Equity curve — cumulative P&LEvery resolved trade at the fixed $100 stake, in resolution order. Shaded band is peak-to-trough drawdown; the dotted tail is unrealised open exposure.
Papertrade Leaderboard · Strategy Simulation Lab
One qualified alert, all isolated strategies, the same entry price, and separate fake balances.
All saved Papertrade coins.
Paper only$100 fake entries · variable Jupiter fee estimateHow it works
Simulation only. No real funds or wallet trades are used. TP/SL levels are locked at entry; some bots
use delayed stops, breakeven-moves, trailing stops, or scaled exits (see each bot's mechanic).
Prices use an adaptive scan interval, with one-minute candle-high/low recovery to backfill brief wicks.
Live market cap comes from DexScreener and may briefly lag Axiom.
Strategy Leaderboard
| # | Strategy | Equity / P&L | Profit / trade | W-L-Open | Current trade | Win Rate | Max DD |
|---|---|---|---|---|---|---|---|
| Loading saved paper-trade rankings… | |||||||
Historical Rank Changes
Saved automatically whenever the leaderboard order changes. Balance-only updates that keep the same order do not create fake rank events.
Loading saved rank history…
Bot Details
Loading the leading bot’s saved trades…
🐦 KOL Tweet Alerts
Open Qualified Alerts
Every qualified coin with an open paper position stays here until the tracker sells it. Click any alert for the complete buy, live path, fee and sell details.
Every open bot paper trade
Each row follows that bot's own TP, SL, trailing, scale-out, or timed-exit rules and shows an after-fee value only when a fresh exact-pool quote exists.
| Coin | Bot / Strategy | Entry | Current | Take Profit | Stop Loss | Exit / Status | If Sold Now | Opened | Full Detail |
|---|
Qualified Coins
Every coin that passed all scanner filters. Paper-tracked at a fixed $100 stake, full exit at TP +50% or stop -30% (whichever hits first). Stats above test that +50/-30 idea across every coin ever scanned, qualified or not — the larger sample.
Qualified Coin StatsPerformance analytics built only from Qualified coins in the selected scope. Open calls are never counted in any win rate.
Waiting for Qualified data…
Does the filter actually add value?Qualified calls against every coin the scanner has ever tracked, on the same +50/-30 paper rules.
Time of day — volume and win rateEvery individual hour in local time. Bars are call volume; the line is win rate. Hollow points mean too few resolved calls to read.
Narrative performanceAverage P&L per call, sorted by what actually made money — not by what was most common.
Entry market cap vs resultAverage P&L per call by entry band. Directly usable for filter tuning.
Does data confidence predict outcome?Win rate and average P&L per confidence band. If Verified does not beat Caution, that is itself a finding.
Time to outcomeHow long wins and losses take to resolve. Tells you whether losers die fast and winners grind, or the reverse.
Excursion analysis — how far did trades run before resolving?Each dot would be one resolved trade: how far up it went (right) against how far down it went (down), with the +50% target and -30% stop drawn in.
Near Misses
Coins the scanner saw but rejected — kept as record-only research, same paper-trade math as Qualified Coins.
NEAR MISS = failed one or more active filtersNo alert or simulated bot entry. It is tracked for research so Self-Learning can discover whether a filter was useful or too strict.
How data quality protects learning
1. Study every resultQualified coins and near misses remain in research totals.
2. Grade the evidenceEach result receives a frozen reliability score.
3. Block weak evidenceScores below 85 cannot change live filters.
4. Explain the fixOpen a row to see the exact pool checks.
Coins needing attention
These records are not automatically bad. Their evidence is stale, incomplete, or inconsistent, so the scanner retries them and keeps them out of learning changes.
Narratives
A read-only backtest of every current Qualified coin. Open trades stay out of win-rate math.
SELECTED NARRATIVE
Choose a narrative
Self-Learn Progress
Forward-only threshold study. Suggestions never change live qualification until the owner reviews and saves them.
🧠 What the learner is doing
1. Record everythingQualified coins and near misses are both followed for 24 hours.
2. Compare outcomesIt checks which entry-time liquidity, volume and holder levels produced more wins.
3. Wait for evidenceA suggestion needs enough trusted training and newest-holdout evidence to reduce lucky overfitting.
4. Owner decidesThe learner never silently changes live filters. Review a suggestion, then save it manually if wanted.
📅 Daily win / loss — all scanned coins
Includes qualified coins and near misses. Expired 24-hour calls count as non-wins so weak results are not hidden.
Near-Miss Learning Lab
Studies every rejected coin, especially missed winners, to show which filters may be too strict. This is research only until enough evidence graduates a rule.
Most common rejection reasons
📈 Qualified-Only Win-Rate Trend
Rolling win rate over the last 10 resolved Qualified calls only. Near misses are excluded, so this graph answers whether the actual scanner selections are improving.
Show newest qualified results:
🌈 Narrative Learning — Qualified Coins
Compares which narratives produced wins and larger sold values. Only resolved Qualified calls with verified data can influence the learned narrative-score floor.
💧 Liquidity floor tuning
Liquidity thresholds are compared against the 80% target. A bar needs at least 20 matching outcomes before it is trusted.
🚩 Fee-volume floor tuning
1h-volume thresholds are compared against the 80% target and need at least 20 matching outcomes before they can change the live floor.
📈 Buy-pressure floor tuning
5-minute net buy-pressure thresholds are tested against the 80% target. A level needs at least 20 matching outcomes before it can change the live filter.
Protocol-fee signal tuning
Compares SOL fee commitment separately for New Pair, Final Stretch and Migrated coins. Only verified outcomes with at least 20 matching samples can tighten a live stage floor.
👥 Top-10 holder ceiling tuning
Loosest holder-concentration ceiling that still keeps the target win rate. The learner LOWERS this to reject concentrated, rug-prone coins.
🔄 Filter Graduation Log
Every time a floor above actually gets raised live, it's logged here — this is the scanner acting on what it learned, not just displaying it.
How it self-improves: every scan records immutable entry-time features and a 24-hour outcome for every
coin it sees, qualified or not. Thresholds shown above are informational until each bucket has enough
samples — more data improves confidence, it cannot guarantee winners.
Filters
Change any threshold and hit Save — it applies to the next scan instantly, no restart. A coin must pass all of these to be Qualified.
Pulse entry strategy now active
- New Pairs: at least 0.8 SOL estimated fees paid (raise to 1 SOL for safer entries).
- Final Stretch: at least 3 SOL; test up to 5 SOL as a stricter variant.
- Migrated: at least 8 SOL.
- KOL confirmation: at least 1 independent account in the 5K–20K follower cohort, applied only when your official X watcher is connected.
- Liquidity and volume: recorded for self-learning, not hard blockers unless you set their fields above zero.
Data honesty: an Axiom sidecar can supply exact Global Fees Paid and lifecycle. With the public DexScreener feed, fees are explicitly estimated from reported 24h volume and public AMM pools are treated as Migrated.
Filter Performance Lab
Compare the original filters with Improved v1. Historical results are labelled as backtests; only coins stamped after activation count as forward proof.
Read-only optimization report
Explains the current result and the next shadow tests. Nothing here changes a live filter.
Saved live filter versions
New filter tests
Custom narratives
Add a narrative name and comma-separated keywords. The first enabled match labels new scans; old trade results stay unchanged.
Your custom qualification rules
Add any number of rules. Every enabled rule is applied on the next scan and appears as a precise failure reason on near misses.
Safe AI collaboration
Other AIs can improve the code through isolated, reviewed branches without touching your live state, secrets, or public server.
Rules are saved in AGENTS.md, CONTRIBUTING_AI.md, and SECURITY.md. No remote shell or arbitrary AI editing endpoint is installed.
Tweeted / KOL Watch
Confirmed recent KOL mentions when the official X API is configured, plus linked-X candidates you can verify manually without fake data.
Confirmed tweeted
Linked-X candidates to check
“Linked X” means the token has an X profile in DexScreener metadata; it does not by itself prove a KOL tweeted it.
History
Past experiments. Restarting saves every bot, trade, open position, coin graph and filter in a compressed archive, then starts fresh with $1,000 per bot. Global learning stays active.
Data cleanup
Permanently remove old records without changing your saved filters. Each button requires two confirmations.
“All historical data” clears qualified coins, near misses, open positions, trades, archives and learned outcomes. Your manual filter settings and custom filter rules remain saved.
ORIGINAL PAPERTRADE · SEPARATE FAKE ACCOUNTS
LOADING 30 BOTS
30-Bot Portfolio
See every bot in one place. Each bot starts with its own $1,000 fake balance and follows a different exit strategy.
Loading all 30 bot accounts…