Forward trades
—
tracking
Today's R—
Open alerts—
Regime—
Next event—
Live record—
The AI deskpaper · improving in public
You're looking at a drawdown.
Good — you're seeing the desk at its least flattering, which is rather the point. The record is — from its peak. The analytics below show what previous drawdowns looked like and how long they took to resolve. Nothing here will be edited if this one resolves badly either.
inspect the drawdown record →Win rate
—
decided trades
Total R
—
cumulative
Combine sim — full record
—
all paper trades, incl. costs — not the live attempt
TODAY
Session map
—00:00 ET06:0012:0018:0024:00
Today's daily bias
loading live reads…Live intraday setups
refreshes every few minutes during the killzoneMarket pulse
updating… the desk's model reads the tape, ≤2×/hrMarket read updates every ~30 min.
High-impact news
the desk stands aside ±30 min around theseCharts
15-minute candles · previous-day levels & alert lines drawn by the desk · —chart data unavailable — retrying on next refresh
PDH / PDL
alert entry
stop
target
killzone hours
news stand-aside
Live alerts
the trades the desk told you to take — logged at send, resolved automatically| Time (ET) | Symbol | Side | Setup | Contract | Qty | Entry | Stop | Target | Status | R |
|---|
No alerts fired yet — they log here the moment the desk sends one.
Eval attempt #1
the LIVE alerts above, run through Apex $50k EOD rules — not the full-record sim in the cards up top
IN PROGRESS
$50,000 / $53,000 target
floor
← drawdown floor · profit target →
target
THE RECORD
Daily recap
end-of-day desk note — what traded today & whyThe first recap generates after the close (~16:15 ET).
Forward test
—No forward trades logged yet.
| Date | Symbol | Dir | Entry | Stop | Target | Outcome | R |
|---|
Performance breakdown
forward-test record by setup, instrument & dayBy setup
By instrument
Highlights
Analytics
distribution, drawdown & month-by-month consistency — forward recordR distribution
Drawdown from peak
By month
| Month | Trades | Win% | R |
|---|
Daily results
each weekday's R — greener is better, redder is worse−R
+R
The desk improves
an AI desk runs its own research program — no idea touches the live book without surviving out-of-sample validationSHIPPEDFVG continuation setupheld in both halves of a six-year split and in the live-era record
SHIPPEDEntry-thrust conviction tierhigh-momentum setups earn priority under the correlation cap — survived validation
REJECTEDSweep gating (FVG-only book)2026-07 · FVG-only looked better everywhere, but sweep stayed positive in the 2018–21 train window — the pre-registered rule says keep the full book
REJECTEDOrder blocksdidn't survive out-of-sample testing
REJECTEDMomentum trend-alignment filterdidn't survive out-of-sample testing
REJECTEDLow-volatility filterdidn't survive out-of-sample testing
REJECTEDMultiple trades per sessiondidn't survive out-of-sample testing
Most ideas die in testing — that's rather the point. What ships is what survives, and the funerals are published too. how the desk works →
ARCHIVE
Briefing archive
every morning desk note, as written — nothing edited after the factNo briefings archived yet — the first one lands ~08:00 ET.