include <!--\s*--> in your ticker regex or use a tolerant capture (\$\s*(?:<!--[^>]*-->)?\s*([A-Z0-9_]+)).
• Creator handle may be missing on some leaderboard rows. On /leaderboard?type=core, the row for portfolios without a public creator @username (e.g. $NUKEFUTURE) has no @username span at all — it shows only the portfolio name and ticker. On /explore/market-beaters, every card has a creator field (display name), but format varies: first-name + last-initial (John P.), single name (Jeremy, Seb, APMIA), or full alias (Hugh Mungus). Don't assume a @handle pattern.
• Image srcSet payloads are huge (~30% of HTML weight) — the cards include 12-width responsive srcSet URLs each ~3KB. Strip <img ... srcSet="..."/> before regex-extracting fields to make parsing 3× faster, or just regex against the original HTML — both work, the latter is fine for 20-card pages.
• Cache-Control is no-store. Every fetch hits the origin (X-Vercel-Cache: MISS). Don't rate-limit yourself based on cache-hit assumptions — but also don't hammer: sustained ≤ 1 req/s is polite for a leaderboard that updates intraday at most.
• Sandbox WebSocket egress note (skill-build environment, not skill-runtime): the Browserbase remote-browser path (connect.usw2.browserbase.com WebSocket) is not reachable from the Vercel Sandbox host that built this skill — only the Browserbase REST API (api.browserbase.com) was reachable. That's why this skill was verified end-to-end via POST /v1/fetch rather than browse open --remote. For agents consuming this skill from environments with full network egress, both paths work; the HTTP path is still strictly preferred for cost.
• dub is a regulated SEC-registered broker-dealer. Treat any clickable button on a portfolio detail page as a potential trade-execution surface. The leaderboard extraction itself touches no execution surface — but during browser fallback, scope to read-only verbs (get html, screenshot, snapshot) and never click a portfolio card.

Expected Output

{
  "success": true,
  "page_url": "https://web.dubapp.com/explore/market-beaters",
  "section": "Top Market Beaters",
  "subtitle": "Portfolios outperforming the market",
  "creator_type": "core",
  "as_of": "2026-05-18T00:00:00Z",
  "card_count": 20,
  "cards": [
    {
      "rank": 1,
      "name": "johndinhphan",
      "ticker": "$JOHNDPHAN",
      "handle": "JOHNDPHAN",
      "creator": "John P.",
      "today_return_pct": "-5.30%",
      "all_time_return_pct": "+543.22%",
      "description": "I find value where others overlook it - safe, growing, and built to last. One year is all it takes.",
      "profile_url": "https://web.dubapp.com/portfolios/JOHNDPHAN"
    },
    {
      "rank": 2,
      "name": "Copy The Cat",
      "ticker": "$COPYDACAT",
      "handle": "COPYDACAT",
      "creator": "Carlos C.",
      "today_return_pct": "-6.54%",
      "all_time_return_pct": "+147.61%",
      "description": "\"Copy the Cat\" monitors the moves of the most successful institutional investors to identify elite starting points for our analysis...",
      "profile_url": "https://web.dubapp.com/portfolios/COPYDACAT"
    }
  ]
}

For the Premium-creator variant, set creator_type: "premium" and page_url: "https://web.dubapp.com/explore/premium/market-beaters" (also 20 cards, same schema).

Failure shapes (rare — only if the HTTP fetch itself fails):

{ "success": false, "reason": "fetch_failed", "status_code": 5xx }

{ "success": false, "reason": "card_grid_missing", "note": "h1 'Top Market Beaters' present but 0 cards parsed — possible markup migration; refetch and inspect the leading anchor pattern" }