The AAA-to-D Crypto Rating Model Explained

— The AAA-to-D Crypto Rating Model Explained —

Key takeaways:

1. A rating model is only useful if it measures quality, not size. Market cap tells you how big an asset is, not how it would perform inside a real portfolio.

2. Crypto Attention's rating is built on portfolio theory. Specifically, it uses the efficient frontier, the same mathematical concept that underlies how institutional money managers construct portfolios. It does not rely on a weighted scorecard of subjective factors.

3. Grades run from AAA to D, using the same letters as traditional credit rating agencies. However, the underlying question is different. This model measures an asset's fit within an optimal portfolio, not its default risk.

4. The model accounts for slippage, the price impact of executing a trade. It does this using the Amihud illiquidity measure. As a result, a $1,000 position and a $500,000 position in the same asset can receive different scores.

5. The final grade comes from a nested process. First, the model identifies the best possible portfolio and ranks those assets highest. It then removes them and repeats the process with the remaining assets. This continues until every tracked asset has been placed.

A rating is only useful if it answers one key question: how would this asset perform in a real portfolio, accounting for the real cost of holding it? Most rating systems on the market today don't ask that question directly. Here's how Crypto Attention's model approaches it.

— What a crypto rating measures —

Most existing rating systems tend to fall into one of two camps:

- The first uses a composite scorecard. It combines weighted factors such as technology, adoption, community size, and security into a single grade. These systems are generally transparent about what they measure.

- The second camp skips quality assessment altogether. Instead, it ranks assets by market capitalization or trading volume. The bigger the asset, the higher the rank.

Neither approach directly answers the question that most investors care about: "How would this asset perform if I invested in it at the size I'm planning to invest?"

A $1,000 position and a $500,000 position in the same asset do not behave the same way. Yet most rating systems do not appear to account for that difference at all.

— Why market cap makes a poor crypto rating input —

Market cap is a poor input for a crypto asset rating. On its own, it doesn't show: - whether an asset is a good investment - how reliably it can be bought or sold - how its risk compares to the rest of the market. A large-cap asset can still be a poor fit for a portfolio. Likewise, a smaller asset can be a strong one. Market cap alone can't distinguish between the two. This is a well-known limitation of existing rating systems. It also serves as the starting point for building something different.

— The math behind Crypto Attention's crypto rating methodology —

The foundational concept is the efficient frontier. It was first formalized by Harry Markowitz in the 1950s, work that later won him the Nobel Memorial Prize in Economic Sciences. Given a set of assets, there is an optimal combination that delivers the highest expected return for a given level of risk. Assets on that frontier are, by definition, the strongest available combination in the market at that time. However, classical portfolio theory assumes liquid markets. In those markets, buying or selling a position does not meaningfully move the price. Most of the crypto market does not meet that assumption. Outside the largest assets, order books tend to be thin. As a result, a large position can move the price before it is even fully entered. To account for this, each asset is scored using the Amihud illiquidity measure. This standard academic metric was first proposed in 2002. It estimates the extent to which a given trade size moves an asset's price. The model subtracts that estimated slippage from an asset's expected risk-adjusted return. As a result, it aims to reflect what an investor would realize, not just the quoted return.

— From score to letter grade —

Finding one efficient frontier only ranks the assets that make the cut. Everything else is excluded from that first pass. To grade every tracked asset, the model removes the top frontier from consideration. It then re-runs the same optimization on what's left. This uncovers a new, slightly less efficient frontier underneath. The process repeats layer by layer until every asset in the tracked universe has a place in the hierarchy. Each layer's raw score is log-transformed and normalized. It is then mapped onto the familiar AAA-to-D scale. These are the same letters used by traditional credit rating agencies like S&P and Moody's, though the underlying question is different. A credit rating estimates the probability of default. This model instead estimates how well an asset would perform as part of an optimal portfolio. It also accounts for the asset's real liquidity constraints.

— A concrete example —

For a tracked universe of about a thousand assets, the number of theoretically possible portfolio combinations works out to roughly 6.67×10²⁴². Testing every combination isn't feasible. That's why the nested frontier search relies on dynamic programming, breaking a large problem into smaller, reusable sub-problems. This is what makes the calculation tractable at that scale.

— The takeaway —

A rating is only as good as the question it answers. A scorecard answers the question "Does this asset check certain boxes?" A market-cap ranking answers "how big is this asset." Crypto Attention's model answers a different question: how would this asset perform in a well-constructed portfolio? That's the difference between a rating that describes an asset and one that's useful for building something with it. These ratings feed directly into Portfolio Construction AI, which uses them to build a live, continuously rebalanced allocation. Ready to see it live? Explore Ratings!

— FAQ —

1. How are crypto ratings calculated? These are calculated either as a weighted scorecard of factors like technology, adoption, and community, or as a simple ranking by market capitalization. Crypto Attention's model works differently. It applies portfolio theory to find the most efficient combination of assets in the market. It ranks those highest, then repeats the process on the remaining assets, layer by layer, until every asset has a grade.

2. What does AAA mean in a crypto rating? AAA in crypto rating means an asset sits at or near the primary efficient frontier. That's the strongest combination of risk-adjusted return available in the market at that time, once slippage and liquidity are accounted for.

3. Is a crypto rating the same as a credit rating? Not quite. A traditional credit rating ( S&P or Moody's) estimates the probability that an issuer defaults on its obligations. AAA-to-D model estimates how well an asset would perform as part of an optimal portfolio. Both use similar letter grades for readability, but they're answering different questions.

4. How do you read a crypto rating? It's best to read a crypto rating as a relative measure, not an absolute one. A higher grade means an asset is closer to the most efficient risk-adjusted combination in the current market, accounting for realistic position sizes and liquidity. It isn't a guarantee of future performance. Grades can shift as market conditions, liquidity, and recent returns change.

5. What factors affect a crypto rating? The factors that affect a crypto rating include an asset's expected risk-adjusted return and its position relative to the rest of the market. Market capitalization plays a role only indirectly, through its effect on liquidity.