Retro Game Pricing

Inspiration

Retro game collectors often struggle to estimate the value of rare items due to limited data. The rapid appreciation of retro gaming assets, like a sealed Super Mario 64 selling for $1.56 million, raises questions about the factors influencing their prices. This project investigates whether stock market fluctuations correlate with the real-world value of retro gaming assets or if other factors, such as scarcity and nostalgia, play a larger role.

What it does

Our project models and predicts retro game prices by analyzing stock market trends and other influencing factors. Using data from stock market APIs and retro game pricing websites, we explore whether stock momentum impacts the valuation of retro gaming consoles.

How we built it

• Data Collection: We gathered stock data using the Market Stacks and Alpaca Markets APIs and collected retro game asset prices from PriceCharting.com.
• Modeling Approaches:
  • Momentum Model: A simplified short-term prediction model that assumes stock and console price trends are correlated.
  • Exponential Moving Average (EMA): A smoother, long-term model that accounts for exponential stock growth.
  • Multivariate Linear Regression (MLR): A more complex model incorporating multiple stock influences on retro game prices.

Challenges we ran into

• The momentum model was highly volatile, leading to negative price predictions.
• Retro games, unlike stocks, are not actively produced, making it difficult to apply traditional stock models.
• Determining the right mix of variables for an accurate prediction was challenging.

Accomplishments that we're proud of

• Successfully implemented multiple pricing models and identified key factors influencing retro game prices.
• Developed a methodology to compare stock market trends with gaming asset prices.
• Proposed a future roadmap involving neural networks for more robust predictions.

What we learned

• Retro game prices are influenced by a combination of scarcity, nostalgia, and economic trends.
• Momentum models alone are insufficient due to market noise, and EMA-based models provide better stability.
• Using machine learning techniques, such as MLR and potentially neural networks, can improve prediction accuracy.

What's next for Retro Game Pricing

• Implement a neural network to capture complex relationships beyond linear regression.
• Normalize stock value changes using sigmoid functions to enhance model sensitivity.
• Expand dataset coverage to include more economic indicators and game-specific attributes.

Built With

• alpaca
• javascript
• marketstack
• node.js
• react