Copy Trading Wallet Rankings: Freeze Eligibility

Copy trading wallet rankings: Use only information available at the historical selection cutoff when ranking eligible wallets.

Wallet ranking research needs point-in-time eligibility before it needs a ranking formula. If the population was chosen using later activity, an otherwise chronological test still knows something about the future. Freezing scores does not repair a future-derived list of participants.

The Reddit discussion disclosed a full-history activity filter. Its results remain unverified; some replies and linked artifacts were unavailable to this review. The ledger below is an original synthetic exercise, not a wallet-copying recommendation.

Copy trading wallet rankings: record three different times

For each qualifying observation retain its event time, the time it became available to your system, and the selection cutoff. A trade can close before the cutoff but arrive afterward. In that case it was unavailable for the historical decision. Preserve both timestamps instead of silently substituting ingestion time for close time.

IETF RFC 3339 specifies an interoperable timestamp format. Formatting alone does not establish clock accuracy, source authenticity, or data availability. Store timezone offsets and the evidence supporting availability alongside the normalized value. A revised label needs its own version and availability record.

A synthetic eligibility trap

Imagine a month-end cutoff and a rule requiring ten closed observations already available. Participant A has twelve by the cutoff. Participant B has two, then reaches twenty next month. A full-history activity filter admits B retrospectively, while the historical rule excludes B at selection. A test built from the retrospective list evaluates a different selector.

Freeze the eligible population, including participants with no usable score, and retain the source IDs behind every score. In this example A qualifies and B does not. Do not erase B from the research database; record why B was excluded at that particular cutoff. B may become eligible in a later, independently frozen snapshot.

Distinguish absence from zero

During the subsequent evaluation window, a participant with no observed closed trades has an unknown measured outcome under a closed-trade scoring rule. Assigning a zero return can misleadingly make inactivity look like successful capital preservation. An observed zero net result and no observations should have different status fields.

Keep the asset or market cohort fixed. A score from one category does not automatically transfer to another. When comparing successive rankings, report entrants, exits, turnover, and the number with actual forward observations. A rank correlation computed on a surviving subset must identify that subset and its size.

Keep the selection experiment small

Choose the eligibility rule, score, observation window, and forward horizon before looking at the next outcomes. A new minimum count, shrinkage rule, or category filter is another candidate. Bailey and colleagues study search-driven overfitting. Preserve rejected variants and reserve subsequent data rather than presenting an adjusted selector as an independent confirmation.

At Tradewink Advisory Research, start with the offline ranking-snapshot guide. Build the two-participant fixture above and verify that adding future observations cannot change earlier eligibility or scores. Then test late-arriving closes and missing forward outcomes. These checks assess chronology and accounting; they do not establish profitable copying, executable capacity, or a production allocation policy.

Continue the research

Trading Strategy Replication: Resolve Conflicting Exits and Limit Order Backtests: Use Counterfactual Fill Ranges develop adjacent checks. Use a small synthetic fixture first; these educational diagnostics do not validate a trading edge or change account settings.