Live System · March 2026

The Wave Rider

We built a transformer to read SEC filings and find hidden biotech signals. We thought we were building a moonshot finder. We actually built something better.

164 binary signals per filing
378 biotech companies
8+ years of regulatory history
6-layer transformer
82%
Win Rate
Filtered recycle windows
75 days
Median Hold Time
From entry to exit target
+62.5%
Avg Return Per Trade
Per entry window, all exits
3.9
Avg Tickers on Screen
Live opportunities at any time

The Discovery

We thought we were building moonshots.
We found something more reliable.

The original thesis: find beaten-down biotech companies with 8+ quarters of strong model conviction, buy at the low, hold until FDA approval sends the stock to the moon. One big bet, 12+ months of waiting. The backtest data told a different story.

🌊

Quality companies oscillate

The 19 companies that passed our conviction gate over 2.5 years don't just go up once and stay up. The market beats them down, the model stays bullish because the SEC filings still show strong fundamentals — and the stock eventually recovers, hits our exit target, and then the market forgets again. The wave resets.

📁

The model's conviction doesn't change

When a stock drops 50% on no fundamental news, our transformer still reads the same SEC filings it read last quarter. Cash runway unchanged. Clinical progress unchanged. Risk flags unchanged. Same quality company at a cheaper price. Fresh wave entry.

📊

11 of 19 tickers recycled

Across the out-of-sample period (July 2023 → March 2026), 11 of the 19 screened companies returned for a second entry window. Some returned 4–6 times. AQST appeared 6 separate times. Each time: beaten-down price, model still bullish. Each time: recovery.

The moonshot bonus is still there

High-conviction names at sub-$1 can still explode on FDA approvals, clinical data, partnership announcements. The waves pay us while we wait for the bigger event. If the moonshot never comes, we still made money riding the oscillations. If it does — and we're holding — that's the bonus.

The key advantage over pure technical traders: they don't know which beaten-down sub-$1 biotech stocks are worth buying. We do. 164 binary signals and 8+ years of regulatory filing history tell us which companies are genuinely strong and which are structurally declining. That's the difference between riding a real wave and drowning in a sinking ship.

System Architecture

From SEC filing to conviction signal

Every quarterly report from 378 biotech companies flows through a four-stage pipeline. No free-text NLP, no price data in the model — just binary yes/no questions about what the company actually filed with regulators.

1
SEC Filing Ingested
10-K and 10-Q reports for 378 biotech companies. 194,000+ filings, back to early 2000s. Updated nightly from EDGAR.
2
164 Primitives Extracted
Each filing is reduced to 164 binary signals — deterministic rules applied to filing text and financial tables. No NLP ambiguity.
3
Transformer Scores
A 6-layer transformer sees the full regulatory history of each company and outputs a forward-return conviction score [0–1].
4
Conviction Gate
Three conditions must all hold simultaneously. When they do, the screener fires and the wave entry window opens.
Gate Condition 1
IC ≥ 0.60
Per-Company Information Coefficient
Spearman rank correlation between the model's historical predictions for this company and its actual 365-day returns. IC ≥ 0.60 means the model has demonstrated predictive power for this specific name — not just the universe average. 211 of 378 companies pass.
Gate Condition 2
mean_q8 ≥ 0.80
8-Quarter Mean Conviction
Simple mean of the model's last 8 quarterly scores. Must be ≥ 0.80 for the gate to open. This filters out one-quarter spikes — we want sustained multi-year conviction, not a single good filing. The transformer must have believed in this company for two full years.
Gate Condition 3
Price ≤ 75% of 52wk High
Price Dip Filter
Avoid chasing stocks where the thesis has already played out. We enter while the stock is still beaten down — the quality is there in the filings, but the market hasn't repriced it yet. This is the wave entry point. The model never sees price data; this filter is applied after scoring.

The Foundation

164 binary signals per filing

Each primitive is a deterministic yes/no question applied to a company's SEC filing. No machine learning in the extraction — just hand-engineered rules applied at scale across 194,000 filings. The transformer learns which combinations of these primitives, sustained over time, predict outperformance.

General Signals
130
Cash runway > 12 months without new funding
Revenue growing year-over-year
Burn rate declining vs. prior quarter
Debt-to-equity below threshold
R&D expenses as % of total expenses
Operating expenses trending stable
Catalyst Signals
24
Phase 2 or Phase 3 trial initiated this quarter
FDA NDA or BLA submission filed
Breakthrough Therapy designation received
Strategic partnership announced
License agreement or royalty deal disclosed
Positive interim data readout mentioned
Risk Signals
10
Going-concern warning issued by auditors
Clinical hold placed by FDA
Key executive or CEO departure
Trial failure or program discontinuation
Significant legal proceedings filed
Dilutive financing at steep discount
Signal breakdown
General (130)
Catalyst (24)
Risk (10)

Case Study

AQST — The Perfect Oscillator

Aquestive Therapeutics appeared on our screener 6 separate times over 2.5 years. Six separate entry windows. Each time: beaten-down price. Each time: model still bullish on the SEC filings. Each time: recovery to exit target. This is the wave thesis in its purest form.

AQST
Aquestive Therapeutics — specialty pharma / CNS
Current price
$3.65
% of 52-week high
48.7%
IC (model accuracy)
0.554
mean_q8 (8-qtr conviction)
0.793
Total entry windows
6
Resolved at target
5 of 5 resolved ✓
Window 6 status
Active ↑
Upside to exit target
+75%

6 Entry Windows — July 2023 → March 2026

Wave
Status
Return
Hold
Wave 1
Target hit
+32.5%
~65 days
Wave 2
Target hit
+47.1%
~88 days
Wave 3
Target hit
+105.4%
~143 days
Wave 4
Target hit
+61.7%
~110 days
Wave 5
Mar 2025 — resolved
+81.4%
183 days
Wave 6
Active now
$3.65
+75% to target
Why AQST keeps returning: Aquestive's fundamentals — specialty pharma pipeline, CNS focus, regulatory progress — are consistently reflected in its SEC filings. The model reads those filings and stays bullish. But the stock repeatedly underperforms the market, drifts back below our entry threshold, and the wave opens again. The model doesn't change its mind just because the stock got cheap. That's the signal.

Recycle Filter

Not all second entries are equal

The recycle backtest revealed a critical insight: some stocks oscillate around genuine quality (AQST, 6/6 wins), while others are in structural decline and just keep generating false re-entries (SYBX: 4 attempts, 0 wins). A simple filter separates them.

📉

Without the filter

Treating every recycle window equally: 70% win rate, +45.5% average return. Structural decliners keep sneaking in. SYBX appeared 4 times after its 2023 spike — all 4 were losses averaging -61%. The model was right that SYBX is a quality company historically; the market had structurally moved on.

With the filter

Rule: only re-enter a recycle window if the previous window hit its exit target. If the last attempt failed (stock never recovered), skip the next entry. Result: 82% win rate, +62.5% average return. 7 trades blocked — those 7 had an average return of −34%.

Unfiltered
70%
win rate · 40 trades · +45.5% avg
Filtered
82%
win rate · 33 trades · +62.5% avg
7 trades blocked · avg return of blocked trades: −34.2% · Blocked: SYBX ×3, APRE ×1, LRMR ×2, SABS ×1

The Edge

What happens without the model?

We ran an ablation study — systematically removing model layers to measure what each one contributes. Same backtest period (2022–2025), same price gate, same time stop. Only the conviction filters change. The answer is not subtle.

Level 1 — No model
Price filter only  (stock ≤ 70% of 52-week high)
What anyone with a stock screener can do. No IC, no conviction, no model. 247 trades, 40% win rate, avg 9 simultaneous positions.
−5.2%
Sharpe 0.21 · DD −51%
SPY +36.7%
Level 2 — Model accuracy only
+ IC ≥ 0.50  (model has been historically accurate on this company)
Knowing the model is accurate isn't enough — you also need to know what it's saying right now. High-IC companies the model is currently bearish on are still bad buys.
−33.1%
Sharpe −0.06 · DD −63%
Worse than no model
Level 3 — Full conviction gate ★
+ mean_q8 ≥ 0.85  (model has been bullish for 8 consecutive quarters)
The mean_q8 filter adds "what the model is saying now." Only 45 trades vs 247 — the model said no to 80% of candidates. Those 80% were losing the money above.
+205.7%
Sharpe 1.44 · DD −27%
vs SPY +36.7%
The selectivity is the strategy. The full model runs 45 trades where the price-only approach runs 247. Each "no" from the conviction gate was worth an average of −1.2 percentage points that it saved you from. The model's job isn't to find opportunities — the price filter does that. The model's job is to say no to 80% of them.

Live Output

Today's screener — March 1, 2026

Four candidates currently pass all three gate conditions simultaneously. Note AQST — wave 6. The same company that has returned to target 5 previous times, now back below our entry threshold with the model still bullish.

Conviction Gate Candidates

● Live · March 1, 2026
Ticker Price % of 52wk Hi IC mean_q8 Upside to Target
ATYR $0.89 13.5% 0.698 0.897 +529%
SYBX $0.62 34.0% 0.581 0.793 +151%
AQST wave 6 ↑ $3.65 48.7% 0.554 0.793 +75%
MRKR $1.43 65.3% 0.502 0.860 +30%
Upside to Target = (52wk high × 0.85) / current price − 1. These are not price targets — they are the exit gate level. The model has no knowledge of current price data; all signals are derived from SEC filings. Price filter is applied after scoring.

What the numbers mean

Out-of-sample period

All results shown use a train/test cutoff of June 30, 2023. No data from July 2023 onward was used in model training. The screener results are fully out-of-sample.

Price data note

Backtests use end-of-day prices. Real execution on sub-$1 names with thin floats would include slippage. We use 30-day entry delays after gate opens to reduce premature entries.

Sample size caveat

19 unique tickers over 2.5 years is a narrow but curated sample. The gate is intentionally selective — not every beat-down biotech qualifies. Selectivity is a feature.

Price independence

The transformer model sees zero price data. All 164 primitives come from SEC filing text and financial tables. The price dip filter is applied after scoring — signal and entry condition are fully independent.

Recycle filter rule

For repeat entries on the same ticker: only re-enter if the previous entry window hit its exit target. This eliminates structural decliners that repeatedly fail to recover. Validated: blocks trades averaging −34%.

Exit target

Exit when price reaches 85% of the 52-week high. A conservative proxy for "thesis has resolved." Does not require predicting FDA decisions or specific catalysts — just that the market reprices the quality we identified.

Portfolio backtest validation

486-configuration parameter sweep (2022–2025, includes bear market). Best config: Sharpe 1.44, +205.7% total return, 73-day avg hold, 45 trades, 55.6% win rate. SPY baseline over same period: +36.7%. The 73-day portfolio avg hold independently confirms the 75-day per-trade median — two methods, same answer.

Two strategies, one signal

The 90-day time stop (primary strategy) maximises Sharpe at 1.44 and captures full wave oscillations. A 14-day time stop is a distinct product — high-frequency oscillator with 141 trades at 13-day avg hold and Sharpe 1.18. Same screener, one parameter change. The primary strategy is the right baseline; the 14-day config is worth separate exploration at smaller size.