MarketsBinance perps
--:-- UTC
Research Sign in

DeFi Yields, Scored

DeFi yields for every major pool, live, each scored out of 100 on what can be measured: size, track record, audit, chain and yield quality.

Loading Pools -- Yields live Protocol data --

Highest USDC yield, 55+

--

Loading

Highest ETH yield, 55+

--

Loading

Highest BTC yield, 55+

--

Loading

Highest SOL yield, 55+

--

Loading

Best DeFi Yields by Trust Score

How the score is built →
# Protocol / pool Trust score APY Rewards TVL vs 30d avg Chain Site
Reading the pool feed.

About this data

Updated Sep 2026

Figures are shown as the source publishes them, and anything we calculate is explained below. How the data is made Report an error General information, not financial advice.

What a DeFi yield actually is

Every number on this page is someone paying you to take a risk. The question is never whether the yield is real, because it usually is. The question is what you are being paid for, and whether the payment is worth it.

There are only three places a DeFi yield comes from. Interest, where a borrower pays to use your asset, which is the lending markets. Fees, where traders pay to use the liquidity you supplied, which is the pools. And emissions, where a protocol prints its own token and hands it to you for showing up. The first two are revenue. The third is marketing, and it stops when the budget does.

The score below cares a great deal about which of the three you are looking at.

How the score is built

Six components, scored out of 100, every one of them a number rather than an opinion. Hover any score in the table and it will show you its own breakdown.

ComponentWhat it measuresWeight
DepthPool TVL, on a log scale from $1M to $10B. Depth is not safety, but a shallow pool is easier to move and easier to leave.22
Track recordHow many days the pool has actually run, and how long the protocol has been listed. Survivorship is the cheapest evidence there is.22
Yield qualityWhether the number is plausible: its level, how much is emissions rather than revenue, how volatile it has been, and whether it agrees with its own 30-day mean.20
Audit statusWhether an audit is on record at the source. Not a judgement of the audit, only of whether one exists.13
ChainThe TVL of the chain settling the position, on a log scale. A yield is only as good as the chain underneath it.13
ExposureSingle-asset, a stable pair, or a volatile pair carrying impermanent loss.10

The caps

Some facts should hold a score down no matter how well a pool does elsewhere, so after the six parts are totalled a ceiling applies:

  • Pays over 100% holds the score to 44, and over 50% holds it to 54, below the line of 55 the header tiles use. A yield that high is not a better version of a normal yield, it is a different thing.
  • Flagged an outlier by the source holds it to 54.
  • Impermanent loss holds it to 74, so no volatile liquidity pool reaches the Strong tier however large it is.
  • No audit on record holds it to 64.

The number is shown with a plain tier beside it, so the table can be read at a glance:

75+ Strong
55+ Fair floor for the header tiles
35+ Weak
0+ Poor
-- no score: nothing on record for the protocol

How to read the scores

The four figures in the header are the highest APY on each asset that scores 55 or more, pays for a single-asset deposit rather than a liquidity pair, and holds at least $10M. Higher numbers exist further down the table; they are further down for a reason.

Sort the table by score and the pattern is the whole page in one move: APY rises steadily as the score falls, which is exactly what it should do. That gap is the risk premium, and it is the price of everything the score is measuring.

What it means in practice is that a yield well above its score's median is not a bargain anyone else has missed. It is a pool being paid more because something about it is worse, and the useful question is which of the six components knows why. Hover the score and find out.

What this cannot tell you

This is the important section, and it is longer than the one above it on purpose.

  • It does not measure smart-contract risk: Nothing here reads the code. The failures that have cost people the most money in DeFi were code failures in large, audited, deeply liquid protocols, and every one of them would have scored well on this page the day before.
  • It does not measure admin keys or governance: Who can upgrade the contract, who can pause it, and who can drain it are the questions that matter most, and none of them is a number in a feed.
  • It does not measure oracle design, which is the single most common attack surface in lending markets.
  • An audit on record is not a clean bill of health: The score knows only that a document exists. It has not read it, and neither has it checked whether the code deployed today is the code that was audited.
  • Depeg risk is not scored: A stablecoin pool is marked as one, but nothing here estimates the chance the stablecoin stops being stable.
  • An APY is a snapshot, not a promise: It is the current rate annualised. Rates move continuously, and the 30-day column is there to show you how much this one has.
  • A score is relative, not absolute: A score in the 90s means the best-measuring pool in a market that is young, largely unregulated and has lost users billions. It does not mean safe.

Not investment advice

Alpha Finance is not a licensed adviser and nothing on this page is a recommendation to deposit into any protocol. A score is a summary of measurable properties, published so you can see how it was reached and disagree with it. Yields, deposits and protocol code change continuously, and every figure here is only as current as the timestamp beside it. You can lose everything you put into any of this.

Questions

How is the trust score calculated?

From six measurable components, scored out of 100: pool TVL, how long the pool and protocol have run, whether an audit is on record, the TVL of the chain it settles on, whether the yield looks plausible and revenue-backed rather than emissions, and whether the position carries impermanent loss. Caps are then applied for facts that should hold a score down outright, such as an APY over 100% or no audit on record.

What does a score of 55 or more mean?

The pool did well on size, track record, audit status, chain and yield plausibility, and sits in the Fair tier or above. It is the line the four header tiles use. A trust score is our own sum of what can be measured. It is not a credit rating, it does not mean the pool is safe, and it is not a recommendation to put money into it.

Why is a pool paying 300% scored so low?

Because the level of a yield is itself a risk signal. Above 50% the yield component of the score falls steeply and the score is held below 55, and above 100% it is held below 45 no matter how the pool does on size or history. A yield that high is usually emissions, a very short-lived fee spike, or a pool measuring something unusual.

What is impermanent loss, and why does it cost points?

If you supply two assets to a liquidity pool and their prices move apart, you end up with more of the one that fell and less of the one that rose. The fees can outweigh it or they can not. It is a real risk a single-asset deposit does not carry, so it costs points and holds the score below the Strong tier.

Does a high score mean the protocol cannot be hacked?

No. Nothing on this page measures smart-contract risk, admin key control, oracle design or governance. Those are the failure modes that have caused most of the large losses in DeFi, and none of them can be read off a TVL number. A score measures what can be measured.

How current are these numbers?

APY, TVL and the pool statistics are live figures, refreshed when the page loads and every ten minutes after that. Protocol metadata, meaning listing date and audit status, is updated on its own schedule, and the header shows the date of the last update, because it changes on the scale of months.

Why do some pools have no score?

The protocol has no metadata on record at the source, so two of the six components cannot be scored. Rather than score it on the four we have and imply a precision that is not there, the page shows a dash and leaves it to you.