Joshua Della Vedova

DV-PMI · Insider screen

Private Information Index (PII)

Of 483,234 wallets tested, 6,292 were flagged at the one-percent significance threshold (flag rate 1.30%). 806 wallets survive the Holm-Bonferroni family-wise correction; 2492 survive the Benjamini-Hochberg FDR procedure. The flag rate ordering across MNPI categories — Action highest, Stochastic lowest — confirms the theoretical prediction that markets where humans can know outcomes in advance show stronger informational signatures.

Categories ordered by a priori MNPI risk: vote markets (elections, awards) are the highest-risk category because outcomes are determined by discrete human decisions; stochastic markets (crypto prices) are the lowest-risk because outcomes are aggregate and leak-resistant.

Flag rate by MNPI taxonomy

MNPI categoryWallets testedFlagged (p < 0.01)Flag rate
Vote 90,797 2,534 2.79%
Action 78,548 755 0.96%
Performance 157,755 1,598 1.01%
Stochastic 134,972 1,613 1.20%

By wallet class

Wallet classWalletsFlaggedFlag rate
Algorithmic 64,385 1,026 1.59%
Sophisticated 64,954 1,353 2.08%
Active Retail 353,895 3,913 1.11%

Comparison with Mitts and Ofir (2026)

The concurrent paper by Mitts and Ofir uses a five-signal heuristic (cross-sectional bet size, within-trader bet size, profitability, pre-event timing, directional concentration) to flag suspicious wallet-market pairs. Applied to the same sample, their approach flags roughly 38.0% of wallets; our microstructure-grounded orthogonality test flags 1.30%. Cohen's kappa between the two flag sets is 0.007. Near-zero concordance indicates the two approaches identify different populations: ours isolates wallets with the structural signature of information (accuracy and execution coupled to a common signal); theirs isolates wallets with large, winning, concentrated bets near event resolution.

Methodology

For each wallet, excess accuracy is defined as accuracy minus max(price, 1 - price), the accuracy achievable by mechanically following the price signal. A one-sided binomial z-test at threshold z > 2.326 (p < 0.01) flags wallets whose accuracy cannot be explained by price-following alone. The theta-epsilon orthogonality violation (a positive correlation between accuracy and execution at the wallet level, not present in the population-level orthogonality established in Paper 1) distinguishes informed traders from lucky ones. MNPI-category assignment uses a rules-and-LLM pipeline validated against hand-coded markets (vote, action, performance, stochastic).

Weekly time series for PII is a future release item; it requires a separate pipeline joining the flagged wallet set with per-wallet trade timestamps across the 671M-row trade panel.

Source paper

Della Vedova, J. (2026). Detecting Informed Trading in Prediction Markets: An Orthogonality Test. SSRN.

Related surveillance indices

PII is the first of eight DV-PMI surveillance indices. The companion indices test for pre-resolution timing concentration, spread patterns in markets with flagged wallets, and the response of flagged-wallet excess accuracy to resolution surprises. See the full surveillance suite →

Data

pii_snapshot.csv · pii_latest.json

Cite this index

@misc{dellavedova2026pii,
  title        = {Private Information Index},
  author       = {Della Vedova, Joshua},
  year         = {2026},
  url          = {https://jdellavedova.com/pii}
}