DV-PMI Surveillance · Timing

Insider Timing Concentration

Cumulative snapshot through 2026-02-28

Wallets flagged by the orthogonality test enter markets at a mean of 29.3 days from resolution, compared to 2.0 months for the non-flagged population. The flagged ratio is 48 percent of the population mean. Concentration in the final 72 hours: 2.03% of flagged-wallet trades versus 1.62% of non-flagged. The pattern is consistent with late-window information leakage; it does not by itself establish material non-public information.

Days-to-resolution	Flagged EA	Non-flagged EA
0-12h	0.031	-0.469
12-24h	0.184	-0.480
24-72h	0.051	-0.335
72h-7d	0.076	-0.280
7d-30d	0.192	-0.355
30d+	0.148	-0.463

Excess accuracy by time-to-resolution bucket. Flagged wallets show small positive excess accuracy in every window; the non-flagged population is sharply negative because the population accuracy is bounded above by the price-implied benchmark.

Headline numbers

	Flagged wallets	Non-flagged wallets
Sample size	6,032	444,016
Mean hours to resolution	704	1460
Median hours to resolution	378	1110
Fraction in final 72 hours	2.027%	1.617%

Methodology

For each wallet, we measure the mean and median hours between trade timestamp and the market's resolution timestamp. The "flagged" population is the set of wallets whose excess accuracy survives a one-sided binomial z-test at p < 0.01 (the Paper 2 PII screen). The non-flagged population is every other wallet that meets the minimum-trades threshold. The metric reports a pattern; it does not establish intent or material non-public information.

Known false-positive sources: (a) flagged wallets that are short-horizon algorithmic strategies rather than informed traders will mechanically trade closer to resolution; (b) markets with short natural windows (sports, single-day events) compress the entire population toward resolution regardless of information.

Limitations

Hours-to-resolution is computed from on-chain block timestamps. Sub-second precision is not available. The metric is wallet-level mean, not trade-level; a wallet that places one early trade and many late trades is still represented by a single mean. This is the convention in Paper 2; an alternative trade-weighted decomposition is a candidate revision item.

Source paper

Della Vedova, J. (2026). Detecting Informed Trading in Prediction Markets: An Orthogonality Test. SSRN. Section: Pre-Resolution Timing Test.

Data

surveillance_insider_timing_latest.json

Other surveillance indices

← All surveillance indices · Informed Trading (PII) · Adverse Selection · Resolution Surprise · Wash Trading

Cite this index

@misc{dellavedova2026timing,
  title        = {Insider Timing Concentration Index},
  author       = {Della Vedova, Joshua},
  year         = {2026},
  url          = {https://jdellavedova.com/surveillance/insider-timing}
}