Want to make massive gains in crypto? Being EARLY to a hot narrative is the key! Discover my secret method for spotting narratives before they go mainstream. Get ahead of the game with this step-by-step guide Exclusive alpha 1/22

Being early to a narrative is crucial for success, and if it's already buzzing on Twitter, chances are you're not as early as you'd like to be. As a narrative trader, finding untapped sources is key. So, where should you look beyond Twitter?

AMAs, DAO forums, Discord, and Medium offer various possibilities. However, these options may not be as helpful initially, given the sheer volume of new projects emerging every month. It's simply not feasible to follow all these channels simultaneously. It's actually impossible.

So what to do? Here is the solution I use: Find global trends among new projects Find outliers among them Automate Let's go over each step.

Find global trends among new projects There are different ways of doing this. However, today, I'm gonna focus on one specific one: VC investments Contrary to popular belief, The project & the team are not the only parameters VCs look for when considering an investment

The most important elements are: 1. Timing: Too early and adoption is hard. Too late and little market share is left. 2. Industry Potential: Is the project in an industry/niche that has tremendous potential in the next years

Good VCs have teams of experts doing their due diligence on these questions. So look at where crypto VCs are putting their money. Now luckily for you, this is becoming easier than always with tools such as @ChainB @DefiLlama.com and @Messari.

Look at global trends and categories. In which categories are VCs investing? Gaming, infrastructure, Dexes, LSD projects?

Find outliers among them Here is where it gets interesting. A new narrative emerges from projects that don't belong to any existing category. They are outliers. They bring something new (AXS, LDO). Find them and research them. Do you see the potential for a new market?

Automate As a ML developer, I like to automate things and get notifications when my patterns get validated. I will describe the process I use in its simplest form. It might be helpful for some of you. I will share some illustrations. It's a bit technical so feel free to skip

First, I download the data from DefiLlama raises. Now remember, we need to find global trends among these projects VCs have invested. The data contains a 'Description' column. You can use simple embeddings such as TF-IDF to get numerical representations of it.

Once I have the embeddings, I use a common unsupervised learning method called Kmeans. I use the elbow method to determine the optimal number of clusters (27). For reference defillama has 33 categories so 27 seems acceptable.

After fitting the data to KMeans, each project will belong to a cluster. To visualize it, I use PCA to transform the data dimensions in 2D and I plot it. If you just take the simple approach described here, this is what you would get. I'll explain how to improve that later

Additionally, you can also plot the labels to check individual projects based on this chart. Specially the ones further away from the center. The next step is to detect outliers. Every project belongs to a cluster and every cluster has a centroid.

I choose to define as outlier anything above the 80% percentile. As measured by the Euclidian distance from the centroid. Finally, I automate this process by running a cronjob that would do that once per week and send notifications on our Discord if there is any new match.

For instance, here are the results when run on 2023 projects. I'd look at lead investors for those projects, the amount raised and what they actually do. Let's have a look at few examples

1. @a16z crypto lead the round with Azra Games: "NFT-themed Legions & Legends". 2. @Dragonfly ＞|＜ lead a round with Econia Labs: "Hyper-parallelized on-chain order book". 3. @Animoca Brands led a round with MoviePass: "Subscription-based movie ticketing service"

These are 3 good examples. It's not just another dex or game. Might be the birth of a new narrative. And it's backed by good VCs I'd then dive deeper in those, looking at what problem they are trying to solve, who are the competitors, how big is the market they are addressing.

Now this was the simplest way of doing it. But as you saw the clusters overlap a lot. Ideas for improvements: 1. Augment the description with additional data. Twitter bio of the project already enhances a lot the clusters 2. Use more sophisticated embeddings 3. Many other tricks

Well, that's all folks! Use this method to find narratives early before it's maintstream. If you don't code, don't worry. I have a no-code method to find narratives early that I will share in the following weeks.

Follow these accounts for more alpha: @Edgy - The DeFi Edge 🗡️ @rektdiomedes @Patrick | Dynamo DeFi @Subli 🦇🔊 @Shiro @Ignas | DeFi Research @🉐 Crypto Linn @DeFI Saint 🦇🔊 @Thor⚡️Hartvigsen @hoeem @slappjakke.eth 🦇🔊 @Fabian D. @Winter Soldier ❄️🙋🏻‍♂️ @Chilla

Thanks for reading. Give me a follow @Crypto Koryo for more alpha on narratives. If you enjoyed, make sure to like and retweet the first tweet linked below.