TL;TR The Bitcoin Factor model quantifies the relationship between normalized indicators (i.e. observed variables) and their underlying latent construct, which is Bitcoin being oversold or –bought, in a 1-factor score. This factor score explains the variance of the indicators in the model and quantifies Bitcoin’s oversold and –bought zones with factor scores ranging from 0 to a 100. NOTE: This model is NOT a price prediction model.
Factor analysis is a useful tool for investigating variable relationships for complex concepts such as socioeconomic status, dietary patterns, psychological scales and in our case if Bitcoin is oversold — or bought. It allows researchers to investigate concepts they cannot measure directly. It does this by using a large number of variables to estimate one or more interpretable underlying factors. The key concept of factor analysis is that multiple observed variables have similar patterns of responses because they are all associated with a latent variable (i.e. not directly measured). CFA is often used in in machine learning models with large numbers of observed variables, where it reduces dimensionality in a unsupervised way.
For example, socioeconomic status is a factor you can’t measure directly. However, you can assess occupation, income, and education levels. These variables all relate to socioeconomic status. People with a particular socioeconomic status tend to have similar values for the observable variables. If the socioeconomic status-factor has a strong relationship with these observed variables, then it accounts for a large portion of the variance in these observed variables. In other words, the factor explains the spread of the data (i.e. variance) in the observed variables. This may sound abstract, but you have all heard off correlation or regression: that’s exactly explaining variance. For example, a value of 0 implies that the predictor variable(s) provide no explanation for the response variable. Vice versa, a value of 1 does imply that the predictor variable(s) provide explanation for the response variable.
In our Bitcoin Factor model we have reduced our analysis to 1 factor. It may be that the indicators underlie more than 1 factor, but for simplicity we stick to 1 factor. Hence, this model is designed for retail investors so simplicity is key. Plus, for a multi-factor model more indicators are preferred. Our theory is that all indicators represent the current status of Bitcoin being oversold or -bought. The indicators included in this factor model are:
- Mayer Multiple
- Market cap/Realized cap (MVRV) Z-score
- Net Unrealized Profit/Loss (NUPL)
- Reserve Risk
- Dormancy Flow
- Realized Value HODL Ratio (RHODL)
- Puell Multiple
- Market cap to Thermo cap ratio
These indicators together give a comprehensive overview of the Bitcoin market based on key-concepts of the 200 daily SMA, realized price, HODL sentiment and miners activity (see links below this article for full explanations per indicator).
Before adding to the model, shown indicators are adjusted for reduced volatility and diminished returns. This is because while the Bitcoin market matures, market bottom and top formations are less pronounced. Moreover, on-chain indicators might lose power due to concentration of speculative liquidity in exchanges, new custody solutions for institutions, the Lightning Network, sidechains etc. By adjusting, these bottom and top formations level out so they still can be recognized and quantified. Also, we have normalized these indicators so bottom formations get a value of 0 and top formations a value of 1 (or close to).
The arrows pointing from the Bitcoin factor to the indicators illustrates correlation arrows (officially they are called factor loadings). The Bitcoin factor score from our model are also normalized so values range between 0 and 1. We have multiplied the score by a 100 to get sensible values and scales.
When we look at which prices Bitcoin was hovering when factor scores where 93 or higher, we see that this was during periods of all-time-highs (ATH) at that moment. Just slightly missing out recent ATH closing price in 2021 ($67,542), yet indicated level ($61,289) was still very near that ATH-peak.
What’s even more impressive is the power of catching oversold DAYS (so not day-ranges but exactly days). Bitcoin prices were hovering around local bottom-days when factors scores were 1 or lower.
To summarize:
- The Bitcoin Factor Model combines indicators as shown here to increase the likelihood of catching bottom and top formations.
- It seems to be specifically usable to quantify oversold zones, where overbought zones are also applicable but needs to be used a bit more conservative.
- The Bitcoin Factor Model will be an on-going developed model with possibly extra or other indicators to constantly improve fit, so keep an eye on it!
- See www.bitnify.app for more info.
Explanation indicators:
