Greg

Friend.Tech has brought attention to web3 social. If you haven’t heard of it, Friend.Tech allows users to tokenize and trade “keys” of their social media influence. These keys give people access to engage with creators and influencers they might not otherwise be able to.

*But who do you choose to invest in?*

*How can you tell if someone’s keys are priced accurately?*

If the platform continues [growing](https://twitter.com/nick_emmons/status/1694402532708094109?s=20), there’s potential for users to profit not only from their own social clout––but also their ability to predict the growth of others’ influence.

No matter what happens with Friend.Tech, this represents a major evolution in financializing social capital itself.

“Social capital” refers to the value an individual can derive from their social network and relationships. As sociologists define it, "It is the ability of an individual to access and mobilize the resources available to them through their social ties for their own benefit."

Our engineering team is developing a tool to help quantify social capital, which can be used to forecast the value of Friend.Tech keys.

How does it work, and how can you use it?
## Using X.com to Measure Social Influence
We believe we can use data from X.com (the platform formerly known as Twitter) and measure a user’s social influence. This influence can be measured, in part, by the potential of a user’s actions–such as:

- Likes
- Posts
- Reposts
- Mentions
- Replies
- Follows

–to trigger further actions by other users. The more User 1’s actions trigger further actions by other users, the greater User 1’s social influence. Different actions have more weight in a user’s social influence. For example, a repost signals higher influence than just a like.

To construct an influence network, we extract user interactions from X.com’s tweet metadata, which includes timestamps, authors, usernames of those who replied/liked/reposted, and number of replies/likes/reposts.

Then we combine tweet interactions with following relationships to build a weighted, directed graph representing the influence network.

The resulting influence graph allows us to quantify and compare users' social influence based on their position within the global influence network.
## How to Rank Individual Users
Once we graph out these interactions, we use multiple algorithms to determine a comprehensive ranking of all users:
### 1.  PageRank Algorithm
This algorithm measures the importance of each user based on the number of incoming relationships–or users that interact with them–as well as the importance (or rank) of each corresponding user who interacts with that user.
### 2. Hyperlink-Induced Topic Search (HITS) Algorithm
This ranks each user based on two scores:

- __An Authority Score__: estimates the social influence of each user.
- __A Hub Score__: estimates the value of that user’s relationship to other users.

### 3. Betweenness Centrality

Betweenness centrality is a calculation that further measures the influence of a user, in this case based on how often it acts as a bridge between other users.

The calculation is based on this equation: 
![Eigenvector centrality algo](//images.ctfassets.net/lhkcghnr8ql4/760IgaL6MLin7Gvss7PMmn/1aad35292fc846236979d0eac775d8a8/Eigenvector_centrality_algo.png)

### 4. Eigenvector Centrality

On the other hand, Eigenvector Centrality calculates a user’s influence based on its connections to other highly influential users.

## How to Rank Users Within a Community
Those calculations represent a user’s individual social influence on X.com. But X.com, like any social network, is not simply a collection of individuals. It is a large community comprised of smaller communities.

As such, we must calculate a user’s rank within the community, as well as the ranking of the community itself.

We use two algorithms to calculate this community ranking:
### Community Participation Ranking
This calculation is based on the ratio of intra-community versus inter-community connections a user has. A high ratio indicates more insular connections within a single community, and therefore potentially lower social influence:
![Community participation algo](//images.ctfassets.net/lhkcghnr8ql4/4s0BcxpekOqVjsDah8I2ol/22c84f30b688d617931f35531e4f0cb8/Community_participation_algo.png)
### Community-Hub Bridge Ranking
The Community-Hub Bridge ranking identifies users who act both as hubs within their own community as well as bridges between communities.

The score balances a user's intra-community connections (weighted by community size) with their inter-community connections (weighted by the number of neighboring communities).

It’s represented by this algorithm:

![Community-hub bridge algo](//images.ctfassets.net/lhkcghnr8ql4/6KZK6HRlXv7a6G4FcAdDxQ/7c54e221b3593e2f73199d857c4ff371/Community-hub_bridge_algo.png)

This allows identifying users who exercise influence both internally in their own community and externally by bridging the overall X.com network.

By factoring in both individual influence scores and community-level metrics, our analysis provides a well-rounded quantification of users' influence within their local networks and the broader Twitterverse.
## You Can Leverage Social Influence Analytics
After our ML models have implemented these calculations, we can rank Friend.Tech users (and potential users) based on their X.com influence.

From that ranking, you can reasonably predict which users’ Friend.Tech shares to buy.

Moreover, when we combine these rankings with market data––such as how many of a user’s keys have sold and so forth and number of unique holders per user––we can make reasonable price predictions.

Our current models focus on three key use cases:

1. Predicting the likelihood of an X.com user joining Friend.Tech using their social profile as input to a binary classifier.
2. Ranking Friend.Tech users by their social capital to infer which shares are best to invest in. Users with more weighted social connections rank higher.
3. Forecasting share price for Friend.Tech users by combining social capital with shares sold, number of holders and other relevant signals.

But this is not a final or exhaustive set of metrics. As Friend.Tech continues perfecting their approach, the applications for this tool could be far-reaching––you could use it for:

- Investing in Friend.Tech users
- Building trading bots
- Gathering alpha
- Building your own tools
- And more

The ranking and valuation models we've built are just the beginning. As we refine our algorithms and incorporate additional signals over time, the accuracy of our forecasts will improve. There is still plenty of room to identify other metrics that further capture the nuances of social capital.

To join our waitlist for this new tool, [fill out the form here](https://forms.gle/PrfsyjUQuj6rRXaM7).


Forecast Friend.Tech key prices by ranking users' X.com influence––here's how.

How to Predict Friend.Tech Key Prices by Measuring Social Influence

About Upshot

Related Posts