In a study of day traders on the Taiwan stock market from 1992 to 2006, Barber, Lee, Liu and Odean found that less than 1% of day traders predictably and reliably earned positive abnormal returns net of fees.
Barber, Lee, Liu and Odean (2014), The cross-section of speculator skill: Evidence from day trading, Journal of Financial Markets 18(1)A trading journal that proves which rules make you money.
Your Bitget fills, your rules, tested on trades they never saw.
Holdout never places orders. The trader makes every decision.
After a losing trade, wait 3 minutes before opening another.
Bitget Demo bot R
- Times broken
- 115
- of 171 round trips
- Effect on net P&L
- -3.79
- 90% interval -4.07 to -3.51 USDT
Holdout never places orders
The product has no code path that does.
CSV stays in your browser
Parsed in a Web Worker on your device.
Read-only keys
A key with trade or withdraw permission is refused when Bitget exposes it.
Every number recomputed by pnpm verify
Offline, no keys, PASS or FAIL for each result.
Traders repeat expensive habits. Nothing proves which rule fixes them.
Bitget reported more than 100,000 users trading rTokens in July 2026, and that 42.87% of first-time rToken buyers increased their positions within seven days.
Bitget, July 2026 transparency report (published August 10, 2026)From fills to a tested rulebook in five steps.
- 01
Import
Drop a Bitget CSV or sync with a read-only key. Parsing happens in your browser.
- 02
Say your rules
Write them in plain words. Holdout shows the compiled rule back so you can confirm it.
- 03
See what breaking them cost
Every break is listed with its trade ids, the dollar cost and an interval.
- 04
Test proposals on unseen weeks
A rule is chosen on your earlier trades and tested once on your later ones, against random rules of the same size.
- 05
Check the next trade
Describe an idea and get one card with the rules it breaks. You decide.
Four things you can check.
A ledger that reconciles
Round trips built from your fills and compared with the exchange figure. A ledger that fails the check stops proposals.
A price on every break
Each stated rule shows how often it was broken and what those trades cost, with a 90% interval.
A verdict on unseen trades
One frozen rule, tested once on the held-out third, against a random sham. Null results are shown.
A rulebook file with a hash
Export the rulebook as JSON with its sha256, then hand it to your agent through the Skill.
Five tools on one ledger.
Audit
Stated rules against your real fills, with the cost of each break.
Proposals
New rules tested on trades they never saw, kept only if they beat random.
Session lens
Profit and loss by US session, across the open and the weekend, stock perp against rToken.
Check a trade
A pre-trade card from your rulebook, session state and market tags.
Skill for your agent
Your rulebook inside Claude Code, Codex, OpenClaw or Cursor, read only.