📖 Case studies · real signals · real on-chain data
Four real signals. Four different stories.
Our system fires three signal types: cluster-buys (2+ elite wallets co-buy), elite single-wallet picks (top-15 wallets buying alone), and stealth accumulation— the pattern Birdeye's new Token Holder Chart API unlocked: holders quietly rising while price stays flat, before any cluster fires. Real on-chain examples of each:
Case study · cluster-buy
$TROLLUDI: +42% in 60 minutes
On May 5th, three separate signals fired on $TROLLUDI across 2 hours as more elite wallets piled in. Demonstrates how the system escalates conviction as a cluster grows — and why entering on the FIRST signal beats waiting for confirmation.
⏱ Signal timeline
2 wallets cluster-buy
First signal fires when wallets ranked #54 and #32 (both top-100 by 30d PnL) independently buy $TROLLUDI within 60 minutes. Score: 36.1/100.
4 wallets — top-25 joins in
An hour later, two more elite wallets including rank #22 (a top-25 trader by PnL) buy in. Score jumps to 43.7. Token has already moved up — but the cluster is strengthening.
Wallet ranks
#22, #31, #36, #53
Top buyer rank
#22 (top 25)
8 wallets pile in — but it's the top
By 12:45 the cluster has grown to 8 wallets (ranks #21 through #53). Score: 62.3. But this is the late-arrival herd — the smart money got in 2 hours earlier. Within minutes, the token starts dumping.
Wallet ranks
#21, #22, #25, #28, #29, #31, #36, #53
🎯
The system caught the move
Three separate signals fired at the right inflection points. Signal #1 (10:35) caught entry, Signal #2 (11:40) caught the +42% peak, Signal #3 (12:45) marked the top.
⚡
Speed beats size
Buying signal #1 at 10:35 = +42% peak. Buying signal #3 at 12:45 = -97% rug. Same token, same alert system, opposite outcomes — entry timing is everything.
📤
Take-profit is the rule
The peak window was real but short. Anyone who held overnight got rugged. The strategy is: enter on early signals, set a +20-50% take-profit, and exit when filled — not when emotional.
🔬 What the system saw on-chain
Every step of this case was driven by real data from the Birdeye Data API. Here's the chain of API calls behind a single signal:
- Wallet polling — Every few minutes,
/trader/txs/seek_by_time fetches recent trades for each of our 494 tracked wallets. - Cluster detection — Our scanner groups buys by token, looking for ≥2 distinct top-rank wallets buying within a 60min window.
- Token enrichment —
/defi/token_overview pulls price, liquidity, market cap. We require min $25k liquidity. - Rug check —
/defi/v3/token/holder verifies top-10 holders own less than 60% (rug filter). - Signal scored & persisted — Score = function of buyer ranks, volume, and recency. Saved to DB.
- Distribution — Telegram broadcast with inline buy buttons (Jupiter, Raydium, Birdeye), Discord webhook fanout, X post if score ≥ 75, web push notifications.
- Live tracking —
/defi/multi_pricesnapshots prices at +1h, +6h, +24h. Rolling max becomes the "peak" metric on the track record page.
Case study · elite single-wallet
$COPE: Rank #2 wallet bought solo · score 100/100
Different signal type, different outcome. When a single ultra-elite wallet (rank #2 by 30d PnL on our leaderboard) buys a token without other wallets joining, the system fires an "elite single" signal — high conviction from caliber alone, but no cluster confirmation.
📊 Two signal types compared
Our system fires both, scores them differently, and lets you filter for which type you want.
Cluster-buy
2+ elite wallets, same token, 60min window
Best for: momentum plays. Multiple wallets converging on a token is the strongest single indicator we track.
Example: $TROLLUDI +42%
Elite single-wallet
Top-15 wallet buys solo, $200+ position
Best for: early signal — sometimes a top-5 wallet sees something before anyone else does.
Example: $COPE +9%, score 100
🦅 Sprint 3 case study · stealth accumulation · first fire
$HATCHER: holders rising, price flat — before any cluster
Birdeye shipped the Token Holder Chart API on May 7. We shipped stealth detection on top of it on May 8. The first fire: $HATCHER, signal #223 — holders went from 186 → 190 (+2.15%) over a 5-minute window while price stayed flat (+16.5% over 24h, no acceleration). No smart-money cluster had fired on it yet. That's the pattern price charts can't show.
Holder growth
186 → 190
+2.15% in 5 min
Detected via Birdeye token_overview holder field, compared against pool snapshot baseline.
Token state
$71.2K
market cap · $23.9K liq
Below typical cluster-signal floor — exactly the population stealth is designed to find.
Score
52 / 100
stealth tier
Score formula: 50 + growth %. Independent of cluster strength — holder velocity IS the signal.
🔬 What the system saw
- T-90 min: Pump-scan endpoint discovers $HATCHER via Helius (recent pump.fun program transaction). Adds to snapshot pool with 186 holders baseline at that moment.
- T-5 min: Detect-signals cron fires. Snapshot age now ≥ 5 min (fresh-enough gate passes). Birdeye token_overview returns current holder count: 190 holders.
- T-0: Growth = (190 - 186) / 186 = +2.15%. Above stealth gate. Price gate clears (24h change in window). Liquidity + mcap gates clear. Stealth signal #223 persists.
- T+1s: Telegram broadcast goes out with custom 🦅 layout (holder-progression line instead of buyer cluster). Grok-4 generates AI commentary referencing the accumulation thesis. HolderChart on the signal page renders the multi-point trajectory.
Want the next one delivered live?
Every signal arrives in Telegram with one-click Jupiter/Raydium buy buttons the moment it fires.