Pipeline

Model Training & Tuning

Continuous model refinement maintaining 91% prediction accuracy through ongoing calibration, Market Scout validation, and segment-specific parameter optimisation across 32,000+ luxury assets.

Model Training & Tuning

Continuous model refinement maintaining 91% prediction accuracy through ongoing calibration, Market Scout validation, and segment-specific parameter optimisation across 32,000+ luxury assets.

Introduction

Reliable predictions require models that evolve with market dynamics. Luxury markets are characterised by shifting collector preferences, macroeconomic influences, and segment-specific cycles—models trained on historical data alone quickly lose relevance.

Prophetic's continuous refinement process ensures our algorithms remain aligned with current market conditions, maintaining the 91% accuracy that underpins all Scores and predictions across our 10 segments.

Note: Model development is an ongoing process. Detailed training methodology remains proprietary.

Training Overview

Continuous Learning Architecture

json

{
  "model_training": {
    "type": "continuous",
    "accuracy_target": "91%",
    "segments": 10,
    "assets_covered": "32,000+",
    "refinement": "ongoing"
  }
}
```

### Training Characteristics

| Characteristic | Description | Benefit |
|----------------|-------------|---------|
| **Data-driven** | Market evidence foundation | Objective basis |
| **Adaptive** | Responsive to market shifts | Current relevance |
| **Iterative** | Continuous improvement cycles | Evolving accuracy |
| **Validated** | Quality-controlled outputs | Reliable results |

> **Important:** Detailed training methodology remains proprietary.

---

## Training Cycle

### The Refinement Loop

Prophetic models operate on a continuous feedback loop:
```
MODEL TRAINING CYCLE
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

     ┌─────────────────┐
     │  Market Data    │
     │   Ingestion     │
     └────────┬────────┘
              │
              ▼
     ┌─────────────────┐
     │    Model        │◀────────────────┐
     │   Training      │                 │
     └────────┬────────┘                 │
              │                          │
              ▼                          │
     ┌─────────────────┐                 │
     │  Market Scout   │                 │
     │   Validation    │                 │
     └────────┬────────┘                 │
              │                          │
              ▼                          │
     ┌─────────────────┐                 │
     │   Outcome       │─────────────────┘
     │   Monitoring    │
     └─────────────────┘
           Feedback Loop

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Cycle Stages

Stage	Process	Output
Ingestion	New transaction data absorbed	Updated datasets
Training	Model parameters adjusted	Refined algorithms
Validation	Market Scout verification	Quality-assured outputs
Monitoring	Prediction vs. actual comparison	Performance metrics

json

{
  "training_cycle": {
    "stages": ["ingestion", "training", "validation", "monitoring"],
    "feedback": "continuous",
    "market_scout": "integrated",
    "quality_gates": true
  }
}
```

> **Note:** Each cycle incorporates learnings from market outcomes.

---

## Data Foundation

### Training Inputs
```
TRAINING DATA LAYERS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  ┌─────────────────────────────────────────┐
  │         External Context                │
  │  ┌───────────────────────────────────┐  │
  │  │      Market Dynamics              │  │
  │  │  ┌─────────────────────────────┐  │  │
  │  │  │   Transaction History       │  │  │
  │  │  │  ┌───────────────────────┐  │  │  │
  │  │  │  │  Asset Attributes     │  │  │  │
  │  │  │  └───────────────────────┘  │  │  │
  │  │  └─────────────────────────────┘  │  │
  │  └───────────────────────────────────┘  │
  └─────────────────────────────────────────┘

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Input Categories

Category	Role in Training	Data Depth
Transaction History	Pattern foundation	100+ years (art)
Asset Attributes	Feature inputs	250 parameters avg
Market Dynamics	Context signals	Real-time
External Factors	Environmental data	Macro indicators

json

{
  "training_data": {
    "transactions": "historical_archive",
    "attributes": "~250_parameters_per_asset",
    "market": "real_time_signals",
    "external": "macro_indicators",
    "quality": "market_scout_validated"
  }
}
```

> **Important:** Training data undergoes rigorous quality validation via Market Scout.

---

## Segment Calibration

### Tailored Model Parameters

Each segment has specifically tuned model parameters reflecting its unique characteristics:
```
SEGMENT CALIBRATION MATURITY
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Segment              Maturity            Data Depth
──────────────────────────────────────────────────
Watches              ████████████████████ Refined    Decades
Sneakers             ██████████████████░░ Refined    Years
Fine Wines           ██████████████████░░ Refined    Decades
Collectible Cards    ████████████████░░░░ Mature     Years
Contemporary Art     ████████████████░░░░ Mature     Century+
Luxury Bags          ██████████████░░░░░░ Mature     Years
High Jewellery       ████████████░░░░░░░░ Established Years
Automobiles          ████████████░░░░░░░░ Established Decades
Real Estate          ██████████░░░░░░░░░░ Adapted    Variable

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Calibration Factors

Segment	Key Calibration Focus	Special Considerations
Watches	Reference-specific, condition weighting	Box/papers premium
Art	Artist trajectory, institutional recognition	Subjective factors
Sneakers	Trend velocity, size availability	Platform premiums
Wines	Vintage variation, critic scores	Storage conditions

json

{
  "segment_calibration": {
    "watches": { "maturity": "refined", "focus": ["reference", "condition"] },
    "art": { "maturity": "mature", "focus": ["artist", "recognition"] },
    "sneakers": { "maturity": "refined", "focus": ["trends", "availability"] },
    "wines": { "maturity": "refined", "focus": ["vintage", "scores"] }
  }
}
```

> **Note:** Each segment has specifically tuned model parameters.

---

## The 91% Accuracy Target

### Accuracy Methodology

Prophetic's **91% accuracy** claim is validated through:
```
ACCURACY VALIDATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

   Historical     Prediction    Actual       Accuracy
   Data           Generated     Outcome      Measured
      │               │            │             │
      ▼               ▼            ▼             ▼
  ┌──────┐       ┌──────┐     ┌──────┐      ┌──────┐
  │██████│  ───▶ │██████│ ──▶ │██████│ ───▶ │ 91%  │
  └──────┘       └──────┘     └──────┘      └──────┘

  Training        Forward       Market        Backtested
  Dataset         Projection    Reality       Performance

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
```

| Validation Method | Description | Frequency |
|-------------------|-------------|-----------|
| **Backtesting** | Predictions vs. historical outcomes | Continuous |
| **Rolling validation** | Recent predictions vs. actuals | Monthly |
| **Cross-segment** | Performance across all 10 segments | Quarterly |

> **Important:** Accuracy is measured across the full prediction range, not cherry-picked outcomes.

---

## Market Scout Integration

### Real-Time Validation

Market Scout provides an additional validation layer beyond model training:
```
MARKET SCOUT IN TRAINING
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

   Model           Market Scout        Published
   Output          Validation          Prediction
      │                │                   │
      ▼                ▼                   ▼
  ┌──────┐         ┌──────┐           ┌──────┐
  │ Raw  │   ───▶  │ AI   │   ───▶    │Verified│
  │Output│         │Check │           │Result │
  └──────┘         └──────┘           └──────┘
                       │
               ┌───────┴───────┐
               │ Fact-checking │
               │ Cross-reference│
               │ Anomaly detect │
               └───────────────┘

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
```

### Validation Functions

| Function | Purpose | Impact |
|----------|---------|--------|
| **Fact-checking** | Verify data accuracy | Error prevention |
| **Cross-reference** | Multi-source validation | Reliability |
| **Anomaly detection** | Flag outliers | Quality assurance |

> **Note:** Market Scout acts as a quality gate before any output reaches users.

---

## Validation Framework

### Quality Gates
```
VALIDATION PIPELINE
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  Model        Validation      Quality       Production
  Output         Tests          Gate          Release
     │             │              │              │
     ▼             ▼              ▼              ▼
 ┌──────┐      ┌──────┐       ┌──────┐      ┌──────┐
 │░░░░░░│ ───▶ │▒▒▒▒▒▒│ ───▶  │▓▓▓▓▓▓│ ───▶ │██████│
 └──────┘      └──────┘       └──────┘      └──────┘

 Generate       Test          Approve        Deploy

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Validation Checks

Check Type	Purpose	Threshold
Consistency	Stable outputs across runs	Required
Accuracy	Market alignment	91% target
Coverage	Segment breadth	All 10 segments
Robustness	Edge case handling	Defined scenarios

json

{
  "validation_framework": {
    "checks": ["consistency", "accuracy", "coverage", "robustness"],
    "gates": "mandatory",
    "threshold": "91%_accuracy",
    "release_criteria": "all_pass"
  }
}
```

> **Important:** Models must pass all quality gates before deployment.

---

## Performance Monitoring

### Continuous Assessment
```
PERFORMANCE MONITORING
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Accuracy
    ▲
95% │  ┌─────────────────────────────────────┐
    │  │░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░│
91% │  │░░░░░░░░░░ Target Range ░░░░░░░░░░░░░│
    │  │░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░│
87% │  └─────────────────────●───────────────┘
    │                        │
    │                  Current: 91%
    └──────────────────────────────────────────▶
              Continuous Monitoring

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Monitored Metrics

Metric	Target	Status
Prediction accuracy	91%	✅ Monitored
Score stability	Low variance	✅ Tracked
Market alignment	Current conditions	✅ Assessed
Segment coverage	All 10 segments	✅ Verified

json

{
  "performance_monitoring": {
    "accuracy": { "target": 0.91, "status": "on_target" },
    "stability": { "variance": "low", "status": "stable" },
    "alignment": { "market": "current", "status": "aligned" },
    "alerts": "automated"
  }
}
```

> **Note:** Performance is continuously monitored against the 91% accuracy target.

---

## Tuning Process

### Parameter Optimisation
```
TUNING WORKFLOW
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

       Performance Signal
              │
    ┌─────────┼─────────┐
    ▼         ▼         ▼
┌─────────┐┌─────────┐┌─────────┐
│ Analyse ││ Adjust  ││Validate │
│  Cause  ││ Params  ││ Change  │
└─────────┘└─────────┘└─────────┘
    │         │         │
    └─────────┴─────────┘
              │
              ▼
       Controlled Deploy

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Tuning Triggers

Trigger	Detection	Response
Market shift	Trend divergence	Parameter review
Performance drift	Accuracy decline	Recalibration
New data patterns	Anomaly detection	Model update
Segment evolution	Market maturation	Factor adjustment

json

{
  "tuning_process": {
    "triggers": ["market_shift", "performance_drift", "new_patterns"],
    "workflow": ["analyse", "adjust", "validate"],
    "approval": "quality_gate",
    "deployment": "controlled"
  }
}
```

> **Tip:** Model confidence indicators reflect current calibration quality.

---

## Model Evolution

### Quality Progression
```
MODEL QUALITY OVER TIME
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Accuracy
    ▲
95% │                              ●───────●
    │                        ●─────┘
91% │ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─●─ ─ ─ ─ ─ ─ ─ ─
    │                 ●─────┘
    │           ●─────┘
    │     ●─────┘
85% │─────┘
    │
    └──────────────────────────────────────────▶
     Initial      Iterations      Current

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
```

| Phase | Focus | Accuracy |
|-------|-------|----------|
| **Initial** | Foundation models | Baseline |
| **Iterations** | Segment calibration | Improving |
| **Current** | Continuous refinement | 91% |
| **Ongoing** | Market adaptation | Maintained |

> **Note:** Model quality improves with each refinement cycle.

---

## Transparency

### What We Disclose
```
MODEL TRANSPARENCY
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

┌─────────────────────────────────────────────┐
│                                             │
│  SHARED WITH USERS:                         │
│  • Confidence levels per output             │
│  • Data quality indicators                  │
│  • Known limitations                        │
│  • Segment coverage depth                   │
│  • 91% accuracy methodology                 │
│                                             │
│  ─────────────────────────────────────────  │
│                                             │
│  PROPRIETARY:                               │
│  • Model architecture details               │
│  • Training methodology specifics           │
│  • Parameter specifications                 │
│  • Weighting formulas                       │
│                                             │
└─────────────────────────────────────────────┘

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
```

| Aspect | Transparency |
|--------|--------------|
| Confidence levels | ✅ Indicated |
| Data quality | ✅ Disclosed |
| Model limitations | ✅ Communicated |
| Accuracy claims | ✅ Methodology shared |
| Training methodology | ❌ Proprietary |
| Parameter specs | ❌ Proprietary |

> **Note:** Prophetic provides confidence context with all outputs.

---

## Quality Standards

### Training Principles
```
TRAINING METHODOLOGY
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

✅ METHODOLOGY INCLUDES:
   • Market data foundation (32,000+ assets)
   • Segment-specific calibration (10 segments)
   • Market Scout validation layer
   • Continuous performance monitoring
   • Controlled deployment process
   • 91% accuracy target maintenance

❌ METHODOLOGY EXCLUDES:
   • Untested model modifications
   • Unvalidated output publication
   • Unstable parameter deployment
   • Non-market speculation

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Important: All model changes undergo rigorous quality control.

Limitations

Training Boundaries

Limitation	Description	Mitigation
Data scope	Based on available history	Continuous expansion
Market coverage	Public transactions only	Multi-source integration
Adaptation lag	Time to integrate changes	Real-time monitoring
Inherent uncertainty	Cannot eliminate	Confidence levels

Appropriate Expectations

json

{
  "training_limitations": {
    "scope": "available_market_data",
    "coverage": "public_transactions",
    "adaptation": "continuous_not_instant",
    "certainty": "probabilistic_not_guaranteed",
    "accuracy_target": "91%_not_100%"
  }
}

Important: Models support decisions but do not guarantee outcomes.

Prediction Algorithms

Portfolio Analytics

Need help? Contact Support

Join our Discord Community

Questions? Contact Sales