Balance
—
Show as a table
| Time (New York) | Account value | Same trades in SPY | Same trades in QQQ | Same trades in DIA |
|---|
Strategies by day
Real round trips from the account, credited to the strategy that bought and the day it sold: profit or loss, then trades, how many made money and the return. In brackets, the backtest on the same day, with the same strategy, on the same symbols. Newest day first.
Loading…
Open positions
Today's price, with the entry (grey), take profit (green) and stop loss (red). Profit or loss here is unrealised until the position is sold, which happens before the close at the latest.
Loading…
Recent trades
Orange: buys. Green and red: sells that closed a round trip, with the realised profit or loss
- Loading…
Watching
Every strategy on every liquid symbol, best recent record first. A position opens only when one signals on the latest bar. Out-of-sample return per trade ± one standard error (number of trades).
- Loading…
Last run's orders
- Loading…
Backtest profile
Flame graph and full profileLoading…
How it works
Every rule below is shared code: the backtest that rates strategies and the live loop that trades them run the same Rust, so what was tested is what trades.
- Hourly. Sixty days of five-minute bars for 91 symbols; the liquid ones are backtested with ten strategies.
- No picking. Choosing strategies by their recent record added nothing in a walk-forward test of 98 sessions, so every strategy trades every liquid symbol, at most 20 positions at once.
- Honest delays. The data feed is fifteen minutes behind, and the backtest only acts on bars it could really have seen.
- Every five minutes. A Cloudflare Worker takes profit at 0.75%, stops a loss at 3% and buys when a strategy signals, 2% of the portfolio a time.
- Flat by the bell. Everything is sold in the last 20 minutes, using the exchange's real closing time, so nothing is held overnight.