DR-DDEI-ARCH-0.1 · Mechanism Laboratory

DDEI INDEX ARCHITECTURE TOURNAMENT

Can Duplios construct a diversified, investable, self-reconstituting digital-economy index that survives its constituents? Research only. Not a live index. Not optimized for backtested return.

RESEARCH · SIMULATION · NOT INVESTMENT PERFORMANCE · NOT INVESTABLE · £0 REAL CAPITAL · PAPER

RESEARCH ONLY · NO LIVE DDEI · DR IS NOT A CONSTITUENT · LIVE_DDEI_PUBLISHED = FALSE · ARCHITECTURE FROZEN = FALSE · DDEI_EXCLUDES_DR = TRUE

Core principle

The constituents can die. The index survives. DDEI is the index. DR is the future tracking instrument. DR is never a DDEI constituent.

DDEI is the index. DR is the future investable asset that would economically track DDEI. This tournament optimizes representation, diversification, investability, survivability, low turnover, low concentration and long-term continuity. It does not select the highest historical return.

Headline size is a research ordinal/Zipf proxy with an era BTC-dominance overlay. It is not a vendor market-cap print and is not a forecast.

Historical reconstruction & backtest

Current status

Live DDEI: not published.
Holdings / PAPER / DR issuance: unchanged.
Price prediction: none. Price decline alone cannot trigger removal.
Size data: research ordinal/Zipf proxies with era BTC-dominance overlays. Not vendor prints.
Return metrics (CAGR, Sortino, path drawdown): INSUFFICIENT EVIDENCE — not used as selectors.
Ready to freeze architecture: NO.

Qualification gates (separate from weighting)

Free-float economic capitalization band
Minimum liquidity band
Market depth where available
Multi-venue pricing
Price confidence
Custody feasibility
Network operational status
Market history / seasoning
Manipulation / insider concentration
Economic functionality
Critical dependency concentration (post-weight)

Eligibility is not weighting. Missing free-float, liquidity, custody or seasoning is not filled in as 1.

Point-in-time snapshots

Later-failed names appear while they still existed. Current survival is not used to build the past. Dead: NXT, CEL, FTT, LUNA, UST, BCC, MULTI. Included while still alive: TRUE.

Id	Date	Split	Named ranks	BTC dominance proxy
2014-12	2014-12-31	DEVELOPMENT	11	82%
2016-12	2016-12-31	DEVELOPMENT	15	72%
2017-12	2017-12-31	DEVELOPMENT	25	46%
2018-12	2018-12-31	VALIDATION	22	52%
2020-03	2020-03-12	VALIDATION	22	64%
2021-05	2021-05-12	DEVELOPMENT	30	40%
2022-04	2022-04-30	DEVELOPMENT	40	41%
2022-11	2022-11-11	VALIDATION	30	38%
2023-03	2023-03-11	VALIDATION	30	42%
2023-07	2023-07-14	HOLDOUT	32	48%
2024-12	2024-12-31	HOLDOUT	40	54%
2025-12	2025-12-31	HOLDOUT	40	56%

Weighting families — 2022-04 panel

Pre-Terra panel still contains LUNA/UST. Literal n ≠ effective n. 20 names dominated by BTC is not 20 exposures.

Family	n	N_eff	Largest	Top 5	HHI	Repr.	Invest.	Max 1-asset zero	Max category zero
RAW MARKET CAP	30	4.29	BTC 43.1%	78.5%	0.233	91.0%	90.7%	43.1%	46.1%
FREE FLOAT CAP	30	3.45	BTC 49.1%	80.2%	0.290	100.0%	94.0%	49.1%	52.4%
CAPPED FREE FLOAT	30	16.27	BTC 10.0%	46.3%	0.061	50.2%	84.7%	10.0%	34.3%
SQRT FREE FLOAT	30	21.05	BTC 10.0%	36.8%	0.048	50.2%	83.1%	10.0%	31.7%
POWER FREE FLOAT	30	21.05	BTC 10.0%	36.8%	0.048	50.2%	83.1%	10.0%	31.7%
EQUAL WEIGHT	30	30.00	BTC 3.3%	16.7%	0.033	35.8%	76.8%	3.3%	26.7%
LIQUIDITY ADJUSTED	30	15.44	BTC 10.0%	47.4%	0.065	50.2%	85.7%	10.0%	34.4%
CATEGORY CONSTRAINED	30	16.27	BTC 10.0%	46.3%	0.061	50.2%	84.7%	10.0%	34.3%
QUALIFIED SMOOTHED	30	16.27	BTC 10.0%	46.3%	0.061	50.2%	84.7%	10.0%	34.3%

Power p sweep (free-float ^ p, 2022-04)

p	N_eff	Largest	Top 5	Representation	1-asset zero
0.35	22.93	10.0%	33.0%	49.0%	10.0%
0.50	21.05	10.0%	36.8%	50.2%	10.0%
0.65	19.58	10.0%	39.8%	50.2%	10.0%
0.75	18.50	10.0%	42.0%	50.2%	10.0%
0.85	17.38	10.0%	44.2%	50.2%	10.0%
1.00	16.27	10.0%	46.3%	50.2%	10.0%

Single-asset cap sweep (capped free-float, 2022-04)

Cap	N_eff	Largest	Top 5	Representation	1-asset zero
5.0%	25.11	5.0%	25.0%	40.2%	5.0%
7.5%	19.11	7.5%	37.5%	45.2%	7.5%
10.0%	16.27	10.0%	46.3%	50.2%	10.0%
12.5%	14.37	12.5%	50.8%	55.2%	12.5%
15.0%	12.76	15.0%	54.1%	60.2%	15.0%
20.0%	9.57	20.0%	60.7%	70.2%	20.0%
25.0%	7.04	25.0%	67.2%	75.9%	25.0%
NONE	3.45	49.1%	80.2%	100.0%	49.1%

Category cap sweep

Categories: NON_PROGRAMMABLE_SETTLEMENT · PROGRAMMABLE_SETTLEMENT · SCALING · FINANCIAL_INFRASTRUCTURE · INFORMATION_ORACLES · INTEROPERABILITY · APPLICATIONS. Relabelling correlated L1s as different sectors is not diversification.

Category cap	N_eff	Largest category	Representation
NONE	16.27	PROGRAMMABLE SETTLEMENT 34.3%	50.2%
40.0%	16.27	PROGRAMMABLE SETTLEMENT 34.3%	50.2%
50.0%	16.27	PROGRAMMABLE SETTLEMENT 34.3%	50.2%
60.0%	16.27	PROGRAMMABLE SETTLEMENT 34.3%	50.2%

Universe size (do not force n)

Target n	Mean literal	Mean eligible	Mean N_eff	Shortage
10	10.00	28.00	10.00	FILLED
20	20.00	28.00	16.34	FILLED
30	26.25	28.00	19.12	INSUFFICIENT QUALIFIED NAMES
50	28.00	28.00	19.72	INSUFFICIENT QUALIFIED NAMES
100	28.00	28.00	19.72	INSUFFICIENT QUALIFIED NAMES

Seasoning (2021-05)

Days	Literal n	Eligible	N_eff
30	25	25	18.99
60	25	25	18.99
90	25	25	18.99
180	25	25	18.99
365	19	19	15.62

Concentration dashboard — 2022-04

Smart baseline

Literal n: 30
N_effective: 16.27
Largest: BTC 10.0%
Top 2: 20.0%
Top 5: 46.3%
Top 10: 71.6%
HHI: 0.061
Largest category: PROGRAMMABLE SETTLEMENT 34.3%
Dependency HHI: 0.074
Loss if largest = 0: 10.0%
Loss if 2nd = 0: 10.0%
Loss if largest category = 0: 34.3%

Research candidate

Literal n: 30
N_effective: 21.05
Largest: BTC 10.0%
Top 2: 20.0%
Top 5: 36.8%
Top 10: 57.3%
HHI: 0.048
Largest category: PROGRAMMABLE SETTLEMENT 31.7%
Dependency HHI: 0.072
Loss if largest = 0: 10.0%
Loss if 2nd = 0: 10.0%
Loss if largest category = 0: 31.7%

Reconstitution path (research candidate)

Turnover is one-way weight distance between consecutive snapshots. Not a monthly PIT tape.

Snapshot	Split	n	N_eff	Largest	Top 5	Turnover
2014-12	DEVELOPMENT	7	7.00	14.3%	71.4%	—
2016-12	DEVELOPMENT	11	10.76	10.0%	50.0%	35.1%
2017-12	DEVELOPMENT	20	16.01	10.0%	43.8%	46.1%
2018-12	VALIDATION	19	15.44	10.0%	45.4%	19.8%
2020-03	VALIDATION	16	13.67	10.0%	48.7%	20.1%
2021-05	DEVELOPMENT	25	18.99	10.0%	39.5%	42.1%
2022-04	DEVELOPMENT	30	21.05	10.0%	36.8%	17.3%
2022-11	VALIDATION	25	18.18	10.0%	41.5%	22.9%
2023-03	VALIDATION	25	18.23	10.0%	41.5%	9.2%
2023-07	HOLDOUT	26	18.63	10.0%	41.0%	5.9%
2024-12	HOLDOUT	30	20.44	10.0%	38.3%	17.0%
2025-12	HOLDOUT	30	20.56	10.0%	37.7%	6.5%

Correlated crash (first-order)

Fully invested long-only index with correlations → 1. Diversification of names does not remove crypto-systemic risk. Not S&P-like absolute safety.

−20% → -20%
−40% → -40%
−60% → -60%
−80% → -80%

Emergency removal vs price

Not a tactical engine. Normal volatility: stay invested. Role: qualification, dependency, venue/custody monitoring, structural disqualification, emergency reconstitution. No price predictions.

Routine reconstitution is not emergency removal. Price fall alone: NO ACTION.

Failure replay

Not every crisis is a constituent death. Holdout cases are not used to pick the architecture.

Case	Kind	Split	Action
Mt. Gox insolvency	VENUE FAILURE	DEVELOPMENT	VENUE NOT ASSET
The DAO exploit	PROTOCOL FAILURE	DEVELOPMENT	EMERGENCY REMOVE
Bitfinex hack	VENUE FAILURE	VALIDATION	VENUE NOT ASSET
Bitcoin 2018 bear	SYSTEMIC MARKET FALL	VALIDATION	NO ACTION
COVID crypto crash	SYSTEMIC MARKET FALL	VALIDATION	NO ACTION
Wormhole	PROTOCOL FAILURE	HOLDOUT	NOT A CONSTITUENT
Ronin	PROTOCOL FAILURE	VALIDATION	EMERGENCY REMOVE
Terra / UST / LUNA	ASSET FAILURE	DEVELOPMENT	EMERGENCY REMOVE
Three Arrows Capital	CUSTODIAN FAILURE	VALIDATION	VENUE NOT ASSET
Celsius	CUSTODIAN FAILURE	VALIDATION	VENUE NOT ASSET
Voyager	CUSTODIAN FAILURE	VALIDATION	VENUE NOT ASSET
Nomad	PROTOCOL FAILURE	HOLDOUT	NOT A CONSTITUENT
FTX / FTT	ASSET FAILURE	DEVELOPMENT	EMERGENCY REMOVE
BlockFi	CUSTODIAN FAILURE	VALIDATION	VENUE NOT ASSET
USDC / SVB	DEPENDENCY FAILURE	VALIDATION	WATCH
Multichain	PROTOCOL FAILURE	HOLDOUT	EMERGENCY REMOVE
Bybit 2025	VENUE FAILURE	HOLDOUT	VENUE NOT ASSET
ETH 2022 drawdown (healthy control)	SYSTEMIC MARKET FALL	VALIDATION	NO ACTION

SHIB and similar economicFn=NO names are excluded by qualification, not by price. That is a design choice, not a false removal of a fallen-but-healthy settlement asset.

Holdout vs development

Candidate N_eff development/validation mean 16.58 · holdout mean 19.88. Simple baseline holdout N_eff 16.13. Rules were not tuned on holdout.

Metrics not used as selectors

CAGR — INSUFFICIENT EVIDENCE
volatility of returns — INSUFFICIENT EVIDENCE
max drawdown of a price path — INSUFFICIENT EVIDENCE
Sortino — INSUFFICIENT EVIDENCE
downside deviation of returns — INSUFFICIENT EVIDENCE
upside participation vs BTC path — INSUFFICIENT EVIDENCE
failure loss avoided in portfolio P&L — INSUFFICIENT EVIDENCE

Cost illustration uses 0.3% round-trip × measured snapshot turnover. Not a broker quote.

Frontier (not a freeze)

Best weighting family: SQRT_FREE_FLOAT is the highest structural-score family on development/validation panels after excluding equal weight (equal weight maximises N_eff and is rejected for representation/investability). CAPPED_FREE_FLOAT remains the smart baseline. INSUFFICIENT EVIDENCE to freeze.
Best p: 0.5 (sqrt-cap / power on structural score, not return). 0.35 is shown in the grid and is rejected as quasi-equal in the tail.
Best single-asset cap: 10.0% (research candidate)
Best routine rebalance: QUARTERLY research candidate. Monthly PIT panel is not in the lab — INSUFFICIENT EVIDENCE to freeze monthly. Fast-entry for top-3 seasoned multi-venue names is a research overlay, not live.
Best seasoning: 90 days research candidate. 365 days starves early eras. 30 days admits unseasoned issuance. INSUFFICIENT EVIDENCE to freeze.
Constituent count range: 10–28 qualified names on tested panels. Do not force 100 when fewer qualify.
Average effective constituents: 16.58
Largest historical weight: 14.3%
Top-5 concentration (path max): 71.4%
Max single-asset zero loss: 14.3%
Systemic −60%: -60%
Failure recall: 100%
False removal rate: 0%
Turnover (mean snapshot): 22.0%
Cost estimate: 0.07%
Upside participation: INSUFFICIENT EVIDENCE — no non-survivorship-biased total-return panel. BTC-weight on a capped index is the structural upside-to-BTC ceiling, not a backtest.
Best simple benchmark: Capped free-float, 10% single-asset cap, ~30 names, 90-day seasoning, quarterly reconstitution (smart baseline).
Complex DDEI beats simple capped index: INSUFFICIENT EVIDENCE

No vendor point-in-time free-float cap panel is stored. Size is an ordinal/Zipf research proxy. Returns, Sortino, and tracking error are INSUFFICIENT EVIDENCE and were not used to pick a winner. Holdout paths exist but rules are not frozen. Live DDEI is not published. Paper holdings are untouched.

DDEI ARCHITECTURE READY TO FREEZE: NO

← Mechanism Lab · DDEI status