Smart scheduler refreshes open/priority instruments only; closed markets reuse their last stored price.
RISK ENGINE
Capital currently at risk€0
Trailing stop10–15%
Cash reserve100%
Position policyNO FIXED CAP
Experiment ends31 AUG
INTELLIGENCE RADAR
Analyst actions + verified filings + selected social signals • whole-market discovery • never an automatic BUY
CONNECTED TO DB
Congress data: Bargo
• derived from official House & Senate STOCK Act filings
Analyst Intelligence readyListening layer for upgrades, downgrades, initiations, target changes and high-conviction calls across the monitored universe.
ANALYST SIGNAL SCOREAnalyst action + price acceleration + abnormal volume + relative strength = priority investigation. Multiple same-direction calls increase conviction. UBER 10 AUG is the benchmark detection case.
VERIFIED INVESTIGATE STRONG SIGNAL UNVERIFIED
EXPERIMENT10 → 31 AUG 2026
Day 1 of 22Benchmark: S&P 500
MOMENTUM LAB
500–1,000 STOCK MONITORING PROJECT
Separate research engine for short-duration momentum moves, historical testing and low-profit / high-frequency paper strategies.
UNIVERSE500–1,000stocks monitored
MODEPAPERno live execution
OBJECTIVESMALL EDGESrepeatable, measured
STATUSBUILD PHASEbacktest engine next
ALPACA CONNECTION
Checking secure paper connection…
CHECKING
Read-only Netlify function • paper account only • no order endpoint implemented.
ACCOUNT EQUITY—
BUYING POWER—
US MARKET—
NEXT OPEN/CLOSE—
HISTORICAL BACKTEST ENGINE • V0.2
Single-stock minute test
Foundation test before we scale to the 500–1,000 stock universe. Fetches Alpaca 1-minute bars and simulates a simple momentum entry with target, stop and time exit.
Ready when you are.
DISCOVERY ENGINE • V0.4
Momentum vs Reversal Lab
Tests two competing ideas on the same historical minute data: quality-filtered momentum continuation versus short-term reversal after a sharp move. The data gets to choose which hypothesis deserves further work.
Each side tests 81 exit/trigger combinations. Quality gate: ≥1.5× recent IEX volume, VWAP context, recent bar direction/acceleration, and QQQ market context. 162 strategy variants total.
Ready. Default run = 25 stocks × 20 trading days = up to 500 stock-days, plus QQQ context. It compares continuation and reversal on exactly the same sample.
Results will show the best Momentum and Reversal hypotheses side-by-side after estimated 0.08% round-trip friction.
REGIME LAB • V0.5
12-Month Market Regime Test
Answers the question: were the recent losses simply a bad market period? It freezes the best rules from the last 20-day test, then runs those exact rules across a much wider year and splits the results by month and by QQQ market regime.
QUALITY MOMENTUMTrigger +0.50% • Target +0.60% • Stop -0.20% • Hold 5 min
REVERSAL / FADETrigger +0.50% • Target +0.40% • Stop -0.20% • Hold 5 min
Regime definitions: QQQ day ≥ +0.35% = UP, ≤ -0.35% = DOWN, otherwise SIDEWAYS. QQQ intraday range ≥ 1.50% = HIGH VOL; otherwise LOW VOL. Default run uses 15 stocks × 12 months and stays under the free-data request rate by pacing calls.
Ready. This is a fixed-rule validation-style test: it does not re-optimise the rules to make the year look better.
Results will show overall performance, each month, and UP / DOWN / SIDEWAYS / HIGH VOL / LOW VOL conditions.
SELECTIVE EDGE LAB • V0.6
Exceptional Signals + Cost Sensitivity
Changes the question from “did a stock move?” to “is this one of the strongest, most unusual moves available right now?” It requires stronger relative volume and acceleration, limits how many trades we take each day, and shows exactly how much transaction friction changes the result.
Fixed exit for this diagnostic: target +0.40% • stop -0.20% • max hold 5 minutes. We compare 1 / 3 / 5 trade slots per day and friction at 0 / 0.02 / 0.04 / 0.08%. Signals are ranked only among stocks visible at that minute — no end-of-day hindsight.
Ready. Default = 15 stocks × 6 months. If anything survives 0.08% friction, we will widen the universe and validate it on unseen dates.
Results will show whether being more selective helps, and the exact friction level at which any apparent edge disappears.
TRADE AUTOPSY LAB • V0.7
Dissect the STRONG / 1-Slot Edge
Stops searching for new rules and instead takes the most promising gross-edge hypothesis from V0.6 — STRONG signals with one daily slot — and asks exactly where its 98-ish trades came from. It breaks the trades down by stock, time of day, weekday, signal quality and market backdrop.
FIXED SIGNALSTRONG • one trade slot/day
FIXED EXIT+0.40% target • -0.20% stop • 5 min
QUESTIONWhere does the gross edge actually live?
This is an autopsy, not optimisation. The STRONG gate and exit rules are frozen. We report raw gross edge first, then show what 2bp / 4bp / 8bp friction would do. The purpose is to identify concentrations and weaknesses before changing any rule.
Ready. Default = the same 15 stocks × 6 months used in V0.6.
Results will show which stocks, hours, weekdays and signal-strength bands generated or destroyed the apparent gross edge.
OUT-OF-SAMPLE VALIDATION LAB • V0.8
Can the Focused Edge Survive Unseen Data?
Freezes the clue found by the Trade Autopsy and tests it on a different six-month period and a different group of stocks. No threshold is changed after seeing the result.
BASELINESTRONG signal • 1 slot/day • all qualifying times/moves
FOCUSED CANDIDATESTRONG • 10:30–11:30 New York • 0.45–0.55% 3-min move • 1 slot/day
FIXED EXIT+0.40% target • -0.20% stop • 5 min • friction shown at 0/2/4/8bp
Validation period defaults to the six months BEFORE the Mar–Aug 2026 discovery/autopsy window. Tickers are also different from the 15-stock autopsy universe. The focused rule is frozen before this run: no optimisation, no peeking, no changing the rule because we dislike the answer.
Ready. The important result is whether the Focused Candidate still has positive average gross edge and whether it survives 2/4/8bp friction on held-out data.
Results will compare Baseline vs Focused overall, month by month, by market regime and by stock.
LARGE-SCALE OUT-OF-SAMPLE LAB • V0.9
Does the Weak Baseline Edge Persist at Scale?
Runs the same frozen Baseline and Focused rules on a much larger, earlier held-out sample. No rule changes, no optimisation and no choosing the winning stocks after the fact.
BASELINESTRONG signal • first qualifying opportunity • 1 slot/day
FOCUSEDSTRONG • 10:30–11:30 New York • 0.45–0.55% 3-min move • 1 slot/day
EVIDENCE TARGET40 new stocks • 12 months • hundreds of trades if the signals are real
Default window ends 28 Feb 2025 — before every dataset used to discover the focused rule. The engine deliberately keeps the same +0.40% target / -0.20% stop / 5-minute exit and the same STRONG filters. Expect a few minutes: progress is shown month-by-month and stock-by-stock.
Ready • 40 held-out stocks × 12 months. Nothing is tuned from the result.
We are looking for persistence: enough trades, positive gross average, survival at realistic friction, and broad contribution across months/stocks rather than one lucky pocket.
EDIT WATCHLIST
Add stock
Ticker is validated with Alpaca, then saved in Supabase. No redeploy needed.
PAPER TRADE
Add position
Saves directly to Supabase using the original trade time. No redeploy needed for future entries.