Manual trading
same fees, slippage and exit rules as the botNo key saved — manual controls are locked.
A manual buy ignores the model's 0.60 threshold and the expected-value gate — it is your call, so it is taken regardless. It is still charged the full 0.1% fee per side plus size-dependent slippage, and still gets a volatility-scaled stop with a 1:2 target. Swing trades have no time limit; they close on a barrier or when the signal is invalidated.
Account funding
—Uses the same trigger key.
A deposit also buys the benchmark that much BTC at today's price, which is how a contribution has to be treated. Adding the cash without adjusting the benchmark would hand the new money every move BTC has made since the run started, and quietly break the one number this bot is judged by — beating buy-and-hold BTC.
Model
—- Version
- —
- Last trained
- —
- Validation accuracy
- —
- Precision at p>0.60
- —
- Training samples
- —
Open positions
| Symbol | Side | Size | Entry | Now | Unrealised | Stop / Target | Model | |
|---|---|---|---|---|---|---|---|---|
| No open positions | ||||||||
Trade history
fees 0.1% per side · dynamic slippage · volatility-scaled 1:2 · no time limit on swing, 5h on scalp| Symbol | Side | Entered | Entry | Exited | Exit | Reason | P&L | % | Model | |
|---|---|---|---|---|---|---|---|---|---|---|
| No trades yet | ||||||||||
Equity vs buy-and-hold BTC
System components
—Each tick runs as eight guarded stages. A failure is attributed to the stage that caused it and contained so it cannot take the others down — the eight-day outage happened because one failing write killed all of them at once and nothing recorded which.
Market news
—Loading…
- Loading headlines…
Bot log
updated —- No log entries yet
Train now
takes about a minuteUses the same trigger key as manual trading.
Trains a fresh model and saves it as a new version. Old versions are never overwritten, so a bad model can always be rolled back. A new model only goes live if it beats random guessing on data it has never seen.
What the model weighs
No trained model yet.
Training queue
| Symbol | Entered | Closed | Outcome | P&L | Label | Usable | |
|---|---|---|---|---|---|---|---|
| No closed trades yet | |||||||
Model versions
append-only — every version is kept for rollback| Version | Trained | Samples | Train acc | Val acc | Precision | Status |
|---|---|---|---|---|---|---|
| No models yet | ||||||
Training log
- No training entries yet
Why this book is usually quiet
—Every candidate must clear the model threshold and show positive expected value after the real round trip. On the current model the measured edge is negative, so most ticks decline every pair. A row saying "hold" is the gate working, not the scanner failing. The table below always states which test each pair failed.
Volatile universe
scanned live — not the five majors| Symbol | 24h range | 24h change | Spread | Depth | Model | Decision | Why | |
|---|---|---|---|---|---|---|---|---|
| Scanning… | ||||||||
Open scalp positions
each expires 5 hours after entry| Symbol | Side | Size | Entry | Now | P&L | Stop | Target | Expires in | |
|---|---|---|---|---|---|---|---|---|---|
| No open scalp positions | |||||||||
Force a scalp
bypasses the gatesUses the same trigger key as manual trading.
Skips the model threshold and the expected-value gate, because forcing a trade the gate refused is the whole point of a manual override. It still pays the real fee and slippage, still gets scalp-width barriers, still expires in five hours, and still respects the exposure ceiling — an override on the entry decision, not on the risk limits.
How this book is configured
Loading…
Scalp trade history
this book only — swing trades are on the Trading tab| Symbol | Side | Opened | Held | Entry | Exit | P&L | Closed by |
|---|---|---|---|---|---|---|---|
| No scalp trades yet | |||||||