Skip to content

Rating log15 Nov 2026Edition 14 min read

One hour after a High rating, 13% of coins stood 70% or more lower.

Every rating of 19 days, checked one hour later, for all five classes. After a Low rating it was 5%.

Data counted on 15 Nov 2026, recording from 25 Oct 2026

Summary

  • 73,447 ratings from 28 Oct 2026 to 15 Nov 2026 were checked one hour later.
  • Across all classes, 9.1% stood 70% or more lower; after a High rating 13.4% (95% interval 12.8% to 13.9%).
  • Very high rests on 184 checked coins, the fewest of any class, so its share is the least certain.

13%

of coins rated High stood 70% or more lower one hour later. The other part had not fallen 70%.

Every rating is checked one hour after it was made, whatever the result. This is the count for all five classes, the misses included.

Stood 70% or more lower one hour latern = 73,447 ratings
  • Low5.5%
  • Moderate10.8%
  • Elevated8.6%
  • High13.4%
  • Very high10.3%
RatingChecked70% or more lowerNot 70% lowerShare lower95% interval
Low11,42862610,8025.5%5.1% to 5.9%
Moderate3,5653843,18110.8%9.8% to 11.8%
Elevated45,2973,91341,3848.6%8.4% to 8.9%
High12,9731,73211,24113.4%12.8% to 13.9%
Very high1841916510.3%6.7% to 15.6%
All classes73,4476,67466,7739.1%8.9% to 9.3%
Every rating from 28 Oct 2026 to 15 Nov 2026, market cap one hour after the rating against the cap at the rating. Interval: 95% Wilson interval of the share.

The share does not rise evenly with the class: Low 5.5%, Moderate 10.8%, Elevated 8.6%, High 13.4%, Very high 10.3%. Very high rests on 184 coins, so its interval is wide. The rating names a risk, it is no forecast of the next hour.

Compared with

The yardstick is the share for all checked ratings together: 9.1% stood 70% or more lower one hour later (95% interval 8.9% to 9.3%). A class says something only where it differs from that. After a High rating the share was 1.47 times the base rate, after a Low rating 0.60 times.

Each class against all classesbase rate 9.1%
  • Low5.5%
  • Moderate10.8%
  • Elevated8.6%
  • High13.4%
  • Very high10.3%
  • All classes9.1%
RatingShare lowerAll classesRatio
Low5.5%9.1%0.60×
Moderate10.8%9.1%1.19×
Elevated8.6%9.1%0.95×
High13.4%9.1%1.47×
Very high10.3%9.1%1.14×
Ratio: the class's share divided by the share of all classes together.

Method

Outcome archive of the hub: for every rated coin the market cap at the rating and at one hour. Counted as lower when the one hour cap was at most 30% of the cap at the rating. Coins without a one hour value are left out (97 in this edition).

Words such as priced, decided and lower are defined in the methodology.

Query ratingLogproof/, outcomes-archive/
import json, glob, gzip, collections
per_day = collections.defaultdict(collections.Counter)
for f in sorted(glob.glob("/var/lib/meterm-hub/proof/proof-*.jsonl")):
    d = f.split("proof-")[1][:10]
    for line in open(f):
        try: per_day[d][json.loads(line).get("risk") or "none"] += 1
        except Exception: pass
cls = collections.defaultdict(lambda: [0, 0, 0])
days = []
for f in sorted(glob.glob("/var/lib/meterm-hub/outcomes-archive/outcomes-*.jsonl.gz")):
    days.append(f.split("outcomes-")[2][:10])
    for line in gzip.open(f, "rt"):
        try: o = json.loads(line)
        except Exception: continue
        h1, mc0 = (o.get("checks") or {}).get("h1"), o.get("mc0")
        c = cls[o.get("risk") or "none"]
        if not h1 or not mc0: c[2] += 1
        elif h1 <= 0.3 * mc0: c[0] += 1
        else: c[1] += 1
print(json.dumps({"ratings": {d: dict(c) for d, c in per_day.items()}, "classes": {k: {"fell": v[0], "held": v[1], "open": v[2]} for k, v in cls.items()}, "days": days}))

Counted on the production hub on 15 Nov 2026. Script: scripts/build-labs-research.mjs, reading only, sums and distributions only.

Limits

A coin rated near the launch floor cannot fall 70% as easily, so its result says little; this edition does not split the classes by market cap. One hour is one snapshot: a coin can fall later. Very high rests on 184 coins only.

Data

Every table of this report as a CSV file. Sums and distributions only, the same rows as on this page.

History and citation

15 Nov 2026
Edition 1.
16 Nov 2026
Base rate of all classes, 95% intervals, summary, data files and citation added. The numbers are unchanged.

Cite as

CON Labs, Rating log. "One hour after a High rating, 13% of coins stood 70% or more lower." Edition 1, 15 Nov 2026. https://www.centralonet.com/labs/research/ratings-one-hour-later

Found an error? Corrections lists every change to data and reports.

2026 · Rating log · Author: CON Labs