On Nimb, "Moved together · 30D" and "Moved opposite · 30D" list markets whose prices rose and fell in step, or in opposite directions, over the past 30 days, and passed statistical checks against noise. They describe the past month; they are not a forecast, and a market that moved opposite is not automatically a hedge.
These lists help market reading by pointing to markets that may react to the same news. Below: the exact rule, why short windows produce false links, and Nimb's own estimate of how often links are wrong.
What does "Moved together · 30D" mean?
Two markets moved together when their price changes over the past 30 days had a positive correlation strong enough to pass Nimb's checks; they moved opposite when the correlation was negative and strong enough. Correlation (ρ, from −1 to +1) measures how consistently two series move in the same direction at the same time. It says nothing about how far they move or why.
The lists appear in the "How it connects" section of the market panel, with up to four markets on each side. A bar shows the strength of each link; hover over it to see ρ and the number of 12-hour steps it rests on, for example "ρ 0.62 over 41 twelve-hour steps". If nothing passes, the panel says "None that passes our checks." The comparison uses each event's lead outcome.
What are the thresholds?
A pair of markets is shown only if every check passes:
| Check | Rule |
|---|---|
| Window | Past 30 days, in 12-hour steps (at most 60 steps) |
| Price changes | Measured in log-odds, so a move from 50% to 55% and one from 90% to 93% are compared fairly |
| Fresh steps only | A step counts only if the price actually moved, Polymarket's price history recorded a new price in that step (not one carried forward) and both ends were between 2% and 98% |
| Joint observations | At least 30 steps where both markets had a fresh move |
| Strength | |ρ| ≥ 0.5 |
| Leave-one-out | Removing the single step that contributes most must leave |ρ| ≥ 0.5 with the same sign |
Nimb's own analysis puts the expected false-discovery rate at about 20% for |ρ| ≥ 0.5 and about 10% for |ρ| ≥ 0.7. Two stricter checks are not applied: a rank-based (Spearman) cross-check and a Benjamini–Hochberg correction for running many tests at once.
Why only on 30D?
Because short windows lie. Two random price series will often look related over a few observations. Nimb's own tests measured this with a "circular-shift" null: it slid one series against the other in time, 200 times, to see how many links appear by chance, and compared that with what the earlier rules flagged.
| Timeframe | Window | Pairs flagged | Expected false share | Median joint obs. |
|---|---|---|---|---|
| 1H–3D | 72 hours, 1-hour steps | 151 | 61% | 8 |
| 7D | 168 hours, 3-hour steps | 212 | 71% | 15 |
| 30D | 720 hours, 12-hour steps | 299 | 43% | 17 |
Under the old rules, 61–71% of the links shown on windows up to 7 days were expected to be false positives, and 43% on 30D. Between 81% and 87% of flagged pairs rested on a single observation: one shared jump, often a news day or a sports kick-off. Most prediction-market prices sit flat for long stretches.
The current rule fixes this by requiring 30 fresh joint observations, which only the 30D window can supply. That is why the panel on every shorter timeframe says links "appear on the 30D view, where there is enough price history to tell them apart from noise". See timeframes.
How many of these links are false positives?
Fewer than before, but not none. With |ρ| ≥ 0.5 and 30 joint observations, Nimb's tests suggest about 1 in 5 shown links to be chance; at |ρ| ≥ 0.7, about 1 in 10. These are estimates from one month of data, not certainties. Treat a link as a prompt to check both markets, not as a finding.
Is a negative link a hedge?
Not by itself. "Moved opposite" means the two prices went in opposite directions over the past month. It does not mean they must keep doing so, or that a position in one offsets a loss in the other at resolution. Two markets can move opposite because traders reacted to the same headline, and then resolve in a way that makes both positions lose.
Nimb reserves the word "hedge" for logical links, where the payouts are tied by the rules of the markets themselves:
| Logical link | Example | Why it holds |
|---|---|---|
| Date ladder | "By March" and "by June" on the same event | If it happens by March, it also happened by June |
| Thresholds | "Above $100k" and "above $120k" on the same asset and date | Above $120k implies above $100k |
| Mutually exclusive set | The candidates in one multi-outcome event | Only one can win, so chances should sum to about 100% |
These links do not need statistics: they follow from how the markets resolve. Multi-outcome sets are explained in multi-outcome markets. The panel does not mark these links, so check them yourself.
Correlation or logical link: which should I trust?
A logical link tells you something about payouts; a correlation only tells you about past prices. If two markets are logically linked and their prices disagree (a "by March" market priced above "by June"), the gap is worth reading, though fees, spreads and thin books often explain it. If two markets merely moved together, look for the shared cause in the news before assuming one will follow the other.
Can I trade on a correlation?
You can, but a 30-day correlation is weak evidence on its own. Some readers treat a strong link between a moving market and a quiet one as possible alpha, an edge the price does not yet reflect; most apparent edge of this kind is noise or is absorbed by fees and spreads. A correlation can break the day the shared cause stops mattering, and with about 60 steps of history even a real link is measured imprecisely.
What can this not tell you?
- Why the markets moved together, or whether they will next month.
- How much one moves when the other does: ρ is not a slope.
- Whether a pair is a hedge at resolution.
- Anything about markets with too few fresh trades: a quiet market cannot show a link, even if one exists.
A grouping by topic, rather than by price, is on themes and factors. Open problems are listed in known limitations, and terms in the glossary.
From reading to execution
Use the lists to find markets worth reading next to the one you are on, then read each one before you use the order ticket. Trading involves risk; see the risks of prediction markets.
Last reviewed 2026-11-07 · Data from Polymarket · Not financial advice · Nimb is not affiliated with Polymarket.