Post-mortem · Case study · Live ongoing

Why we stopped selling trading signals

For 10 months we ran automated Kraken bots on real money. They won 71.8% of 682 individual trades — and still lost 63 percentage points against passive BTC buy-and-hold under compound math.

In compound terms — if a position had been the entire account — that win rate produced a -94.5% return against BTC's -31% in the same window. Underperformance: -63 percentage points.

In actual terms — on a $50-per-trade ceiling, $700 starting capital — the account ended at $427. That's -39% vs HODL's -31%. Real underperformance: -12 percentage points.

Both numbers are true. The first one shows what happens when asymmetric losses (-3.8% avg) eat asymmetric wins (+1.0% avg) under compound math. The second shows position-sizing saved the worst case but didn't eliminate the edge gap.

Win rate is not edge. Compound geometry is.

Swiss Whale Intelligence is now an information service, not a trading product. The bot still runs — every new trade appears below as the lab continues. The ledger refreshes each Friday with SHA256-hashed CSVs. Reproducible, downloadable, ongoing.

Period: July 2025 — present · 747+ trades · 71.8% hit rate · -63 pp compound · -12 pp real · SHA256 verified

📊 Archive Downloads

Data format: Strategy names, UTC timestamps, return percentages. Schema identical across snapshots.

Update frequency: Daily at 02:00 UTC; lifetime CSV regenerated on every archive rebuild.

Privacy: Percentage returns only, no absolute position sizes.

🌞 Complete lifetime snapshot (since July 2025):
📥 latest_all.csv 🔐 SHA256 📁 Browse Archive
🕐 90-Day Rolling Window (recent trades only):
📥 latest_90d.csv 🔐 SHA256

The full archive breaks the lifetime ledger into weekly, monthly and quarterly reports — each with its own SHA256 hash. The metadata.json index gives you the full SHA256 map.

747
Trades on real capital
72.6%
Hit rate · individual trades
-63 pp
Compound vs BTC HODL
SHA256
Integrity hash per snapshot

🔒 Integrity Verification

Each CSV snapshot ships with a SHA256 integrity hash. This detects file-level tampering after publication; it is not an on-chain anchor.

Verify command:
curl -s https://swisswhaleintelligence.com/ledger/latest_all.csv | openssl dgst -sha256
curl -s https://swisswhaleintelligence.com/ledger/latest_all.csv.sha256

Archive scope & caveats:

  • Live Kraken-bot trades from July 2025 onward, plus legacy rows from earlier experiments (flagged by strategy name).
  • UTC timestamps throughout; schema identical across weekly, monthly and quarterly snapshots.
  • Percentage returns only — no absolute position sizes (privacy).
  • We still run strategies internally with our own capital. New rows append as our generator pipeline completes each cycle; existing rows are never rewritten. The lifetime CSV is regenerated on every successful archive rebuild.

📈 What we learned

Win rate is not edge. 71.8% of trades closed in profit — and the strategy still lost 63 percentage points against passive HODL under compound geometry. The math: avg win +1.0%, avg loss -3.8%. That 1:3.8 asymmetry needs ~80% win rate to break even on compound returns. We had 72%.

Fees are the real boss. 85% of trades historically ran as Kraken takers (0.28% fee per leg = 0.56% round-trip). On +1% avg wins, that's 56% of the gross gain gone before any P&L. Maker-only orders (0.18% fee) became standard May 2026; the difference shows in fresh trades below.

The pivot. Our current product is an on-chain intelligence service — live whale feed, SOPR, HODL waves, exchange reserves, OFAC sanction labels, entity attribution. It does not trade on your behalf and does not publish buy/sell signals. The bot still runs as research lab — its trades append to this ledger weekly. Strategies are tested forwards on $400 of real capital, not on your wallet.

Last updated: 2026-07-10 18:00 UTC
Operated by CE Services, Switzerland 🇨🇭