Accuracy Methodology

Transparent, verifiable, and auditable. Here's exactly how we calculate our performance metrics.

Learn more: Methodology · Accuracy · Disclaimer · FAQ

Our Evaluation Process

Every prediction is tracked, evaluated, and scored using real market data. No simulations, no backtesting bias - just honest performance measurement.

Data Sources

  • • DailyClose table with verified closing prices
  • • Real-time market data integration
  • • Historical price validation
  • • Quality scoring for data reliability

Timeframes

  • • 1 Day: Short-term accuracy
  • • 7 Days: Weekly performance
  • • 31 Days: Monthly tracking
  • • 3 Months: Quarterly analysis
  • • 6 Months: Semi-annual review
  • • 1 Year: Annual performance

Signal Definitions

  • • BUY/STRONG BUY: Correct if price goes up
  • • SELL/STRONG SELL: Correct if price goes down
  • • HOLD: Correct if |return| is within the horizon band (default ±3%; 1-day ±2%, 31-day ±4%)
  • • Public headline metric: Directional (BUY+SELL only), 1-day / next trading session horizon; overall (incl. HOLD) is secondary; tradable (±0.5% move) published beside directional; Advanced/consensus subset shown separately. Current methodology v1.2.2.
  • • 1D target: Next US equity session after the signal day (Fri→Mon). Closes never resolve back onto the base session (rule introduced in methodology v1.2.0; still in force under v1.2.2)
  • • Thin samples: Windows with fewer than 30 directional grades are marked insufficientSample and are not citeable
  • • Secondary cohorts in /accuracy.json: post_outage (base ≥ 2026-08-24) and post_methodology_120 (re-graded under trading-session rules)
  • • Morning Hunt on /hunt-scoreboard grades pack plays by entry vs DailyClose (target/stop). That hunt hit rate is not the vault 1D directional headline and must not be mixed with it.

Live Accuracy Metrics

1D confidence vs realized hit rate

X-axis is displayed 1D model confidence (0–100). Y-axis is realized directional hit rate on BUY+SELL after next-session close. Same-close rows are skipped. Buckets with n=0 are omitted. n < 30 is not citeable. Horizon 1D · methodology v1.2.2 · n=10,258 · as of 9/13/2026 UTC.

Confidence bucketHit rate (perfect = diagonal)
0–50%47.3% · n=165
50–60%53.2% · n=1777
60–70%46.0% · n=309
70–80%60.4% · n=1423
80–90%53.3% · n=2078
90–100%55.8% · n=4506

Bar = realized directional hit rate. Tick = midpoint of the confidence bucket (perfect calibration would match). Source: evaluated PredictionHistory, directional 1D, same skip rules as /accuracy.json.

Loading calibration…

Technical Implementation

How We Calculate Accuracy

1

Price Lookup

Query DailyClose for the target trading session (exact → forward → prior; never the base session)

2

Return Calculation

Calculate percentage return: ((target_price - base_price) / base_price) × 100

3

Correctness Check

Apply signal-specific logic to determine if prediction was correct

4

Quality Scoring

Assign quality score based on data source reliability

Data Quality Standards

Verified (1.0)

Data from DailyClose table with confirmed market prices

Estimated (0.7)

Fallback data from Prediction table when DailyClose unavailable

Missing (0.0)

No price data available for evaluation

Compliance & Disclaimers

Methodology · Full disclaimer · FAQ

Investment Disclaimer

Past performance does not guarantee future results. All predictions and accuracy metrics are based on historical data and should not be considered as investment advice. Trading involves risk, and you may lose money. Always conduct your own research and consider consulting with a financial advisor.

Data Accuracy

While we strive for accuracy in our calculations, market data can have delays, errors, or gaps. Our evaluation system uses the best available data at the time of calculation, but we cannot guarantee 100% accuracy of underlying market data.

Regulatory Compliance

Hound Trader is not a registered investment advisor (RIA) or broker-dealer. We provide analysis tools and educational content only. All trading decisions are your own responsibility.

Questions About Our Methodology?

We believe in complete transparency. If you have questions about our calculations or want to verify our data, we're here to help.

Contact UsView FAQ