Here are two tokens from the same corner of the market, read off our record on October 23, 2026.

Optimism (OP) carries a grade of F / Failed, with a score of 52.8. It is excluded, so our ranking does not give it a position at all.

POL (ex-MATIC) carries a grade of C / Speculative, with a score of 51.1, at ranking position 32.

Both are Ethereum scaling assets. Both were last scored at the same moment in our record. OP scores 1.7 points higher than POL. OP is the one with the F.

If you hold either of these, that is the kind of thing that makes a rating look broken. It is not broken. It is the framework doing the specific job it was built to do, and this is a good week to walk through it, because the two assets are close enough together that nothing else can explain the gap.

Our Friday piece is normally a read of an outside incident. The four news sources this routine is cleared to read all refused it this morning, so we did what our own rules tell us to do in that case: take the incident from our own dated record instead, and say so. That is the designed second path, not a shortcut. Everything below is first hand from our own published record, with the date it was read.

The first thing to know is that we do not calculate the letter from the score. They are two separate things our model publishes, and a reader who treats the letter as a label stuck on top of the number will misread a case like this one every time.

The score is a composite out of 100. The bands are written down and they are public: A is 80 and above, B is 60 to just under 80, C is 40 to just under 60, and F is below 40.

Run OP's 52.8 through those bands and you land in the C band. Run POL's 51.1 through them and you land in the C band as well. POL carries the C. OP does not.

So the bands are not the whole story. They are not the only thing that sets the letter: a tripped screen places an asset in F at any score, including a score in the A band. That is the second road into an F, and it is the one that matters here.

An F in our record means one of two things happened.

• The asset scored below 40.

• Or the asset failed a screen, at whatever score it was carrying.

A low score is not the only road into an F. Our record does not say which road a given asset took, and we do not guess. We are not going to tell you which screen OP did or did not clear, because our published record does not carry that field yet, and inventing one would be worse than leaving the question open.

What we can tell you, flatly, is the consequence. An F in our record is an exclusion rather than a bad place in the queue.

This is the part that trips people up, and it is worth being precise about.

POL has a ranking position: 32, on the October 23, 2026 record. OP does not have one. Our ranking does not place an excluded asset at the bottom. It leaves it out.

That distinction matters more than it sounds. "Ranked 32nd" is a statement about where an asset sits relative to everything else we rate. "Excluded" is not a worse version of that statement. It is a different statement: our model declined to place the asset in the ordering at all. Somebody scanning a list from the top would never reach OP, not because it finished behind the others, but because it is not in the list.

You can hold an asset we exclude. People do, in size. What you cannot do is read its absence from a ranking as a quiet place near the bottom of one.

The letters are the headline. The research record underneath them is where the reasoning lives, and on these two it reads as a pair of near-identical complaints about near-identical strengths.

On Optimism, our model's standing assessment records real activity. Its usage note reads: "OP Mainnet shows meaningful network activity: ~$433.6M TVL and $4.62M 30d DeFi fees, while the OP Stack repo shows strong development activity with 645 commits and 54 contributors over 90d. Usage is real, though not top-tier." Those figures were written into our record when the asset was scored and are not a live reading. On its ecosystem, our model calls OP "a major L2 ecosystem with a strong OP Stack developer base" whose evidence "supports durable network effects and competitive positioning."

Then the limit, and our model puts it in plain terms: "Tokenomics remain mediocre because the evidence does not show strong fee/burn accrual to holders." On valuation, "the supplied data do not show exceptional fees or usage relative to market value."

On POL, almost the same shape. Usage: "Polygon PoS shows meaningful on-chain activity: TVL about $713.5M and 30-day DeFi fees about $36.5M. That supports real network usage, though the evidence does not separate fully organic demand from incentives." Again, recorded at scoring time, not a live reading. Then the limit: "the supplied evidence does not establish fee/burn capture or strong holder accrual. Token value linkage remains limited relative to usage."

Read those side by side. Two networks with genuine activity on them. Two tokens our model cannot connect to that activity on the evidence in front of it. The composite scores land 1.7 points apart, which is about as close as two separate assessments get.

And the letters are opposite. That is precisely the information the letter is carrying that the score is not.

Here is the practical version, with no advice in it.

If you own a scaling token and you have been watching its score, the score is telling you how our model weighed the evidence. It is not telling you whether our model is willing to rank the asset at all. Those are two separate readings, published side by side, and in this pair they point in opposite directions.

Our model also states an impersonal limit at each letter. On October 23, 2026, our model's maximum single position read 0% for the F-graded asset and 2% to 3% or less for the C-graded one. Those are grade-level limits our model states for every asset carrying that letter. They are not a statement about your portfolio, your goals or your holdings, and we would not make one.

We would rather be the research company that marks its own edges.

Our record does not publish why OP carries an F rather than a C. It does not publish which screen was involved, or whether a screen was involved at all rather than a scoring outcome. It does not publish a per-dimension score you can add up yourself. Between our October 22 and October 23 records, by the way, nothing about either asset's score, grade or position changed, so there is not even a movement here to misattribute.

Being able to tell you why a grade sits where it does is a thing we are building, not a thing we are refusing. Until it is in the published record, the honest answer is that the letter and the score are both ours, both dated, and the gap between them on this pair is real.

The useful version of all of this is not reading it again next month. It is being told the morning it stops being true.

That is Alerts: pick the crypto assets you care about and ask to be told when our letter grade on one of them changes. It is free with a free CryptoGrade account. It is our grade on the assets you chose, and it is the one thing in this piece you cannot get from a chart, because what changes is our assessment rather than the asset's price.

On this pair specifically, the thing worth hearing about is not 52.8 or 51.1 on any given morning. It is the morning one of those letters reads something else.

CryptoGrade does not work commercially with the assets or projects it covers, and it does not sell data. Nobody we grade pays us. That is the only reason a piece like this can name two well-known tokens, print both letters, and then print our own model saying that neither one establishes value accrual to holders.

Both assets have a page if you want to check our work: Optimism at https://cryptograde.ai/tokens/optimism and POL at https://cryptograde.ai/tokens/pol-ex-matic . The framework both were put through is written up at https://cryptograde.ai/framework