Introduction and methodology

Background

Tokens with high Fully Diluted Valuations (FDVs) and low initial circulating supply have been a hot topic of discussion within the industry. FDV is a key metric in finance, representing the total market value of an asset. Within the cryptocurrency market specifically, FDV represents the market value of a cryptocurrency if all existent tokens were in circulation. On the other hand, initial market cap (IMC) refers to the market value of a token at the time of its token generation event (TGE) considering its floating supply and its launch price. 

‍

A question arises - why do projects within the crypto industry decide to not have 100% of its token supply available at launch? The answer is that it allows them to optimize their tokenomics to ensure long-term sustainability. The motivation behind this strategy is to avoid the entire value of the FDV being available for selling straight away, causing significant decreases in the asset’s price. Limiting the initial supply allows the token to face much less sell pressure, supporting its price.

‍

As the industry has evolved, new strategies and mechanisms have been developed to have a more calculated approach for token distribution. Projects have different tranches that represent the amount of tokens being used for different purposes. While there is no “one fits all” tokenomics model, projects often allocate specific percentages for investors, core team & early contributors, community incentives, marketing, treasury, and liquidity for the initial launch on centralized and/or decentralized exchanges. This is very different from how the industry started, with Proof of Work models creating tokens and putting them into circulation only after someone created value for that particular chain by validating blocks, and whilst this is still a sustainable model, projects nowadays prefer a more traditional approach that allows venture capital and other traditional finance ideologies to come into play.

‍

In the end, not launching with 100% of tokens unlocked straight away reflects a broader understanding of market dynamics and the importance of sustainable growth in the cryptocurrency space. By leveraging the correct distribution mechanisms, projects have an opportunity to enhance sustainability, manage volatility, build strong communities, or even comply with regulations.

‍

Among the distribution and launch strategies that have been conceived, the High FDV and Low IMC model has gained prominence in recent years. This approach is supposed to offer several advantages over traditional distribution methods, including improved ability for price discovery and enhanced market sentiment. 

‍

What is low float, high FDV anyways?

• Low float and high FDV is a strategy that crypto projects can use for their initial token launch.
• It consists of the combination of high valuations with low circulating (or floating*) token supply at launch.

*The difference between circulating and floating supply is rarely spoken about, but from our experience it seems to be that circulating applies to all vested tokens, whereas floating applies to all vested tokens that are on the market which doesn’t include staking contracts, lock ups, vaults, and other ways the token can be taken off the market.

‍It’s been a common activity for projects to use this launch strategy as it’s believed that allocating a low circulating supply at launch can translate to quicker price appreciation, following the most fundamental principle in economics that explains how the availability of an asset relative to consumer demand affects its price. In this case, the law of demand and supply would, on paper, point to higher price appreciation if an asset, in this case a token, is scarce at launch. This is because more buyers are competing for a limited number of tokens, leading to higher bids ultimately driving up the price until a point of equilibrium is reached.

While having a low circulating supply could help drive buying pressure up due to scarcity, upcoming token unlocks could have the opposite effect on the token. With a larger percentage of the supply due to be unlocked, there would be increased selling pressure after the token’s launch, as early holders and investors might look to realize profits and the token becomes less scarce simultaneously as the project’s initial support fades. This surge in selling pressure can outweigh the initial scarcity-driven demand, potentially leading to a decline in the token's price over time. Understanding this balance between initial scarcity and the impact of unlock schedules is critical to evaluating a token's performance trajectory, and thus, this becomes one of the key reasons to perform a statistical analysis that finds the relationship between a high FDV and low IMC/float (used interchangeably hereafter) and long term token performance. 

‍Objective

‍While there has been a wave of projects using this strategy for their launch, it has been seen that many of them don’t perform well in the long run. With this analysis, we are aiming to analyze the relationship between a token’s initial FDV & IMC, and its performance. 

‍Methodology

‍Firstly, the analysis called for the identification of tokens that launched under the High FDV and Low IMC model. As there is no public repository or directory that directly displays all tokens launched under this model, the selection process was more rudimentary. Said process began by going through a list of recent token launches, ideally over the last year. For each of the tokens that was launched over this analyzed time period, their total and initial circulating supply was collected. Only once we had a considerable amount of tokens, we sorted the list by the initial circulating supply in ascending order, allowing us to have a list of around 16 tokens that had a relatively low initial circulating supply, with the lowest initial circulating supply being 1.11% (PRIME), and the largest being 15.4% (PIXEL).

While the High FDV and Low Float model has been popularized as a launch strategy for tokens in recent years, its actual application in the industry reveals that the model’s concept is nuanced. Based on prior research and general industry standards, most discussions around this model focus heavily on the “High FDV” component, often treating it as a sign of market confidence on a new project. However, the “Low Float” aspect, which refers to the low initial circulating supply, is far less standardized and, in many cases, not even genuinely that low.

‍

One of the primary obstacles encountered during this research was the scarcity of tokens that truly fit the High FDV/Low Float model. While we were set out to analyze an extensive data base, we quickly discovered that many tokens claimed to follow this model but did not actually meet the criteria for having a genuinely low initial circulating supply. The “low float” often appears subjective, as there is no set industry standard for how low a circulating supply should be to qualify. This inconsistency further complicates efforts to define and study the impact of the model.

‍

The lack of a clear standard for what classifies as “low float” not only highlights the nuanced understanding of this model but also points to the need for further investigation into its implementation. While this was a challenge for this report, it also offers an opportunity to explore how the model is being used and whether its potential advantages are true. This research, therefore, serves as both an analysis of the High FDV/Low Float model and a commentary on its limitations and lack of standardization within the industry.

‍

Nevertheless, once the list of tokens to be analyzed was ready, the token’s prices and details required for the analysis were collected, getting the token’s prices at launch, allowing us to later collect the prices one week, one month, and two months after their launch. This part of the process was done through CoinGecko’s API, which allowed us to get accurate prices.

‍

Statistical analysis

For the analysis we utilized several statistical models and tools that allowed us to measure the relationship between our selected independent variables (predictors) and token price performance over three timeframes: 1 week, 1 month, and 2 months post-launch.

‍

• Predictors
  
  • Initial FDV (Fully Diluted Valuation): The total valuation of the token at the time of launch, assuming all tokens are in circulation.
  • IMC (Initial Market Cap): market capitalization of a newly-launched token, considering its initial circulating supply of tokens by their launch price
  • FDV/IMC Ratio: ratio that compares a token's FDV to its IMC at the time of launch, allowing to see how much of the token's value is currently unlocked versus the potential value if all tokens in the supply were available in the market.

We incorporated several statistical models, including:

1. Multiple Linear Regression: to assess the direct relationship between predictors and token returns.
2. Ridge Regression: a regularized linear regression method designed to mitigate issues of multicollinearity between features.
3. Random Forest Regression: non-linear model capable of capturing more complex relationships within the data.
4. Lasso Regression: regularized linear regression technique that performs feature selection by shrinking certain coefficients to zero.

The analysis was performed using Python due to the availability of libraries such as pandas for data manipulation, numpy for numerical computations, and scikit-learn for implementing machine learning models. Additionally, the statsmodels library was utilized for detailed statistical tests and statistical data exploration.LimitationsThis research was subject to several limitations, primarily due to our scope and the data availability at the time of analysis. Thus, this report serves more as an exploratory representation to discover potential patterns.

We acknowledge that for a statistical analysis to be statistically significant, you need a very large number of observations which in this case is not viable. Our analysis includes 16 tokens, reflecting the limited availability of tokens launched under the High FDV and Low IMC model in recent months. While analyzing only 16 tokens may not appear to be representative of a broader market, this approach was necessary because the High FDV and Low IMC model is a more recent breakthrough and there aren’t many tokens that were recently launched with it.

Our analysis focused on short and mid-term effects of our predictors on price performance, and additional timeframes could be introduced in future analysis to assess the long-term effects of these predictors on price dynamics.

‍

The models don’t account  for other factors such as overall market conditions and the effect of market leaders on tokens, which also has a great impact on price performance.

‍

Despite our limitations, this research can offer valuable insights and will act as the foundation for more in-depth analysis that can answer what drives a token’s price performance in the future. We are setting the stage for further exploration into the factors that drive token price performance and even whether the High FDV and Low IMC model affect other traits.

Analysis Results

Regression Analysis

1-Week Price Performance Regression Analysis

In the regression model for the 1 week returns with the selected predictors Initial FDV (x1), IMC (x2), and the FDV/IMC Ratio (x3), the results suggest that while the constant term or baseline value has a meaningful contribution, none of the independent variables (x1, x2, x3) show statistically significant relationships with the dependent variable. What this means is that while the starting point of the model has a strong influence on the model, the predictors (Initial FDV, IMC, FDV/IMC Ratio) are additional factors that, in this analysis, do not show any meaningful or significant impact on shaping the result.

‍

The results of this analysis are summarized as follows:

• R-Squared: 0.28, it indicates that the model explains 28% of the variance in price performance. While this is low, it suggests some explanatory power.
• Adjusted R-squared: 0.100, reflecting a decline when accounting for the number of predictors, which indicates that some predictors may not actually contribute anything to the model.
• MSE: 6,233,230.61
• P-Values:
  
  • Constant: 0.002 (statistically significant at the 5% level)
  • Predictors: they’re all not statistically significant, as their p-values are above 0.05
    
    • X1 (FDV): 0.872
    • X2 (IMC): 0.428
    • X3 (FDV/IMC Ratio): 0.215
• T-Statistics:
  
  • Constant: 3.952, which is above a typical threshold of 2, confirming statistical significance.
  • Predictors: all of which are too low to suggest significance
    
    • X1 (FDV): 0.164
    • X2 (IMC): -0.821
    • X3 (FDV/IMC Ratio): -1.309

 1-Month Price Performance Regression AnalysisWhen applying the regression model to explain the influence of the same predictor variables over the 1-month price performance, the model explains only about 33% of the variation in 1-month price performance, meaning most of the factors driving price performance are not captured by these independent variables. The constant (baseline) is significant, meaning there is some price performance when none of the predictors have any effect. However, none of the three predictors (Initial FDV, IMC, or FDV/IMC Ratio) show a significant or reliable impact on price performance, appearing to be weak or even negligible.The results of the analysis for 1-month price performance are summarized as follows:

• R-Squared: 0.330, indicating that the model explains 33% of the change in the 1-month price performance. While this is higher than for 1-week performance, it still suggests the model has limited explanatory power.
• Adjusted R-Squared: 0.163, showing how after accounting for the predictors, the model explains only 16.3% of the variance, highlighting that some predictors are not contributing meaningfully.
• MSE: 3,914,248.90
• P-Values:
  
  • Constant: 0.001 (statistically significant at the 5% level)
  • Predictors: none of them are statistically significant
    
    • X1 (FDV): 0.962
    • X2 (IMC): 0.495
    • X3 (FDV/IMC Ratio): 0.201
• T-Statistics:
  
  • Constant: 3.952, which is above a typical threshold of 2, confirming statistical significance
  • Predictors: all of which are too low to suggest significance
    
    • X1 (FDV): -0.048
    • X2 (IMC): -0.703
    • X3 (FDV/IMC Ratio): -1.353

‍2-Month Price Performance Regression AnalysisLastly, by applying the regression model to explain the effect of the predictor variables over the 2-month price performance, it showcased that the influence of FDV, IMC, or FDV/IMC Ratio over price is only explained to about 18.9% in the 2-month period. Once again, this indicates that most of the factors that do influence price performance are not statistically represented by the selected variables in this analysis. 

‍

For this time period after launch, the constant is significant, meaning that there is some price performance when none of the predictor variables have any effect. The analysis also showcased that none of the three predictor variables attribute a significant or reliable price impact, appearing to be negligible. The results for the 2-month price performance are summarized as follows:

• R-Squared: 0.189, indicating that the model only explains about 18.9% of the change in price 2 months after the tokens launched. The R-Squared is even lower than what was seen in the 1-month analysis, suggesting that these predictors are also poor at explaining the price change over a longer time frame.
• Adjusted R-Squared: -0.013, indicating that after accounting for the effect of predictors, the model has no meaningful variance, implying that the predictors are not contributing to the regression model.
• MSE: 8,233,018.26
• P-Values:
  
  • Constant: 0.003, statistically significant at a 5% significance level
  • Predictors: none of them are statistically significant
    
    • X1 (FDV): 0.837
    • X2 (IMC): 0.438
    • X3 (FDV/IMC Ratio): 0.501
• T-Statistics:
  
  • Constant: 3.741, being above a typical threshold of 2 means that it is statistically significant
  • Predictors: all T-Statistics are too low, indicating no significance
    
    • X1 (FDV): 0.210
    • X2 (IMC): -0.802
    • X3 (FDV/IMC Ratio): -0.694

‍Exploratory Analysis: Correlation HeatmapWe decided to include a correlation heatmap to visually summarize the relationships between key metrics in the dataset. It provides a straightforward way to identify patterns and potential interactions between variables, acting as a summary of the dynamics between our factors.The heatmap displays correlation coefficients which measure both strength and direction of the linear relationship between two variables. Positive correlations, closer to +1, indicate that as one variable increases, the other tends to increase, while negative correlations, closer to -1, suggest the opposite relationship. Lastly, a correlation coefficient of zero indicates that there is no linear relationship between the two variables.Key takeaways from this heatmap include:

• Token prices across the three timeframes (1 week, 1 month, and 2 months) are strongly correlated, which mainly suggests that the price movements are consistent over the observed periods.
• The predictors (FDV, Initial IMC, and FDV/IMC Ratio) show weak negative correlations with token returns, strengthening the findings from our regression models that these variables do not significantly explain price performance.
• Interestingly, the FDV/IMC Ratio has a moderate positive correlation with the token’s Initial Price, suggesting this metric may have some influence on early price dynamics.

‍

Other Statistical Models: Ridge, Lasso, and Random Forest Regression

‍So far, the statistical analysis models have been ineffective in explaining the price performance across all analysed time frames. To address this, additional statistical models were explored to potentially enhance the scope of the analysis, as mentioned in the methodology section: Ridge Regression, Lasso Regression, and Random Forest Regression. The outcomes of these models are presented next.

‍For 1-Week Price Performance

Ridge Regression:

• MSE: 4,952,419.53
• R-squared: -20.23

• Coefficients:
  
  • X1: -145.14
  • X2: -851.58
  • X3:  -808.39

The Ridge model performed poorly, with a Mean Squared Error (MSE) of 4,952,419.53 and an R-squared value of -20.23. This indicates that the model fails to explain the variance in price performance. The coefficients suggest a negative relationship with all predictors, but their magnitude is large, which could be an indicator of multicollinearity issues.Lasso Regression:

• MSE: 6,172,914.85
• R-squared: -25.47
• Coefficients:
  
  • X1: 239.72
  • X2: -1276.67
  • X3: -987.57

The Lasso model also performed poorly, outputting an MSE of 6,172,914.85 and an R-squared value of -25.47, even worse than Ridge Regression. While Lasso shrinks the coefficients more significantly, the large coefficient values for IMC and FDV/IMC Ratio suggest instability in the model's ability to explain the effects on the desired variable.Random Forest Regression:

• MSE: 3,531,208.84
• R-squared: -14.14
• Coefficients:
  
  • X1: 0.31
  • X2: 0.62
  • X3: 0.05

The Random Forest model produced the lowest MSE of 3,531,208.84 but had an R-squared value of -14.14, which was the worst among all models. Despite showing an ability to capture non-linear relationships, the model failed to explain the changes in price performance. The feature importance analysis ranked IMC as the most influential predictor, followed by FDV and the FDV/IMC Ratio accordingly, but the poor overall performance suggests the data lacks meaningful patterns for prediction.For 1-Month Price PerformanceRidge Regression

• MSE: 3,580,009.03
• R-squared: -72.19
• Coefficients:
  
  • X1: -255.99
  • X2: -555.52
  • X3: -623.32

For the 1-month price performance analysis, the Ridge model resulted in a Mean Squared Error (MSE) of 3,580,009.03 and an R-squared value of -72.19. While this is still a poor fit, it is an improvement compared to the 1-week period, mainly considering the MSE decrease. In this case, all the coefficients are negative, indicating an opposite relationship with all predictors, which could indicate that higher values for FDV, IMC, and FDV/IMC Ratio may correlate with lower price performance. However, the overall model fit remains weak and the effect of the predictors cannot be statistically significant.Lasso Regression:

• MSE: 3,905,441.91
• R-squared: -78.84
• Coefficients:
  
  • X1: -55.55
  • X2: -786.16
  • X3: -730.61

The Lasso model shows an MSE of 3,905,441.91 and an R-squared value of -78.84, which is slightly worse than Ridge Regression. The coefficients were also negative for all predictors, with FDV and IMC being penalized considerably more. This suggests that Lasso may be over-penalizing the coefficients, reducing its ability to capture the true relationships between predictors and the dependent variable.Random Forest Regression:

• MSE: 4,472,732.55
• R-squared: -90.44
• Feature Importances:
  
  • X1: 0.47
  • X2: 0.39
  • X3: 0.12

The Random Forest model had the highest MSE of 4,472,732.55 and the worst R-squared value of -90.44 among the three models. Feature importance analysis showed that the FDV was the most significant predictor for 1-month price performance, in contrast to the 1-week period, where IMC had the highest importance. Despite this, the model’s performance remains poor, indicating that it still fails to capture meaningful patterns in our data.

‍

For 2-Month Price Performance

Ridge Regression:

• MSE: 6,297,347.22

• R-squared: -119.70

• Coefficients:

• X1: -64.06
• X2: -10
• X3: -451.36

The Ridge model resulted in a Mean Squared Error (MSE) of 6,297,347.22 and an R-squared value of -119.70, indicating a poor fit. The coefficients for all predictors were negative, similar to previous periods. However, the smaller magnitudes of the coefficients suggest weaker relationships between the predictors and price performance.Lasso Regression:

• MSE: 8,153,267.40
• R-squared: -155.28 (Worse than Ridge.)
• Coefficients: Similar to Ridge, but more shrunk coefficients.
  
  • X1: 376.80
  • X2: -1498.58
  • X3:  -626.45

The Lasso model performed worse than Ridge, with an MSE of 8,153,267.40 and an R-squared value of -155.28. The coefficients are more heavily shrunk due to Lasso's regularization. This is particularly evident for IMC and FDV/IMC Ratio, which have been penalized further, reflecting potential over-penalization.Random Forest Regression:

• MSE: 9,213,557.46
• R-squared: -175.60 (This is still poor, and even worse than the previous periods.)
• Feature Importances: IMC is the most important predictor here as well, which is consistent with the 1-week return period.
  
  • X1: 0.34
  • X2: 0.40
  • X3: 0.25

The Random Forest model yielded the highest MSE of 9,213,557.46 and the worst R-squared value of -175.60, reflecting a failure to capture meaningful patterns in the data. Feature importance analysis revealed that IMC is the most significant predictor, followed by FDV and FDV/IMC Ratio, consistent with results from the 1-week return period. However, the poor overall performance highlights the inadequacy of this model for the given data.

‍Conclusion

This research was set to analyze the relationship between Initial FDV, Initial IMC, and the FDV/IMC Ratio with token price performance over three timeframes: 1 week, 1 month, and 2 months post-launch. 

‍

Despite leveraging multiple statistical models and incorporating tools to ensure robust analysis, the findings consistently revealed weak and statistically insignificant relationships between these predictors and price performance. 

‍

The analysis’ nature, combined with the limitations such as a small data base and exclusion of external factors suggests that the High FDV and Low IMC model alone was not sufficient to explain price performance. 

‍

However, the report provides a foundation for more studies to include additional factors like market conditions, token utilities, or vesting schedules, paving the way for a deeper understanding of what really drives token price performance.

‍