Evaluation methodology and latest public snapshot

Evaluation period: 30 days (ending 2026-10-05T19:08:36.878Z). Metric version: terminal_direction_v1.

Records / verified outcomes / pending: 529 / 394 / 135
Direction metric sample N: 394
SMAPE sample N: 394
Direction consistency: 60.4%
Model reference price proximity: 94.7%
Mean SMAPE: 5.3%
Data snapshot: 2026-10-05T19:08:36.878Z

Model or version

LSTM Neural Network: Direction metric sample N 7, SMAPE sample N 7, Direction consistency 85.7%, Model reference price proximity 98.3%
Transformer: Direction metric sample N 19, SMAPE sample N 19, Direction consistency 63.2%, Model reference price proximity 95.6%
G-Prophet2026V2: Direction metric sample N 24, SMAPE sample N 24, Direction consistency 54.2%, Model reference price proximity 95.5%
G-Prophet 2026V1: Direction metric sample N 163, SMAPE sample N 163, Direction consistency 60.7%, Model reference price proximity 94.7%
Ensemble: Direction metric sample N 177, SMAPE sample N 177, Direction consistency 59.3%, Model reference price proximity 94.4%
Random Forest: Direction metric sample N 4, SMAPE sample N 4, Direction consistency 75.0%, Model reference price proximity 92.3%

Market

A-Shares: Direction metric sample N 382, SMAPE sample N 382, Direction consistency 60.2%, Model reference price proximity 94.8%
Crypto: Direction metric sample N 1, SMAPE sample N 1, Direction consistency 100.0%, Model reference price proximity 99.4%
US Stocks: Direction metric sample N 3, SMAPE sample N 3, Direction consistency 66.7%, Model reference price proximity 97.6%
HK Stocks: Direction metric sample N 8, SMAPE sample N 8, Direction consistency 62.5%, Model reference price proximity 87.8%

Horizon (days)

1 Horizon (days): Verified sample 81, Direction consistency 55.6%
3 Horizon (days): Verified sample 208, Direction consistency 57.7%
5 Horizon (days): Verified sample 2, Direction consistency 100.0%
7 Horizon (days): Verified sample 103, Direction consistency 68.9%

All-time model and version results

All-time verified samples: Direction consistency N=14,327, SMAPE N=14,331; Direction consistency: 51.3%; Model reference price proximity: 90.5%.

Ensemble: Direction metric sample N 1,621, SMAPE sample N 1,621, Direction consistency 51.3%, Model reference price proximity 91.8%
LSTM Neural Network: Direction metric sample N 126, SMAPE sample N 126, Direction consistency 53.2%, Model reference price proximity 91.4%
Transformer: Direction metric sample N 379, SMAPE sample N 379, Direction consistency 50.7%, Model reference price proximity 91.1%
G-Prophet 2026V1: Direction metric sample N 11,575, SMAPE sample N 11,579, Direction consistency 51.3%, Model reference price proximity 90.4%
G-Prophet2026V2: Direction metric sample N 537, SMAPE sample N 537, Direction consistency 48.6%, Model reference price proximity 88.5%
Random Forest: Direction metric sample N 89, SMAPE sample N 89, Direction consistency 67.4%, Model reference price proximity 86.8%

Direction consistency compares whether the model reference price and the later observed price are on the same side of the recorded current price. Its sample also requires a valid recorded current price and model reference price. Price proximity is 100% minus mean SMAPE between the model reference price and observed price; it is not a return, and its separate sample requires valid predicted and actual prices with a nonzero denominator. Both metrics use records with status correct or incorrect and a verified actual price; pending records are excluded. The public endpoint does not separately classify overdue records from other pending records.

A dash means a metric has no eligible samples or its sample count is unavailable; it is not a zero score. This is a retrospective aggregate, not a forecast of future performance. Model stability is a reported score, not a calibrated probability. Sample size, market, horizon, model version, and snapshot time are part of the result.

Public data source: GET /api/v1/public/accuracy-dashboard?days=30

G-Prophet

Public Research Snapshot

Model Evaluation Dashboard

Direction consistency, model reference price proximity, and stability across verified research records

Historical aggregate for research review. Each result should be read with its verified sample size, market, horizon, and model version.

Algorithm Comparison

Ensemble

94.4%Price Proximity · N 177

Direction Accuracy 59.3%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 94%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 177⏳ Pending 53

G-Prophet 2026V1

94.7%Price Proximity · N 163

Direction Accuracy 60.7%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 69.4%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 163⏳ Pending 73

G-Prophet2026V2

95.5%Price Proximity · N 24

Direction Accuracy 54.2%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 72.4%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 24⏳ Pending 8

Transformer

95.6%Price Proximity · N 19

Direction Accuracy 63.2%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 84.8%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 19⏳ Pending 1

LSTM Neural Network

98.3%Price Proximity · N 7

Direction Accuracy 85.7%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 71.2%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 7⏳ Pending 0

Random Forest

92.3%Price Proximity · N 4

Direction Accuracy 75%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 43.8%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 4⏳ Pending 0

Performance by Market

A-Shares

94.8%Price Proximity · N 382

Direction Accuracy 60.2%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 81.6%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 382⏳ Pending 131

HK Stocks

87.8%Price Proximity · N 8

Direction Accuracy 62.5%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 62.7%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 8⏳ Pending 3

US Stocks

97.6%Price Proximity · N 3

Direction Accuracy 66.7%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 77%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 3⏳ Pending 1

Crypto

99.4%Price Proximity · N 1

Direction Accuracy 100%Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct|Avg Stability 68%Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

✅ Verified 1⏳ Pending 0

Model Evaluation Dashboard

Sample Size

394

Matured records with verified market outcomes

Validated completed predictions count used as sample size for these statistics.

Direction Accuracy

60.4%

Share of verified records with matching model and observed direction

Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct

Price Proximity

94.7%

How close model reference prices are to observed prices (100 - mean SMAPE)

Research Records

529

Research records created in the selected period

Total predictions made by all users in the selected period. Includes both pending and completed predictions.

Avg Stability

80.6%

Average model stability score

Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

Active Users

Users who made predictions in this period

Unique users who made at least one prediction in the selected period.

Naive Baseline Comparison

Sanity Check

Model metrics are shown beside simple references calculated on the same matured, verified sample, so no percentage stands alone.

Last 30 Days

Sample 394 · Actual up 147 / non-up 247 · majority class Non-up

Direction Accuracy

Model direction

60.4%

Majority-class rule

62.7%

Price Proximity

Model reference price

94.7%

Assume no price change

94.9%

All-Time Sample

Sample 14,327 · Actual up 6,517 / non-up 7,810 · majority class Non-up

Direction Accuracy

Model direction

51.3%

Majority-class rule

54.5%

Price Proximity

Model reference price

90.6%

Assume no price change

92.3%

These naive baselines use the same verified sample as the model metrics. They only test whether the model adds information beyond simple rules; they are not trading strategies or returns.

Metric Trend

Price Proximity

Direction Accuracy

Public Segmented Evidence

Aggregated verified samples by market, horizon, and stability bucket; no portfolio or trading-strategy metrics are included.

Market × Prediction Horizon

Prioritize segments with at least 50 samples

Segment	Samples	Direction Accuracy	95% Range	Price Proximity	Avg Stability
A-Shares · 3D	208	57.7%	50.9-64.2%	94.4% · N 208	90.2%
A-Shares · 7D	91	69.2%	59.1-77.8%	94.2% · N 91	74.9%
A-Shares · 1D	81	55.6%	44.7-65.9%	96.5% · N 81	67.4%
HK Stocks · 7D	8	62.5%	30.6-86.3%	87.8% · N 8	62.8%
US Stocks · 7D	3	66.7%	20.8-93.9%	97.6% · N 3	77%
A-Shares · 5D	2	100%	34.2-100%	95.9% · N 2	65%
Crypto · 7D	1	100%	20.7-100%	99.4% · N 1	68%

Prediction Horizon

Horizon	Samples	Direction Accuracy
1D	81	55.6%
3D	208	57.7%
5D	2	100%
7D	103	68.9%

Stability Calibration

Current data does not show a clear positive relationship between stability and direction accuracy. Treat it as model state context.

Stability/Direction Corr

-0.0333

Verified Samples

394

<50%6 · 83.3%

50-60%27 · 59.3%

60-70%62 · 64.5%

70-80%72 · 59.7%

80-90%71 · 62%

90-100%156 · 57.7%

Confidence Buckets (Cumulative)

These are cumulative samples (for example, ≥70% includes every sample with stability at or above 70%), showing sample size and direction consistency only.

All394 · 60.4%

≥60%361 · 60.1%

≥70%299 · 59.2%

≥80%227 · 59%

Explore quantitative model research

Create model reference paths and review them against subsequent market outcomes

Start Research Already have an account? Log in

Warming up the engine

Evaluation methodology and latest public snapshot

Evaluation period: 30 days (ending 2026-10-05T19:08:36.878Z). Metric version: terminal_direction_v1.

Records / verified outcomes / pending: 529 / 394 / 135
Direction metric sample N: 394
SMAPE sample N: 394
Direction consistency: 60.4%
Model reference price proximity: 94.7%
Mean SMAPE: 5.3%
Data snapshot: 2026-10-05T19:08:36.878Z

Model or version

LSTM Neural Network: Direction metric sample N 7, SMAPE sample N 7, Direction consistency 85.7%, Model reference price proximity 98.3%
Transformer: Direction metric sample N 19, SMAPE sample N 19, Direction consistency 63.2%, Model reference price proximity 95.6%
G-Prophet2026V2: Direction metric sample N 24, SMAPE sample N 24, Direction consistency 54.2%, Model reference price proximity 95.5%
G-Prophet 2026V1: Direction metric sample N 163, SMAPE sample N 163, Direction consistency 60.7%, Model reference price proximity 94.7%
Ensemble: Direction metric sample N 177, SMAPE sample N 177, Direction consistency 59.3%, Model reference price proximity 94.4%
Random Forest: Direction metric sample N 4, SMAPE sample N 4, Direction consistency 75.0%, Model reference price proximity 92.3%

Market

A-Shares: Direction metric sample N 382, SMAPE sample N 382, Direction consistency 60.2%, Model reference price proximity 94.8%
Crypto: Direction metric sample N 1, SMAPE sample N 1, Direction consistency 100.0%, Model reference price proximity 99.4%
US Stocks: Direction metric sample N 3, SMAPE sample N 3, Direction consistency 66.7%, Model reference price proximity 97.6%
HK Stocks: Direction metric sample N 8, SMAPE sample N 8, Direction consistency 62.5%, Model reference price proximity 87.8%

Horizon (days)

1 Horizon (days): Verified sample 81, Direction consistency 55.6%
3 Horizon (days): Verified sample 208, Direction consistency 57.7%
5 Horizon (days): Verified sample 2, Direction consistency 100.0%
7 Horizon (days): Verified sample 103, Direction consistency 68.9%

All-time model and version results

All-time verified samples: Direction consistency N=14,327, SMAPE N=14,331; Direction consistency: 51.3%; Model reference price proximity: 90.5%.

Ensemble: Direction metric sample N 1,621, SMAPE sample N 1,621, Direction consistency 51.3%, Model reference price proximity 91.8%
LSTM Neural Network: Direction metric sample N 126, SMAPE sample N 126, Direction consistency 53.2%, Model reference price proximity 91.4%
Transformer: Direction metric sample N 379, SMAPE sample N 379, Direction consistency 50.7%, Model reference price proximity 91.1%
G-Prophet 2026V1: Direction metric sample N 11,575, SMAPE sample N 11,579, Direction consistency 51.3%, Model reference price proximity 90.4%
G-Prophet2026V2: Direction metric sample N 537, SMAPE sample N 537, Direction consistency 48.6%, Model reference price proximity 88.5%
Random Forest: Direction metric sample N 89, SMAPE sample N 89, Direction consistency 67.4%, Model reference price proximity 86.8%

Public data source: GET /api/v1/public/accuracy-dashboard?days=30

G-Prophet

Public Research Snapshot

Model Evaluation Dashboard

Direction consistency, model reference price proximity, and stability across verified research records

Historical aggregate for research review. Each result should be read with its verified sample size, market, horizon, and model version.

Algorithm Comparison

Ensemble

94.4%Price Proximity · N 177

✅ Verified 177⏳ Pending 53

G-Prophet 2026V1

94.7%Price Proximity · N 163

✅ Verified 163⏳ Pending 73

G-Prophet2026V2

95.5%Price Proximity · N 24

✅ Verified 24⏳ Pending 8

Transformer

95.6%Price Proximity · N 19

✅ Verified 19⏳ Pending 1

LSTM Neural Network

98.3%Price Proximity · N 7

✅ Verified 7⏳ Pending 0

Random Forest

92.3%Price Proximity · N 4

✅ Verified 4⏳ Pending 0

Performance by Market

A-Shares

94.8%Price Proximity · N 382

✅ Verified 382⏳ Pending 131

HK Stocks

87.8%Price Proximity · N 8

✅ Verified 8⏳ Pending 3

US Stocks

97.6%Price Proximity · N 3

✅ Verified 3⏳ Pending 1

Crypto

99.4%Price Proximity · N 1

✅ Verified 1⏳ Pending 0

Model Evaluation Dashboard

Sample Size

394

Matured records with verified market outcomes

Validated completed predictions count used as sample size for these statistics.

Direction Accuracy

60.4%

Share of verified records with matching model and observed direction

Method: Correct Directions / Total Verified × 100% Rules: • Predicted > Current → Predicted Up • Actual > Current → Actually Up • Same direction = Correct

Price Proximity

94.7%

How close model reference prices are to observed prices (100 - mean SMAPE)

Research Records

529

Research records created in the selected period

Total predictions made by all users in the selected period. Includes both pending and completed predictions.

Avg Stability

80.6%

Average model stability score

Average model stability score for each prediction. It reflects volatility and internal consistency, not the empirical probability of being correct. Range: 0% - 100%

Active Users

Users who made predictions in this period

Unique users who made at least one prediction in the selected period.

Naive Baseline Comparison

Sanity Check

Model metrics are shown beside simple references calculated on the same matured, verified sample, so no percentage stands alone.

Last 30 Days

Sample 394 · Actual up 147 / non-up 247 · majority class Non-up

Direction Accuracy

Model direction

60.4%

Majority-class rule

62.7%

Price Proximity

Model reference price

94.7%

Assume no price change

94.9%

All-Time Sample

Sample 14,327 · Actual up 6,517 / non-up 7,810 · majority class Non-up

Direction Accuracy

Model direction

51.3%

Majority-class rule

54.5%

Price Proximity

Model reference price

90.6%

Assume no price change

92.3%

These naive baselines use the same verified sample as the model metrics. They only test whether the model adds information beyond simple rules; they are not trading strategies or returns.

Metric Trend

Price Proximity

Direction Accuracy

Public Segmented Evidence

Aggregated verified samples by market, horizon, and stability bucket; no portfolio or trading-strategy metrics are included.

Market × Prediction Horizon

Prioritize segments with at least 50 samples

Segment	Samples	Direction Accuracy	95% Range	Price Proximity	Avg Stability
A-Shares · 3D	208	57.7%	50.9-64.2%	94.4% · N 208	90.2%
A-Shares · 7D	91	69.2%	59.1-77.8%	94.2% · N 91	74.9%
A-Shares · 1D	81	55.6%	44.7-65.9%	96.5% · N 81	67.4%
HK Stocks · 7D	8	62.5%	30.6-86.3%	87.8% · N 8	62.8%
US Stocks · 7D	3	66.7%	20.8-93.9%	97.6% · N 3	77%
A-Shares · 5D	2	100%	34.2-100%	95.9% · N 2	65%
Crypto · 7D	1	100%	20.7-100%	99.4% · N 1	68%

Prediction Horizon

Horizon	Samples	Direction Accuracy
1D	81	55.6%
3D	208	57.7%
5D	2	100%
7D	103	68.9%

Stability Calibration

Current data does not show a clear positive relationship between stability and direction accuracy. Treat it as model state context.

Stability/Direction Corr

-0.0333

Verified Samples

394

<50%6 · 83.3%

50-60%27 · 59.3%

60-70%62 · 64.5%

70-80%72 · 59.7%

80-90%71 · 62%

90-100%156 · 57.7%

Confidence Buckets (Cumulative)

These are cumulative samples (for example, ≥70% includes every sample with stability at or above 70%), showing sample size and direction consistency only.

All394 · 60.4%

≥60%361 · 60.1%

≥70%299 · 59.2%

≥80%227 · 59%

Explore quantitative model research

Create model reference paths and review them against subsequent market outcomes

Start Research Already have an account? Log in

Model Evaluation and Direction Consistency - G-Prophet