> ## Documentation Index
> Fetch the complete documentation index at: https://docs.beliefsystems.xyz/llms.txt
> Use this file to discover all available pages before exploring further.
# Market Selection & Weighting
> The published, rules-based criteria every prediction market must pass to enter a Belief Index series – and the formula that sets each constituent's weight.
**Criteria current as of:** July 2026
Markets enter a Belief Index series by passing a published, rules-based screen – not because someone liked them. This page specifies the eligibility criteria a prediction market must meet to enter a series, and the formula that determines how much weight it carries once admitted.
This page covers *which* markets enter an index and *at what weight*. For how admitted markets are
valued, see [NAV Methodology](/indices/nav-methodology). For how composition evolves over time,
see [Perpetual Series](/indices/perpetual-series).
**Published for transparency, not as a binding commitment.** These criteria describe how Belief
Systems currently constructs and reconstitutes its indices. They are forward-looking guides:
Belief Systems may revise them, or depart from them in a specific case, at its discretion. Some
live series also predate the current rule set, so readers may observe compositions that would
not result from applying today's criteria – that drift is expected, not an error.
## One Rule Set, Applied Twice
There are two moments a market can enter a series:
1. **At series creation**, when the initial composition is assembled.
2. **At reconstitution**, when a Perpetual series adds markets on its published [review cadence](/indices/perpetual-series#review-cadence).
Both paths read from a single canonical rule set – the same thresholds, the same weighting formula. A market that qualifies at creation would qualify at reconstitution, and vice versa. Any divergence between the two paths is treated as a defect, not a feature.
This is the same discipline that makes an equity index investable: the rules *are* the methodology, and the methodology does not bend per event.
## Eligibility Screens
A candidate market must pass **every** screen below. These are hard thresholds, applied identically to every candidate.
| Screen | Current threshold | Why it exists |
| -------------------------- | --------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Order book depth** | ≥ \$3,000 across the top five price levels, both sides combined | Depth is what makes a midprice trustworthy and a position enterable at reasonable size. Thin books produce noisy NAV and expensive rebalances. |
| **Bid-ask spread** | ≤ \$0.02 | A tight spread indicates healthy two-sided participation. Wide spreads mean the midprice is an average of disagreement, not a price. |
| **24-hour trading volume** | ≥ \$1,000 | Depth without volume can be stale quotes. Volume confirms active price discovery is actually happening. |
| **Time to resolution** | Between 30 days and 10 years | A market about to resolve would enter the index and exit almost immediately, churning composition for no signal. The upper bound guards against mis-dated markets, not against genuinely long-dated ones. |
| **Price level** | Midprice between \$0.05 and \$0.95 | Markets near certainty carry little remaining information about the theme – they behave like cash with tail risk rather than a live probability signal. |
| **Market state** | Active and open for trading | A market must be live and tradeable to be held. |
| **Category screens** | Sports and entertainment markets excluded | Belief indices measure macro, geopolitical, and policy expectations. High-volume recreational categories are screened out regardless of liquidity. |
Thresholds are stated as current values. They are revised deliberately and prospectively – see
[How Thresholds Are Set](#how-thresholds-are-set-and-revised) below. A revision never
retroactively changes an existing composition.
## Judgment Screens
Quantitative screens establish that a market is *investable*. A second set of screens, applied through human review, establishes that it *belongs*:
| Screen | What it means |
| ------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Directional fit** | Every constituent is held on an explicit side – YES or NO – chosen so that the position gains when the index's stated direction of expectations strengthens. If a market cannot be cleanly mapped to the index's direction, it is dropped. |
| **Binary outcomes only** | Markets with scalar or multi-outcome structures ("how many X will happen") are excluded. Directional indices require unambiguous binary exposure. |
| **One horizon per event** | When the same question trades at multiple resolution dates, exactly one is selected – generally the longest liquid horizon, which survives more review cycles. |
| **Resolution clarity** | Settlement rules must be unambiguous and verifiable. Markets with subjective or disputed resolution criteria introduce oracle risk and are avoided. |
| **Theme diversification** | No single sub-theme should dominate the basket. An index measuring a broad risk domain should not quietly become an index of one situation. |
A series targets **5 to 20 constituent markets**. Below five, a single market's resolution moves the index too violently for "diversified" to mean anything. Above roughly twenty, additions tend to either duplicate existing exposure or reach below the quality bar. A narrow theme with seven strong candidates makes a perfectly good basket; a broad theme with thirty is trimmed for clarity.
## Weighting
Admitted markets are weighted by a published formula rather than discretion:
```
raw_score = sqrt(order_book_depth × 24h_volume)
weight = raw_score / sum(all raw_scores)
```
**In plain English:** a market earns weight by being both deep *and* actively traded. The geometric mean means neither input can carry the other – a deep but dormant market, or a busy but shallow one, earns less weight than a market that is both. The square root dampens outliers so a single very large market cannot mechanically dominate the basket.
Two adjustments are applied to the natural weights:
Any market whose weight exceeds 50% is capped there, and the excess is redistributed
proportionally across the uncapped markets. This repeats until no market is over the cap. The
cap is anti-concentration: an index should never be one market wearing a costume.
Any market whose natural weight computes below 1% is **excluded entirely** and the survivors are
renormalized. This is an inclusion rule, not a weight clamp – survivors keep their relative
weights. The floor is anti-fragmentation: a position too small to hold materially adds
operational surface without adding meaningful exposure.
Final weights are normalized to sum to exactly 1.0. Both inputs – depth and volume – are observed at composition time, so weights are a point-in-time snapshot. They do not drift with market conditions between reconstitutions; see [Known Limitations](/indices/nav-methodology#known-limitations) in NAV Methodology.
Three candidate markets pass all eligibility screens:
| Market | Order book depth | 24h volume | raw\_score = sqrt(depth × volume) | Natural weight |
| ------ | ---------------- | ---------- | --------------------------------- | -------------- |
| A | \$40,000 | \$20,000 | 28,284 | 68.2% |
| B | \$10,000 | \$10,000 | 10,000 | 24.1% |
| C | \$4,000 | \$2,500 | 3,162 | 7.6% |
Market A's natural weight (68.2%) exceeds the 50% cap. It is capped at 50%, and the remaining 50% is split between B and C in proportion to their raw scores:
| Market | Final weight |
| ------ | ------------ |
| A | 50.0% |
| B | 38.0% |
| C | 12.0% |
No market falls below the 1% investability floor, so no exclusions apply. In a larger basket, a market whose natural weight computed to, say, 0.4% would be dropped and the remaining weights renormalized.
Each series records its weighting scheme in its methodology, displayed on the series detail page.
The formula above is the standard scheme for series created and reconstituted under the current
methodology; some earlier series record a simpler scheme, such as equal weighting.
## How Thresholds Are Set and Revised
The numeric thresholds are not arbitrary. They are derived empirically from the distribution of markets actually held by live Belief indices – anchored near the conservative end of that distribution, so the screens encode what has historically made a good constituent rather than a hopeful guess.
Revisions follow the same discipline as the rest of the methodology:
* Thresholds are reviewed on a deliberate cadence, not tuned ad hoc.
* Any revision applies **prospectively** – to future selections, never to existing compositions.
* The current values are published on this page, and the "current as of" date at the top reflects the last review.
## What These Rules Do Not Decide
* **Removals.** Markets are never removed by discretion. Constituents exit a series only by resolving – see [Perpetual Series](/indices/perpetual-series) for how settled markets leave the basket.
* **Valuation.** How an admitted market is priced into NAV is specified separately in [NAV Methodology](/indices/nav-methodology).
## Further Reading
How composition evolves through reconstitution, and how costs are disclosed.
How constituent prices become an index level – transparent and replicable.
Foundational concepts: composition, normalization, lifecycle.
Risks including concentration, correlation, resolution, and reconstitution.