Live

Machine Learning & Similarity

Conyr distills each qualifying wallet’s full trading history into a behavioral fingerprint — a vector that captures how a wallet trades, not just its PnL. Read a wallet’s profile, or run nearest-neighbor search to find wallets that behave like it.

What the fingerprint captures

The profile summarizes a wallet across six behavioral dimensions:

Hold timing — how long positions are held and how exits are paced
Exit discipline — how cleanly winners and losers are closed
Performance — return shape and risk-adjusted consistency
Velocity & sizing — trading cadence, position sizing, and diversification
Infrastructure — venue and automation patterns
Provenance — funding origin and account history

Each dimension is reduced to normalized behavioral scores; similarity is computed over the resulting vector.

A wallet needs enough closed-position history to be profiled. Below that threshold has_vector is false and similarity is unavailable — Wallet Quality and Labels still work at lower volume.

Profile

L2 GET /v1/wallet/{address}/profile

Returns the wallet’s taxonomy plus its behavioral scores, each on a normalized 0–1 scale (with the underlying raw value).

curl -H "Authorization: Bearer $API_KEY" \
  "https://api.conyr.ai/v1/wallet/7xKXtg.../profile"

{
  "wallet": "7xKXtg...",
  "has_vector": true,
  "trader_type": "DAY_TRADER",
  "performance_level": "PROFITABLE",
  "features": {
    "win_rate":        { "raw": 0.63, "normalized": 0.72 },
    "profit_factor":   { "raw": 2.1,  "normalized": 0.68 },
    "trade_frequency": { "raw": 12.5, "normalized": 0.55 }
  },
  "trade_badges": ["CONSISTENT_WINNER", "RUNNER_KEEPER"],
  "identity_badges": ["ROOT_CEX"]
}

features is an illustrative subset — the response carries the wallet’s full behavioral score set keyed by name. Treat the keys as opaque scores; their selection and scaling are part of Conyr’s model.

Similarity

L2 GET /v1/wallet/{address}/similar

Nearest-neighbor search over the behavioral vector space. Returns ranked look-alikes with a similarity score and human-readable shared_traits / key_differences.

Parameter	Type	Default	Description
`limit`	integer	`20`	Max results (max 100)
`min_trades`	integer	—	Minimum trade-count filter
`performance_level`	string	—	Filter, e.g. `PROFITABLE`

curl -H "Authorization: Bearer $API_KEY" \
  "https://api.conyr.ai/v1/wallet/7xKXtg.../similar?limit=10&performance_level=PROFITABLE"

{
  "query_wallet": "7xKXtg...",
  "results": [
    {
      "wallet": "4mNqR...",
      "similarity": 0.94,
      "trader_type": "DAY_TRADER",
      "performance_level": "PROFITABLE",
      "total_pnl_usd": 45000.0,
      "win_rate": 0.65,
      "shared_traits": ["high win rate", "runner keeper", "quick exit timing"],
      "key_differences": ["larger position sizing", "more DEX diversity"]
    }
  ],
  "count": 10
}

Use case: “find wallets that trade like this one.” Filter by performance_level to surface only profitable look-alikes. Both endpoints cache 60s.

Labels & Taxonomy Provenance & Funding