Experiments
Everything we've tried, including the failures. Real results only. Status: live testing retired idea
Sniper bot on Base (real money) live
Scans new Base token launches, runs each contract through our risk scanner, and buys only what passes. $30 per trade (early trades $10), exits at +10% / −5%, one position at a time, approvals revoked after every sell.
- Dates (ET)
- Oct 24 – now
- Trades
- 38 closed
- Win %
- 29% (11 W / 27 L)
- Net
- −$49.39 (−$1.30/trade, after all gas)
Verdict: Losing so far. Two rugs/honeypots cost $40.02; without the rug losses, trading was roughly +$9.7. The scanner got stricter after each loss.
Sniper bot, paper mode first run retired
Same sniper, simulated $10 buys before going live.
- Dates (ET)
- Oct 24, 5:54–6:24 PM
- Trades
- 1
- Win %
- 0%
- Net
- −$9.04
Verdict: The one paper buy (SIFI) was rugged by a liquidity pull 25 minutes in. Too short to judge; led to the liquidity-pull watcher.
Micro-scalps on Base retired
Tiny $10 swaps on WETH, cbBTC and EURC trying to catch small moves. Three variants: momentum at 0.10%, momentum at 0.25%, and EURC mean-reversion. Real pool fees, slippage and gas counted.
- Dates (ET)
- Oct 26 1:54 PM – Oct 27 4:21 PM
- Trades
- 61 (36 / 21 / 4)
- Win %
- 31% (28% / 43% / 0%)
- Net
- −$0.26 (−$0.29 / +$0.11 / −$0.08)
Verdict: No edge: costs ($0.69 total) ate the moves, and no trade ever hit its target or stop. 4 positions were left open when it stopped.
“cb scalp” flat rules (backtest) retired
Fixed exits on Base coins: +2% take profit, −3% stop, 6-hour max hold, $30 trades, 1.2% assumed round-trip cost.
- Dates (ET)
- 183 days of history
- Trades
- 86
- Win %
- 30%
- Net
- −$45.24 (−$0.53/trade)
Verdict: Loser.
“cb scalp” flat rules (live paper) retired
Same flat rules, run forward on live prices.
- Dates (ET)
- Oct 26 4:27 PM – Oct 28 1:55 PM
- Trades
- 132
- Win %
- 40%
- Net
- −$42.27 (−$0.32/trade)
Verdict: Matches its backtest. The flat +2% / −3% rule is retired.
Adaptive exits (backtest) retired
Per-coin targets based on trend and volatility, up to 72-hour hold, $30, 1.2% assumed cost.
- Dates (ET)
- 183 days of history
- Trades
- 93
- Win %
- 38%
- Net
- −$20.69 (−$0.22/trade)
Verdict: Loses less than flat rules, still loses.
Adaptive exits (live paper) testing
Same adaptive rules, run forward on live prices.
- Dates (ET)
- Oct 26 5:42 PM – Oct 28 5:25 PM
- Trades
- 48 closed
- Win %
- 38%
- Net
- +$12.35 (+$0.26/trade)
Verdict: The only live paper winner, but it ran 2 days, its own backtest lost money, and 28 positions were never closed (not counted). Needs a longer rerun before we trust it.
Daily “Combo mix” and futures paper accounts retired
A $100 mixed stock/crypto breakout account and a $781.58 long-only ETH+SOL futures account, simulated daily.
- Dates (ET)
- Oct 22 – Oct 27
- Trades
- 0 closed (1 open AAPL)
- Win %
- —
- Net
- −$0.13 (open) / $0.00
Verdict: Barely traded and stopped updating after Oct 27. No conclusion.
Kalshi “40¢” 15-minute up/down (backtest, all strategies) retired
Idea from a viral post: buy BTC/ETH 15-minute up-or-down contracts priced around 35–50¢. Four strategies (base, late-window, momentum, fade), $1 per bet, Kalshi fees included.
- Dates (ET)
- Historic 15-minute markets
- Trades
- 3,309 bets
- Win %
- 40–45% (break-even ~44–45%)
- Net
- −$133.14 (sum of 4 strategies)
Verdict: No overall edge. Only late-window BTC beat break-even.
Kalshi late-window BTC (paper) testing
The one variant that looked positive in the backtest: 245 bets, 47.8% won vs 44.1% break-even, +$21.06 (+8.6¢ per $1). Since Oct 30, 3:26 PM it is the only strategy the paper tracker runs.
- Dates (ET)
- Oct 30 – now
- Trades
- 4 settled so far
- Win %
- 25% (1 W / 3 L)
- Net
- −$1.48
Verdict: Far too few bets to say anything yet. We want 100+ before any real money.
Kalshi paper, all strategies retired
Live paper bets on all four strategies before we narrowed to late-window BTC.
- Dates (ET)
- Oct 30, 2:00–3:40 PM
- Trades
- 35
- Win %
- 37% (13 W / 22 L)
- Net
- −$4.65 (−$0.13/bet)
Verdict: Too small a sample. Replaced by the late-window BTC-only test.
Live paper bets
Contract risk scanner live
Every new token is checked before the bot will touch it: tax ≤ 5%, a real round-trip test sell, honeypot and anti-sniper traps, insider/creator dumps, liquidity size and drops, drainer contracts, copycat names, verified source, holder concentration, v4 hook safety.
- Dates (ET)
- Oct 24 – now
- Trades
- —
- Win %
- —
- Net
- —
Verdict: Running. Most new tokens are rejected.
Scam-defense research, Phase 1 live
Defensive only: went back through our own logs to see which checks caught bad tokens and which missed.
- Dates (ET)
- Oct 24 – Oct 28
- Trades
- 40 tokens flagged
- Win %
- —
- Net
- —
Verdict: The insider-dump guard worked 8 out of 9 times. It missed 4 tokens (SIFI, AOP, UDTF, PANGEA), and those misses became new checks. 4 gaps are written up as proposed fixes.
Binance hack tracing live
Following the stolen funds across chains on-chain and posting verified updates.
- Dates (ET)
- Oct 29 – now
- Trades
- —
- Win %
- —
- Net
- —
Verdict: Ongoing. See Investigations.
Paper and backtest numbers are simulations, not real money. Only the Base sniper trades with real funds. NFA 🦉