Transparent & Data-Driven

Signal Performance & Methodology

We believe in radical transparency. Every confidence score is backed by real backtested data. See exactly how our signals perform across different sources, confidence levels, and market conditions.

Alpha Opportunities Snapshot-Rank Backtest

Hypothetical $10,000 model on daily alpha_opportunities_snapshot ranks (confidence ≥ 55, same floor as mobile). Top 10 and Top 20 long/short sleeves use non-overlapping 3-trading-day holds: enter next open, exit at close.

Point-in-time snapshot rank per direction (Top 10 / Top 20)
Equal-weight long and short legs; no overlap while a hold is open
No commissions, borrow, slippage, or liquidity constraints

Source

Daily snapshot rows ranked within each direction. Excludes optional web-only filters (60% source win rate, market-cap tiers).

Method

Next trading day open → close after 3 trading days. Illustration only; not investment advice.

Past simulated performance does not guarantee future results. Large single-window moves may reflect thin or corporate-action price prints.

Largest 3-day snapshot holds (Top 10 sleeve)

Single-name legs from non-overlapping windows: snapshot rank, enter next open, exit close after 3 trading days.

Symbol	Side	Rank	Snapshot	Hold	Return

One volatile name can dominate a window; ranks are as-of snapshot date.

50.0%

Raw Win Rate

0.11%

Raw Avg Return

238,049

Backtested Signals

Evaluation Horizon

Raw win rate, average return, and backtested signal count above include only resolved signals for the top 500 symbols by market capitalization in our universe (symbols with a known market cap in our database)—not every name we have ever signaled.

Alpha Performance (vs SPY)

The naked truth: performance after removing market movement

0.11%

Avg Alpha (Excess Return)

49.8%

Beat SPY Rate

0.16%

SPY Avg Return (Same Period)

What is Alpha?

Alpha measures how much our signals beat (or lag) the market. A +1% alpha means signals returned 1% more than SPY over the same period. This strips out market beta—in a +5% SPY rally, a signal returning +6% has +1% alpha. This is the true measure of signal value.

Backtested Performance

Every signal type is tracked against actual market outcomes to calculate real win rates and returns.

Adaptive Learning

Confidence scores adjust automatically as new market data comes in. Poor performers get downweighted.

Full Transparency

No black boxes. See sample sizes, win rates, and methodology for every signal source and type.

Signal Source Reliability

Performance rankings by data source

Short Interest Intel

4 signal types

76.5%

0.28%

0.10%

50.3%

136

Tail Risk

1 signal types

71.4%

5.79%

0.56%

55.4%

Technical Analysis

2 signal types

56.3%

1.31%

4.88%

66.0%

5,014

Insider Trading

3 signal types

54.6%

1.13%

0.68%

54.2%

2,159

Market Events

1 signal types

54.5%

0.81%

-0.40%

46.6%

4,939

Sentiment

2 signal types

53.9%

0.90%

-0.09%

49.9%

343

FTD Intel

6 signal types

53.3%

0.27%

-0.16%

48.0%

17,038

Fundamental Analysis

5 signal types

51.6%

0.20%

-0.17%

48.4%

1,037

AI Deep Analysis

8 signal types

51.3%

-0.02%

0.62%

53.4%

1,262

Statistical Extremes

15 signal types

51.0%

0.53%

0.88%

50.6%

18,211

Earnings Drift

3 signal types

50.9%

0.18%

-0.03%

49.3%

230

Congressional Trading

2 signal types

50.3%

0.01%

-0.02%

50.3%

84,738

ML Forecasts

2 signal types

50.3%

-0.05%

0.04%

49.7%

7,659

Fundamentals

3 signal types

50.2%

0.32%

0.21%

51.1%

616

Cognitive Load

1 signal types

49.5%

0.08%

-0.04%

49.5%

32,450

Macro Events

13 signal types

49.4%

-0.11%

-0.70%

45.5%

6,323

Institutional Holdings

5 signal types

48.8%

0.06%

-0.01%

48.8%

11,561

Warn

2 signal types

48.6%

-0.53%

-0.63%

50.9%

109

Dark Pool Activity

5 signal types

48.5%

-0.03%

49.7%

66,872

Options Flow

1 signal types

47.6%

0.00%

0.0%

Short Volume Intel

4 signal types

47.4%

-0.85%

-0.22%

48.3%

753

Form 8-K Material Events

9 signal types

47.1%

-0.35%

-0.20%

47.9%

1,712

Options Flow

3 signal types

45.9%

-1.16%

-0.88%

43.4%

233

Top Performing Signal Types

Win rate > 60% with at least 20 samples

Form 8-K: CEO Departure

Form 8-K Material Events

Bearish

                    92.0%
                

High

→

25

Momentum Z-Score: Strong Up

Statistical Extremes

Bullish

                    91.4%
                

High

→

526

Volume Z-Score Surge

Statistical Extremes

Bullish

                    89.9%
                

High

→

405

Momentum Z-Score: Strong Down

Statistical Extremes

Bearish

                    89.9%
                

High

→

379

Volume Z-Score Surge

Statistical Extremes

Bearish

                    89.2%
                

High

→

242

Revenue Deceleration

Unknown

Bearish

                    84.2%
                

High

→

38

Gap Up

Technical Analysis

Bullish

                    80.6%
                

High

→

346

Gap Down

Technical Analysis

Bearish

                    79.9%
                

High

→

299

Short Interest: Extreme DTC

Short Interest Intel

Bullish

                    76.5%
                

High

→

136

RSI Z-Score: Overbought

Statistical Extremes

Bearish

                    74.7%
                

High

→

79

How We Calculate Confidence (0-100 Scale)

Our 3-step formula for signal scoring

Base Confidence

We calculate a weighted average of all agreeing signals:

Base = Σ(Win Rate × Weight) / Σ(Weight)

Example: Congress (55% win rate, 0.70 weight) + Dark Pool (65%, 0.78) = 61 base

Confluence Boost

Multiple independent signals get a bonus:

2 signals: +3 points
3 signals: +7 points
4+ signals: +10 points

Diversity Bonus (0-8 points)

Diverse signal types and sources get additional boost:

Type diversity: +0 to +5 points
Source diversity: +0 to +3 points

Example: 4 different signal types from 3 sources = +8 diversity bonus

Final Score

Base + Confluence + Diversity = Final Confidence (0-100)

Example: 61 + 7 + 8 = 76 (High Confidence)

Confidence Tiers

75-100: High Confidence

65-74: Medium Confidence

55-64: Low Confidence

Below 55: Not displayed publicly

Our Signals Learn From Performance

Adaptive weighting based on recent performance

Unlike static systems, our signal weights adapt based on recent performance:

Example: Dark Pool Signals

All-Time: 65% win rate → 0.78 weight

Last 30 Days: 72% win rate → 0.82 weight ↑

Recent strong performance increases the signal's influence on future confidence scores.

Weight Calculation

30-Day Win Rate	Weight
≥70%	0.90 (Maximum influence)
65-69%	0.80 (High influence)
60-64%	0.70 (Good influence)
55-59%	0.60 (Moderate influence)
50-54%	0.50 (Low influence)
<50%	0.40 (Minimal influence)