Alert Outcome Analysis — April 27–28, 2026
Period: Apr 27–28, 2026 Total alerts: 31 (1 removed — butthole, false alert, dev self-launch) 2x definition: ATH market cap > $70K post-migration Data sources: Birdeye OHLCV, DexScreener, manual verification
Results by Tier
| Tier | 2x | Total | Win Rate |
|---|---|---|---|
| EXTREME | 3 | 10 | 30% |
| FAST | 2 | 6 | 33% |
| MODERATE | 2 | 6 | 33% |
| SLOW | 1 | 4 | 25% |
| STALLED | 2 | 5 | 40% |
| TOTAL | 10 | 31 | 32% |
Winners (confirmed ATH > $70K)
| Tier | Name | Peak MC | Traders in BC |
|---|---|---|---|
| EXTREME | Vesting | $307K | 1 |
| EXTREME | 1 billion speedrun | $70K+ | 2 |
| EXTREME | Dwayne | ~$68K | 1 |
| FAST | non profit coin | $162K | 3 |
| FAST | chetgpt | $100K | 3 |
| MODERATE | Justice For Luca Cella Walker | $230K | 1 |
| MODERATE | Judge Network | ~$67K | 2 |
| SLOW | werld coin | $82K | 1 |
| STALLED | Bork | $186K | 1 |
| STALLED | Michael jackson Chimpanzee | $128K | 1 |
Near-Misses (ATH $55–69K)
| Tier | Name | ATH MC | Gap to 2x |
|---|---|---|---|
| EXTREME | LOCKED IN | $56K | -$14K |
| MODERATE | MurderGPT | $61K | -$9K |
| SLOW | open asshole | $65K | -$5K |
| MODERATE | MolechGPT | $54K | -$16K |
| EXTREME | Elon Musk Must Win | $54K | -$16K |
Hard Failures (ATH ≤ $40K — never left migration MC)
Scams Pump the Hardest, OMNITRADING, Israeli shekel, WeLoveDicks, Scamcoin, Scams And Profit 500, Legit Coin, For Profit Coin, Dog In Vest, Unfazed, OpenLie, Helping Hand, AmericanReserveModernizationAct, Sam Snakeman, up
Pattern Analysis
Pattern 1 — “Scam” in the name = near-certain failure
Every token with “scam”, “profit from scam”, or a scam-themed name failed without exception:
| Token | Result |
|---|---|
| Scams Pump the Hardest | ❌ $38K |
| Scams Pump the Hardest (2nd) | ❌ $36K |
| Scams And Profit 500 | ❌ $33K |
| Scamcoin | ❌ never left migration MC |
| Sam Snakeman | ❌ never left migration MC |
These tokens attract bot volume during the BC fill (hence known traders triggering alerts) but have no organic buyer narrative post-migration. The name telegraphs zero conviction. This is a filterable signal.
Pattern 2 — EXTREME tier is mostly bot battles that immediately dump
10 EXTREME alerts. 3 winners. The 7 losers all peaked within 33–38K) — meaning they pumped at launch, never had post-migration momentum, and dumped immediately. The EXTREME fill speed reflects bot competition for supply, not market demand for the token.
The 3 EXTREME winners all had distinguishable narratives: Vesting (part of a “vest/invest” meta cluster), 1 billion speedrun (meme with clear identity), Dwayne (celebrity tie-in). The losers had either no narrative or oversaturated names.
EXTREME ≠ quality. EXTREME = competition. The winners are the ones with a reason to keep running after bots exit.
Pattern 3 — The “vest” narrative cluster
Three tokens launched in rapid succession (~19:40 PT): Vesting (10s, ✅ 56K). This was a meta-cluster — one name inspired copies.
Vesting won because it was first. Dog In Vest and LOCKED IN were copies riding the same narrative. Only the original captured the full momentum. The copies had BC buyers (known traders entered all three) but the market only sustained one version.
When the same concept spawns 3+ tokens within minutes, only the original is worth trading.
Pattern 4 — STALLED tier has the highest win rate (40%)
Both STALLED winners (Bork 128K) were slow-BC conviction plays — known traders had been accumulating for 30+ minutes before migration. A slow BC fill with sustained known-trader buying is a stronger quality signal than a 5-second EXTREME fill driven entirely by bots.
A trader accumulating for 30 minutes has conviction. A bot buying in 5 seconds is just execution.
Pattern 5 — Near-misses cluster around $54–65K
5 tokens peaked in the 70K threshold. These are tokens with a real post-migration run but insufficient momentum to sustain 2x. Common traits: 1 trader in BC, weak or ambiguous narrative, moderate tier. The signal was real but the setup wasn’t strong enough for full follow-through.
Pattern 6 — Multi-copy tokens almost always lose
“Dog In Vest” migrated 5 separate times in one session. “1 billion speedrun” migrated twice. “LOCKED IN” migrated 3 times. In every case, only the first version had a chance. All subsequent copies failed.
The bot alerts on all of them because it sees traders in BC — but the traders themselves are likely arbitraging the launch, not expressing conviction in the token. Duplicate name detection would eliminate a meaningful slice of losing alerts.
Synopsis
The detector is correctly identifying when known traders are active — but being active in BC does not guarantee a token runs. The 32% win rate across all tiers hides a more nuanced picture: quality of the underlying token matters as much as trader presence.
The clearest finding is that two variables predict failure more reliably than tier or trader count: a scam-themed name and being a copy of a token that already migrated. Remove those two categories and the win rate improves significantly on the remaining pool.
STALLED tier is underrated. Slow-BC conviction trades outperform EXTREME alerts on a win-rate basis. High speed means high competition, not high quality.
EXTREME tier is overrated. High speed means high competition, not high quality. Most EXTREME tokens are bot battles that die within minutes of migration. The winners in EXTREME had strong standalone narratives that kept buyers engaged post-migration.
Suggestions for Improving Alert Quality
1. Add a name blacklist filter
Filter or downgrade alerts where the token name contains: scam, rug, profit, ponzi, or obvious scam-bait phrases. Every scam-named token in this sample failed. This is a zero-cost quality filter with no confirmed false negatives in the data.
BLACKLIST_KEYWORDS = ["scam", "rug", "ponzi", "exploit", "drain"]
def is_blacklisted(name: str) -> bool:
return any(kw in name.lower() for kw in BLACKLIST_KEYWORDS)Expected impact: Removes ~5 losing alerts per session, likely 0 winners lost.
2. Detect and flag duplicate token names
Track token names seen within the current session. If the same name (or close variant) has already migrated and been alerted, flag the new alert as a copy.
SESSION_NAMES_SEEN = set() # normalized token names this session
# If name in SESSION_NAMES_SEEN → append "[COPY]" to alert titleExpected impact: Removes most of the “Dog In Vest copy #4” style alerts. In this sample, 5+ tokens were clear copies.
3. Raise minimum threshold for single-trader EXTREME alerts
Single-trader EXTREME alerts won 2/8 (25%) — Vesting and Dwayne both won but are outliers with strong narratives. Multi-trader EXTREME (2+) won 1/2. Consider requiring a minimum buy size of 1+ SOL for single-trader EXTREME hits before alerting, filtering out low-conviction small buys.
4. Surface trader buy size more prominently in alerts
The buy size (SOL) is in the alert but needs to be the headline signal. A trader putting 5 SOL in BC is categorically different from 0.5 SOL. Higher buy = stronger conviction = better predictive value. Weight alerts by total SOL committed, not just trader count.
5. Track cumulative SOL across all traders as a composite score
Instead of “2 traders in BC”, surface “4.2 SOL in BC across 2 traders”. Total SOL committed by known traders is a better quality signal than trader count alone. Add a total_bc_sol field to alerts and log it for outcome tracking.
6. Add outcome tracking to the bot
Currently there’s no feedback loop. The bot fires alerts but never learns whether they worked. Add a lightweight post-migration price check at t+30min and t+60min for alerted tokens, logged to the DB. Over time this builds a dataset for improving signal weights.
Key Takeaways
- 32% overall win rate — 10/31 alerts resulted in 2x+
- STALLED is the best tier (40%) — slow-BC conviction trades outperform bot-speed trades
- EXTREME is noisy (30%) — speed alone doesn’t predict quality
- Scam-named tokens: 0% win rate — filterable, zero false negatives
- Copy tokens: 0% win rate — duplicate name detection would clean this up
- Near-misses cluster at $54–65K — close but a real gap; narrative quality determines if they break through
P&L Simulation — 0.1 SOL per Alert, Sell at ATH
Assumptions
- Entry MC: $33K (pump.fun standard migration price) for all tokens
- Exit: peak ATH market cap — theoretical sell at exact top
- Position: 0.1 SOL per trade, no fees or slippage
- Hard failures with no confirmed ATH estimated at $33–35K (never meaningfully left migration MC)
- SCAM token included as if the bot caught it — it was designed to, the Telegram bug blocked delivery
Total trades: 32 (31 alerted + 1 missed) Total deployed: 3.2 SOL
Winners
| Token | Tier | Peak MC | Mult | Return | P&L |
|---|---|---|---|---|---|
| Justice For Luca Cella Walker | MODERATE | $230K | 6.97x | 0.697 SOL | +0.597 SOL |
| Vesting | EXTREME | $307K | 9.30x | 0.930 SOL | +0.830 SOL |
| Bork | STALLED | $186K | 5.64x | 0.564 SOL | +0.464 SOL |
| non profit coin | FAST | $162K | 4.91x | 0.491 SOL | +0.391 SOL |
| Michael jackson Chimpanzee | STALLED | $128K | 3.88x | 0.388 SOL | +0.288 SOL |
| chetgpt | FAST | $100K | 3.03x | 0.303 SOL | +0.203 SOL |
| werld coin | SLOW | $82K | 2.48x | 0.248 SOL | +0.148 SOL |
| 1 billion speedrun | EXTREME | $70K | 2.12x | 0.212 SOL | +0.112 SOL |
| Dwayne | EXTREME | $68K | 2.06x | 0.206 SOL | +0.106 SOL |
| Judge Network | MODERATE | $67K | 2.03x | 0.203 SOL | +0.103 SOL |
| Subtotal | 3.948 SOL | +2.948 SOL |
Near-Misses
| Token | Tier | Peak MC | Mult | Return | P&L |
|---|---|---|---|---|---|
| open asshole | SLOW | $65K | 1.97x | 0.197 SOL | +0.097 SOL |
| MurderGPT | MODERATE | $61K | 1.85x | 0.185 SOL | +0.085 SOL |
| LOCKED IN | EXTREME | $56K | 1.70x | 0.170 SOL | +0.070 SOL |
| MolechGPT | MODERATE | $54K | 1.64x | 0.164 SOL | +0.064 SOL |
| Elon Musk Must Win | EXTREME | $54K | 1.64x | 0.164 SOL | +0.064 SOL |
| Subtotal | 0.880 SOL | +0.380 SOL |
Hard Failures
Peaks estimated at $33–38K — breakeven or negligible gain
| Token | Tier | Est. Peak | Mult | Return | P&L |
|---|---|---|---|---|---|
| Scams Pump the Hardest | EXTREME | $38K | 1.15x | 0.115 SOL | +0.015 SOL |
| Scams Pump the Hardest #2 | EXTREME | $36K | 1.09x | 0.109 SOL | +0.009 SOL |
| OMNITRADING | FAST | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| WeLoveDicks | FAST | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| Scams And Profit 500 | FAST | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| Legit Coin | STALLED | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| For Profit Coin | STALLED | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| Dog In Vest | EXTREME | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| Unfazed | SLOW | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| OpenLie | FAST | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| Helping Hand | MODERATE | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| AmericanReserveModernizationAct | STALLED | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| up | MODERATE | $35K | 1.06x | 0.106 SOL | +0.006 SOL |
| Israeli shekel | SLOW | $34K | 1.03x | 0.103 SOL | +0.003 SOL |
| Scamcoin | EXTREME | $33K | 1.00x | 0.100 SOL | 0 SOL |
| Sam Snakeman | EXTREME | $33K | 1.00x | 0.100 SOL | 0 SOL |
| Subtotal | 1.687 SOL | +0.087 SOL |
SCAM — Missed Alert (Bot Was Built to Catch This)
The Telegram bug (& in BullX URL) blocked delivery on all 6 send attempts. The detect was correct — this was a known-trader hit that should have gone out.
| Token | CA | Peak MC | Mult | Return | P&L |
|---|---|---|---|---|---|
| SCAM | 6AVAUKa9...N2wpump | $16M | 485x | 48.485 SOL | +48.385 SOL |
Full Summary
| Category | Trades | Deployed | Returned | P&L |
|---|---|---|---|---|
| Winners | 10 | 1.0 SOL | 3.948 SOL | +2.948 SOL |
| Near-Misses | 5 | 0.5 SOL | 0.880 SOL | +0.380 SOL |
| Hard Failures | 16 | 1.6 SOL | 1.687 SOL | +0.087 SOL |
| SCAM (missed) | 1 | 0.1 SOL | 48.485 SOL | +48.385 SOL |
| TOTAL | 32 | 3.2 SOL | 55.000 SOL | +51.800 SOL |
Without SCAM — 31 alerted trades only
| Deployed | 3.1 SOL |
| Returned | 6.515 SOL |
| Net P&L | +3.415 SOL |
| ROI | +110% |
Including SCAM
| Deployed | 3.2 SOL |
| Returned | 55.000 SOL |
| Net P&L | +51.800 SOL |
| ROI | +1,619% |
SCAM alone: 0.1 SOL → 48.5 SOL (+48,400%)
What This Tells Us
The 31-alert session was profitable without any outlier — +110% ROI selling every winner at peak. In practice you won’t hit exact ATH, but even at 50–60% capture on each winner the book is green.
Hard failures cost almost nothing. The tokens that dumped immediately to $33–35K returned nearly the full 0.1 SOL — they’re closer to breakeven than losses, especially pre-fee. The real drag on quality is alert fatigue, not capital loss.
The SCAM miss is the painful one. A single Telegram URL bug converted a 485x trade into a no-trade. At 0.1 SOL entry that’s 48.4 SOL left on the table from one character (& → %26). That bug is fixed.
P&L Simulation — 2x Take-Profit vs Sell at ATH
Scenario: What if the bot auto-sold every position the moment it hit 2x (i.e., token MC reached 33K migration entry)?
Rules:
- Tokens that reach $66K MC → sell at 2x → 0.2 SOL returned on 0.1 SOL entry
- Tokens that never reach $66K → sell at ATH (same as base simulation)
- Same 32 trades, same 0.1 SOL position size
Winners — capped at 2x
All 10 winners had ATH above $66K so all would have triggered the 2x sell.
| Token | Peak MC | ATH Mult | 2x TP Return | Left on Table |
|---|---|---|---|---|
| Vesting | $307K | 9.30x | 0.200 SOL | 0.730 SOL lost |
| Justice For Luca | $230K | 6.97x | 0.200 SOL | 0.497 SOL lost |
| Bork | $186K | 5.64x | 0.200 SOL | 0.364 SOL lost |
| non profit coin | $162K | 4.91x | 0.200 SOL | 0.291 SOL lost |
| Michael Jackson Chimpanzee | $128K | 3.88x | 0.200 SOL | 0.188 SOL lost |
| chetgpt | $100K | 3.03x | 0.200 SOL | 0.103 SOL lost |
| werld coin | $82K | 2.48x | 0.200 SOL | 0.048 SOL lost |
| 1 billion speedrun | $70K | 2.12x | 0.200 SOL | 0.012 SOL lost |
| Dwayne | $68K | 2.06x | 0.200 SOL | 0.006 SOL lost |
| Judge Network | $67K | 2.03x | 0.200 SOL | 0.003 SOL lost |
| Subtotal | 2.000 SOL | 2.242 SOL lost |
Near-Misses — unchanged (ATH never reached $66K)
| Token | Peak MC | Return | P&L |
|---|---|---|---|
| open asshole | $65K | 0.197 SOL | +0.097 SOL |
| MurderGPT | $61K | 0.185 SOL | +0.085 SOL |
| LOCKED IN | $56K | 0.170 SOL | +0.070 SOL |
| MolechGPT | $54K | 0.164 SOL | +0.064 SOL |
| Elon Musk Must Win | $54K | 0.164 SOL | +0.064 SOL |
| Subtotal | 0.880 SOL | +0.380 SOL |
Hard Failures — unchanged
1.687 SOL returned, +0.087 SOL profit (same — never approached 2x)
SCAM — capped at 2x
| Token | Peak MC | ATH Return | 2x TP Return | Lost |
|---|---|---|---|---|
| SCAM | $16M (485x) | 48.485 SOL | 0.200 SOL | 48.285 SOL |
Full Comparison
| Scenario | Trades | Deployed | Returned | P&L | ROI |
|---|---|---|---|---|---|
| ATH exit (31 trades) | 31 | 3.1 SOL | 6.515 SOL | +3.415 SOL | +110% |
| 2x take-profit (31 trades) | 31 | 3.1 SOL | 4.567 SOL | +1.467 SOL | +47% |
| ATH exit (32 w/ SCAM) | 32 | 3.2 SOL | 55.000 SOL | +51.800 SOL | +1,619% |
| 2x take-profit (32 w/ SCAM) | 32 | 3.2 SOL | 4.767 SOL | +1.567 SOL | +49% |
What This Tells Us
2x TP underperforms ATH by 63% on normal trades and by 97% including SCAM.
The damage is concentrated in 3 trades: Vesting (9.3x), Justice For Luca (6.97x), and Bork (5.64x) — together they account for 1.891 SOL of the 1.948 SOL difference. A 2x take-profit would have exited all three at $66K and missed everything above.
The SCAM comparison is extreme — 2x TP captures 0.2 SOL on a 485x trade. A fixed take-profit turns a life-changing trade into a rounding error.
However, ATH exit is theoretical. Nobody sells at the exact top. A realistic 2x TP is actually executable — the bot can place a limit order or monitor MC and auto-sell. The ATH numbers assume perfect timing which doesn’t exist in practice.
The real answer: a tiered exit is better than either extreme — take 50% off at 2x, let 50% ride with a trailing stop. That captures the guaranteed double on every winner while keeping exposure on the outliers that go to 5–10x.
P&L Simulation — Tiered Exit (Suggested Strategy)
Rules:
- Entry: $33K MC, 0.1 SOL per trade
- TP1: sell 50% at 2x ($66K MC) → 0.05 SOL slice exits at 2x
- TP2: sell 30% at 3x ($99K MC) → 0.03 SOL slice exits at 3x
- TP3: sell remaining 20% at trailing stop −20% from peak → 0.02 SOL slice exits at peak × 0.80
- For tokens that never hit 2x: trailing stop fires on full position at peak × 0.80
Winners — Hits 3x+ (6 tokens, TP1 + TP2 + TP3 all fire)
| Token | Peak | TP1 (50%@2x) | TP2 (30%@3x) | TP3 (20%@peak×0.8) | Total | P&L |
|---|---|---|---|---|---|---|
| Vesting | 9.30x | 0.100 SOL | 0.090 SOL | 0.149 SOL | 0.339 SOL | +0.239 SOL |
| Justice For Luca | 6.97x | 0.100 SOL | 0.090 SOL | 0.112 SOL | 0.302 SOL | +0.202 SOL |
| Bork | 5.64x | 0.100 SOL | 0.090 SOL | 0.090 SOL | 0.280 SOL | +0.180 SOL |
| non profit coin | 4.91x | 0.100 SOL | 0.090 SOL | 0.079 SOL | 0.269 SOL | +0.169 SOL |
| MJ Chimpanzee | 3.88x | 0.100 SOL | 0.090 SOL | 0.062 SOL | 0.252 SOL | +0.152 SOL |
| chetgpt | 3.03x | 0.100 SOL | 0.090 SOL | 0.048 SOL | 0.238 SOL | +0.138 SOL |
Winners — Hits 2x but not 3x (4 tokens, TP1 fires, rest trail)
Remaining 50% (TP2+TP3 combined) exits at peak × 0.80
| Token | Peak | TP1 (50%@2x) | Remaining 50% @trailing | Total | P&L |
|---|---|---|---|---|---|
| werld coin | 2.48x | 0.100 SOL | 0.099 SOL | 0.199 SOL | +0.099 SOL |
| 1 billion speedrun | 2.12x | 0.100 SOL | 0.085 SOL | 0.185 SOL | +0.085 SOL |
| Dwayne | 2.06x | 0.100 SOL | 0.082 SOL | 0.182 SOL | +0.082 SOL |
| Judge Network | 2.03x | 0.100 SOL | 0.081 SOL | 0.181 SOL | +0.081 SOL |
Winners subtotal: 2.427 SOL returned on 1.0 SOL — P&L: +1.427 SOL
Near-Misses — No TP fires, full position exits at trailing stop (peak × 0.80)
| Token | Peak | Trailing Exit | Return | P&L |
|---|---|---|---|---|
| open asshole | 1.97x | 1.576x | 0.158 SOL | +0.058 SOL |
| MurderGPT | 1.85x | 1.480x | 0.148 SOL | +0.048 SOL |
| LOCKED IN | 1.70x | 1.360x | 0.136 SOL | +0.036 SOL |
| MolechGPT | 1.64x | 1.312x | 0.131 SOL | +0.031 SOL |
| Elon Musk Must Win | 1.64x | 1.312x | 0.131 SOL | +0.031 SOL |
Near-misses subtotal: 0.704 SOL returned on 0.5 SOL — P&L: +0.204 SOL
Hard Failures — Full position exits at trailing stop (peak × 0.80)
| Token | Peak | Trailing Exit | Return | P&L |
|---|---|---|---|---|
| Scams Pump the Hardest | 1.15x | 0.920x | 0.092 SOL | −0.008 SOL |
| Scams Pump the Hardest #2 | 1.09x | 0.872x | 0.087 SOL | −0.013 SOL |
| 13 tokens @1.06x | 1.06x | 0.848x | 0.085 SOL ×13 = 1.105 SOL | −0.078 SOL total |
| Israeli shekel | 1.03x | 0.824x | 0.082 SOL | −0.018 SOL |
| Scamcoin | 1.00x | 0.800x | 0.080 SOL | −0.020 SOL |
| Sam Snakeman | 1.00x | 0.800x | 0.080 SOL | −0.020 SOL |
Hard failures subtotal: 1.526 SOL returned on 1.6 SOL — P&L: −0.074 SOL
SCAM — Tiered exit on 485x
| Slice | Size | Exit | Return |
|---|---|---|---|
| TP1 (50%) | 0.05 SOL | 2x | 0.100 SOL |
| TP2 (30%) | 0.03 SOL | 3x | 0.090 SOL |
| TP3 trailing (20%) | 0.02 SOL | 485x × 0.80 = 388x | 7.760 SOL |
| Total | 0.1 SOL | 7.950 SOL |
vs ATH simulation: 48.485 SOL. The trailing stop fires at 388x instead of the 485x peak — still captures 7.76 SOL on the 20% slice.
Full Comparison — All Three Strategies
| Strategy | 31 Trades | Deployed | Returned | P&L | ROI |
|---|---|---|---|---|---|
| ATH exit | 31 | 3.1 SOL | 6.515 SOL | +3.415 SOL | +110% |
| 2x take-profit | 31 | 3.1 SOL | 4.567 SOL | +1.467 SOL | +47% |
| Tiered exit | 31 | 3.1 SOL | 4.657 SOL | +1.557 SOL | +50% |
| Strategy | 32 Trades (w/ SCAM) | Deployed | Returned | P&L | ROI |
|---|---|---|---|---|---|
| ATH exit | 32 | 3.2 SOL | 55.000 SOL | +51.800 SOL | +1,619% |
| 2x take-profit | 32 | 3.2 SOL | 4.767 SOL | +1.567 SOL | +49% |
| Tiered exit | 32 | 3.2 SOL | 12.607 SOL | +9.407 SOL | +294% |
What the Math Tells Us
On normal trades (no SCAM), tiered barely beats pure 2x TP (+50% vs +47%) — the difference is the near-miss and failure trailing stops recovering more than a hard -20% exit would. Both are far below ATH because ATH assumes perfect top-tick execution which doesn’t exist in practice.
On SCAM (485x), the gap is massive: tiered returns 7.95 SOL vs 0.20 SOL for pure 2x TP. The 20% trailing slice is what makes outliers worth holding — a 388x exit on a 20% position returns more than a 2x exit on the full position.
The tiered structure solves the core tension: lock in guaranteed profit on every winner at 2x, capture mid-range runners at 3x, and stay exposed on outliers via the trailing slice. It’s also fully executable — TP1 and TP2 are limit orders, TP3 is a trailing stop. No manual decisions required once the trade is open.
P&L Simulation — Worst Case (Hard Failures + Near-Misses → $0)
Scenario: Every trade that didn’t win goes to zero. No recovery, no stop loss catch, no trailing exit — full rug on all 21 non-winners. Tests how each strategy holds up when losers are total losses.
Winners unchanged. Losers return 0 SOL.
31 Trades — No SCAM
| Category | Trades | Deployed | Returned | P&L |
|---|---|---|---|---|
| Winners | 10 | 1.0 SOL | — | — |
| Near-misses | 5 | 0.5 SOL | 0 SOL | −0.500 SOL |
| Hard failures | 16 | 1.6 SOL | 0 SOL | −1.600 SOL |
| Total losses | 21 | 2.1 SOL | 0 SOL | −2.100 SOL |
| Strategy | Winners Return | Total Return | P&L | ROI |
|---|---|---|---|---|
| ATH exit | 3.948 SOL | 3.948 SOL | +0.848 SOL | +27% |
| 2x take-profit | 2.000 SOL | 2.000 SOL | −1.100 SOL | −35% |
| Tiered exit | 2.427 SOL | 2.427 SOL | −0.673 SOL | −22% |
32 Trades — With SCAM
| Strategy | Winners | SCAM | Losses | Total | P&L | ROI |
|---|---|---|---|---|---|---|
| ATH exit | 3.948 SOL | 48.485 SOL | 0 SOL | 52.433 SOL | +49.233 SOL | +1,538% |
| 2x take-profit | 2.000 SOL | 0.200 SOL | 0 SOL | 2.200 SOL | −1.000 SOL | −31% |
| Tiered exit | 2.427 SOL | 7.950 SOL | 0 SOL | 10.377 SOL | +7.177 SOL | +224% |
What This Reveals
2x TP goes negative in the worst case. −35% without SCAM, −31% even with it. Capping all winners at 2x while taking full losses on 21 trades is a losing book — the wins don’t cover the losses.
Tiered is also negative without SCAM (−22%) but recovers dramatically with SCAM (+224%) because the 20% trailing slice turns the 485x into 7.95 SOL. Pure 2x TP only gets 0.20 SOL from SCAM regardless of how big it goes.
ATH holds up best (+27% even in worst case) because the higher per-winner returns (avg 3.95x vs 2.0x for 2x TP) generate enough to absorb 21 total losses. Still theoretical — you can’t sell at ATH.
The critical takeaway: this scenario is exactly why the stop loss matters. The tiered strategy includes a trailing stop that catches failures at peak × 0.80 instead of zero. In the realistic tiered simulation that returned +50% ROI, hard failures recovered to ~0.85x and near-misses to ~1.4x — neither went to zero. This scenario shows what happens if you skip the stop loss entirely and let everything ride to the bottom. Don’t skip the stop loss.
Synopsis
The worst-case scenario stress-tests every strategy against total capital loss on 21 of 31 trades. The results are unambiguous.
2x take-profit breaks. It’s the only strategy that loses money even when a 485x trade is in the book (−31%). Capping winners at 2x while absorbing full losses on losers produces a losing book regardless of outliers. The math doesn’t work: 10 winners at 2x return 2.0 SOL, 21 losers at zero cost 2.1 SOL. You’re underwater before SCAM even enters the equation.
Tiered survives the outlier, not the grind. Without SCAM, tiered is negative (−22%). The stop loss is what separates the realistic tiered simulation (+50%) from the worst case — the trailing stop catches losers at ~0.85x instead of zero, which is the difference between a +1.5 SOL book and a −0.7 SOL book. The 20% trailing slice is what keeps the strategy alive on outliers: SCAM alone returns 7.95 SOL under tiered vs 0.20 SOL under 2x TP.
ATH holds up best in worst case (+27%) because its higher per-winner returns (avg 3.95x) generate enough to absorb 21 full losses. This confirms that holding runners matters more than protecting against losers — the losers are nearly breakeven in practice anyway.
The stop loss is the single most important variable in this dataset. It’s not about which take-profit strategy you use — it’s about not going to zero on the 68% of trades that don’t win. A trailing stop at −20% from peak transforms worst-case losers from 0 to 0.85x, which is the difference between a losing and winning book.
All Simulations — Master Comparison
| Strategy | 31 Trades (normal) | ROI | 32 Trades (w/ SCAM) | ROI |
|---|---|---|---|---|
| ATH exit | +3.415 SOL | +110% | +51.800 SOL | +1,619% |
| Tiered exit | +1.557 SOL | +50% | +9.407 SOL | +294% |
| 2x take-profit | +1.467 SOL | +47% | +1.567 SOL | +49% |
| ATH, worst case | +0.848 SOL | +27% | +49.233 SOL | +1,538% |
| Tiered, worst case | −0.673 SOL | −22% | +7.177 SOL | +224% |
| 2x TP, worst case | −1.100 SOL | −35% | −1.000 SOL | −31% |
Trader Leaderboard — April 2026
Scoring: 1 point per coin bought in bonding curve that hit $70K+ ATH post-migration.
| Rank | Trader | Points | Winning Coins |
|---|---|---|---|
| 🥇 1 | dv | 5 | 1 billion speedrun, non profit coin, chetgpt, Judge Network, Bork |
| 🥈 2 | chester | 2 | Vesting, 1 billion speedrun |
| 🥈 2 | theo | 2 | non profit coin, Judge Network |
| 🥈 2 | parsiix | 2 | non profit coin, werld coin |
| 🥈 2 | kevnszn | 2 | chetgpt, Justice For Luca Cella Walker |
| 6 | trenchman | 1 | Dwayne |
| 6 | decu | 1 | chetgpt |
| — | unknown | — | Michael Jackson Chimpanzee |
dv is the standout — 5 winning coins across 4 different tiers (EXTREME, FAST, MODERATE, STALLED). Consistent presence on quality setups regardless of speed. chester, theo, parsiix, and kevnszn each had 2 winners. Michael Jackson Chimpanzee buyer unconfirmed.
Leaderboard Connections & Pattern Analysis
dv is the single most reliable signal in this dataset
dv appeared on 5 of the 10 winners — and across every tier. EXTREME (1 billion speedrun), FAST (non profit coin, chetgpt), MODERATE (Judge Network), STALLED (Bork). This is not a trader who scalps fast tokens — this is a trader with genuine conviction across all migration speeds. In a session with 31 alerts, dv was in the BC on 5 winners and none of the hard failures appear in their wallet.
Implication: dv in BC is the closest thing this dataset has to a standalone buy signal. If the system fires an alert and dv is one of the traders, that alert deserves elevated attention regardless of tier.
Trader consensus = higher peak MC
The two FAST winners with the highest peaks were both 3-trader alerts:
| Coin | Traders | Peak MC |
|---|---|---|
| non profit coin | dv, theo, parsiix | $162K |
| chetgpt | kevnszn, decu, dv | $100K |
The two MODERATE winners also had multi-trader presence (Judge Network: dv + theo, Justice For Luca: kevnszn). Compare to single-trader winners like werld coin (186K) — Bork is the outlier, explained by the STALLED pre-watch pattern below.
Implication: This supports Suggestion #5 — total SOL and trader count are composite quality signals. More known traders in BC = more conviction = higher ceiling. A 3-trader alert from dv, theo, and parsiix is categorically different from a 1-trader alert from an unknown S-tier wallet.
The STALLED conviction pattern holds
Pattern 4 identified STALLED as the highest win-rate tier. The leaderboard confirms why: Bork (STALLED, $186K) had dv accumulating through a slow BC fill. A trader with dv’s track record sitting in a bonding curve for 30+ minutes is not a bot — it’s a thesis. The system’s original instinct to skip STALLED alerts was wrong. Pre-watch on slow fills with known traders is one of the strongest setups in the data.
EXTREME winners required a second known trader
Every EXTREME winner had at least one high-conviction trader behind it:
- Vesting — chester (solo, $307K) — narrative: “vest” meta was live, chester identified it first
- 1 billion speedrun — dv + chester (2 traders, $70K) — two traders agreeing on an EXTREME fill is a meaningful signal
- Dwayne — trenchman (solo, $68K) — celebrity tie-in narrative
The 7 EXTREME losers had either no known traders or a single low-conviction entry. Pattern 2 said EXTREME = bot competition, not demand — this confirms it. The bot fills fast, but the traders who stuck around to buy narratives (vest meta, celebrity name) are the ones who won. EXTREME + named trader with conviction = real. EXTREME alone = noise.
dv + theo is a pattern worth watching
dv and theo appeared together on 2 winners: non profit coin and Judge Network. They also appeared separately on other winners. When both are in BC on the same token, the win rate in this sample is 2/2. Small sample but worth flagging — if these two are aligned on a setup, the signal is stronger.
Key Trader Signals for Future Alerts
Based on this session, here’s how to weight traders when reading an alert:
| Signal | What it means |
|---|---|
| dv in BC | Elevated priority — 5/10 winners, 0 confirmed hard failures |
| dv + any second trader | Strong consensus signal — multi-winner in both combos |
| kevnszn in BC | Quality narrative filter — both wins had strong stories (chetgpt, Justice For Luca) |
| 3+ traders in BC | Highest ceiling — both 3-trader alerts hit $100K+ |
| Solo STALLED trader | Check if it’s slow accumulation — Bork was dv sitting in BC for 30+ min |
| EXTREME + 0 known traders | Bot battle — skip or size down |
Related
- home-solana-apr2026 — EXTREME narrative case study
- migration-speed-signal — speed tier framework
- narrative-triggers — what makes a token run post-migration
- migration-detector-architecture — system architecture