Every pick gets a Quality Score the moment we surface it. The score is built on three things: how locked-in the chokepoint actually is, how much real demand is pointing at it, and where the company sits in our supply chain graph — a map of 4,300+ companies and 19,800+ supply relationships we've built node by node. The graph is the part that's hard to fake. It's what tells us a Korean specialty chemicals maker is structurally exposed to HBM memory demand, or that a small Taiwanese fiber assembly shop sits on the only path to TSMC's next-gen optics. Higher score, higher conviction. Then we just watch what the picks do. So far, the high-score names have a — win rate. The low-score names have lagged. Same picks, same scoring, no rewriting after the fact.
Win Rate by Score Bucket
Loading...
Each bar shows the win rate for picks within that score range only — not cumulative. The top bucket (8+) shows where the scoring system separates highest-conviction picks from the rest. Sample size is still growing.
—
Top vs. Bottom
Return gap between best and worst scores
—
Supply Chain Boost
Extra return from mapped companies
—
Gain / Loss Ratio
Dollars won per dollar lost, top tier
Performance Breakdown
—
Combined Win Rate · — avg · — picks
▶ Show realized vs unrealized breakdown
Realized (closed)
Win Rate—
Avg Return—
Positions—
Unrealized (open)
Win Rate—
Avg Return—
Positions—
Realized returns skew negative because stop-losses fire early on losers while winners continue to run. This is by design — the system cuts losses fast. Unrealized positions are still open.
Track record still building?
Get monthly performance updates as the sample grows. We'll show you the numbers as they develop — wins, losses, everything.
Free. Monthly. Unsubscribe anytime.
Full Score Breakdown
Discrete score buckets. Each row shows only picks within that range — not cumulative. Shows where the scoring separates winners from losers.
Score Range
Picks
Win Rate
Median Return
Avg Return
Profit Factor
How the Tiers Work
The score ranks picks into three tiers. Runner is the primary product — the highest-conviction picks with supply chain confirmation.
Runner (Score 7+)
Highest confidence. Multiple sources agree AND the company sits at a critical supply chain position. Stop-loss at –20%.
Watchlist (Score 4–6)
The system tracks these but they lack full conviction — fewer sources or a less critical supply chain role. Not a primary recommendation. Stop-loss at –15%.
Avoid (Score 0 or below)
The system flags these as low quality. We track them to prove the scoring works — and they consistently lose money.
Losses shown are realised exits. When a pick hits its stop-loss, we close it and show the actual exit price. The stock may keep falling after — that continued decline is shown as context, not as our loss. Avoid-tier stocks that fell the furthest? The system flagged them as low quality before they dropped. Watchlist picks (formerly “Core”) are tracked but not primary recommendations. The "With Stops" column in the score table above shows aggregate returns with risk limits applied.
Best and Worst Picks
Biggest wins and biggest losses since February 2026. Losses show the realised exit return (where the stop triggered), with post-exit stock movement shown as context.
Best Picks
Worst Picks (Realised Exit Returns)
Runner Tier Performance
Highest-conviction picks identified by the V4 quality model
—
PICKS
—
WIN RATE
—
AVG RETURN
—
W/L RATIO
—
BEST PICK
—
WORST LOSS
Loading data...
Detected
Ticker
Tier
QS
Sev
Thesis
Src
Return
Peak
Status
Catalyst Watch
Structural chokepoints with converging near-term catalysts. Updated daily.
Loading catalyst data...
Supply Chain Node
Structural chokepoint with active catalyst convergence
SEV 4
Burst active
Events: 3+
Full catalyst watch rankings are available to Pro subscribers.
Past performance does not guarantee future results. All investments involve risk including loss of principal. Returns shown are from detection date to current price and do not represent actual trades. This is not financial advice.
How It Works
What we scan: 34 independent data sources — congressional trades, hedge fund filings (13F), options flow, short interest, defense contracts, lobbying spend, prediction markets, technical analysis, and more. Each source is checked daily.
What triggers a pick: When two or more unrelated sources independently flag the same stock within a rolling window, the system scores it. More sources agreeing = higher score.
The supply chain layer: We built a map of 2,000+ companies and the materials they depend on. If a flagged stock also sits at a critical point in this map (e.g., it makes something that can't be easily replaced), the score goes up further.
Graph-only picks: Some stocks are flagged purely by their position in the supply chain map — they control a critical material or process with no easy substitute. These don't need multiple scanner sources to qualify. Graph-sourced picks have outperformed scanner-sourced picks consistently.
Quality Score (0–10): Combines source count, how critical the company is in the supply chain, how big the opportunity is relative to company size, and sector relevance. Stocks are grouped into tiers: Runner (score 7+), Core (4–6), and Avoid (3 or below). Convergence-only picks must also be in the supply chain graph, have a quality score of 5+, or have 4+ independent sources — otherwise they are automatically demoted to Avoid.
Entry price: Closing price on the day the system first flags the stock. For picks identified before our live tracker launched (March 20, 2026), entry prices are sourced from historical market data at the original flag date. These "historically-sourced" picks are shown separately from live-tracked picks.
Returns: Measured from entry price to current price. Updated daily. Open positions shown as-is — nothing is cherry-picked or removed.
Risk management: Each tier has an automatic stop-loss: Runner –20%, Core –15%. High-quality runners (score 7+) get wider trailing stops to avoid being shaken out of strong moves. All parameters are self-calibrating — the system tracks every exit, observes what happens afterward, and adjusts stop widths based on real outcomes. If stops are too tight (stocks keep recovering after being stopped out), they widen automatically.
Deduplication: One pick per stock per 30-day window, so the same stock can't inflate the numbers by being counted multiple times.
No trading costs modeled: Returns don't include commissions or slippage. Real-world returns would be somewhat lower.
What We Show Honestly
Small sample: The live system has been running since February 2026 — less than three months. The Runner tier has — picks with a — win rate, which is promising but not yet statistically bulletproof. We show everything so you can watch the track record build.
One market regime: This period has been mostly sideways-to-down (S&P 500 fell ~4%). We haven't yet seen how the system performs in a strong bull market or a crash. That data will come with time.
Survivorship bias: We only track currently listed stocks. Companies that got delisted or acquired aren't in the dataset, which may slightly inflate results.
Disclosure delays: Some data sources (like congressional trade disclosures) can be filed weeks after the actual trade. We measure from disclosure date, not trade date — so some of the move may have already happened before we see it.
The system learns: Both the scoring model and the position management rules adjust based on real outcomes. Stop-loss widths, trailing stop percentages, and re-entry thresholds are all calibrated from actual exit data using a Bayesian feedback loop — performance characteristics will shift as the system accumulates more observations.
What Would Change Our Mind
We believe the approach works. Here's what would make us reconsider:
If the top-scored picks (Runner tier) drop below a 50% win rate over any 3-month period, the scoring model may need rethinking.
If the staircase pattern breaks — meaning lower-scored picks start outperforming higher-scored ones — the quality score isn't working.
If stocks in our supply chain map stop outperforming those found by data alone, the map isn't adding value.
If the results only hold in one sector (e.g., it's really just a semiconductor momentum play), then the edge is narrower than we think.
Every pick is timestamped and tracked from the day it's flagged. If the pattern breaks down, the data on this page will show it — we can't hide it. See our full methodology page for deeper detail on how the system works.
See What the System Finds Next
The supply chain map tracks 2,000+ companies across semiconductors, defense, energy, and robotics. When a small company controls something critical and demand is growing, the system flags it.