@Sakura please summarize this article, thanks uwu.
TLDR:
Backtests can be misleading if based on revised data; using Point-in-Time (PiT) metrics ensures historical accuracy. 
Key Points:
- Misleading Backtests: Backtests using revised data can lead to incorrect conclusions.
- Point-in-Time Data: PiT metrics are immutable and reflect only the information available at the time.
- Look-Ahead Bias: Using updated data introduces biases that can distort trading strategies.
- Performance Comparison: Strategies based on PiT data often perform worse than those using revised data.
- Importance of Accuracy: Accurate historical data is crucial for reliable trading analysis.
In-depth summary:
The article discusses the pitfalls of backtesting trading strategies using revised data, which can lead to misleading results. It highlights a hypothetical trading strategy based on the movement of coins off exchanges, suggesting that such movements can indicate bullish trends. However, the author emphasizes that a single day of outflows is not enough to determine a trend, and a moving average crossover is used to confirm sustained outflows.
The core issue arises from the fact that many metrics, especially those related to on-chain data, are often retroactively revised. This means that the data available today may differ from what was known at the time of trading decisions, leading to a phenomenon known as look-ahead bias. To illustrate this, the article compares two backtests: one using standard exchange balance data and another using Point-in-Time (PiT) data, which is immutable and reflects only the information available at the time it was computed.
The results show that while the standard backtest may appear favorable, the PiT-based strategy performs significantly worse, highlighting the importance of using accurate historical data. The author concludes that backtests can be deceptive if they rely on incorrect or revised data, and only immutable PiT metrics can ensure that historical analysis is conducted accurately.
ELI5:
When you test a trading strategy, using old data that has changed can make it look better than it really is. It’s like checking your test answers after the teacher has given you the right ones! Using Point-in-Time data means you only look at what you knew back then, which helps you see the real picture. 
Writers main point:
The primary point the author is making is that backtests can be misleading if they are based on revised data, and using Point-in-Time metrics is essential for accurate historical analysis. 