Methodology
How TellTales reads a holder base, trait by trait, with every threshold in a versioned parameter file.
How TellTales reads a holder base
- Snapshot every holder of the mint. Off-curve owners (pools, vaults, bonding curves) are set aside as programs and listed with their share.
- Replay each profiled wallet's successful, balance-changing transactions. Trades are read from balance deltas, so every DEX and router is covered, including stablecoin-quoted trades (converted with hourly SOL/USD). A buy only counts when the wallet itself paid; co-signed bundle buys are flagged.
- Build a FIFO cost-basis ledger per wallet, then score the traits below. Every rate is shrunk toward a prior and carries a 90% interval, so small samples cannot look certain.
- Simulate the holder base: each holder sells according to their own history, weighted by balance and conditioned on today's price versus their cost. Holders without usable history use the population prior, always shown separately.
- Group holders by who first funded them. Young funders that seeded many holders are distributors; exchange-like hubs are reported but never merged into one entity.
Traits
Capitulation
Of the times a position fell ≥30% below its FIFO cost, how often the wallet sold ≥25% of it, versus holding through. Holds need a known price; a missing price is “unknown”, never “dead”. Small leftovers (≤25% of peak) after profit-taking are moonbags, not holds.
FOMO re-entry
After a sell, the next buy of the same token within 24h is a re-entry. It is a whipsaw when bought back ≥10% above the wallet's own exit price. The FOMO tax is what it overpaid against its exit - never more than the buy itself.
Averages up
When adding to a position (≥10% above or below), whether it buys above its running average cost (chasing) or below it (dip buying).
One-shot exits
Share of closed positions exited in a single sell of ≥90% of the holding (clipping), versus scaling out.
Take-profit multiple
Median multiple of cost at which profitable sells (≥1.05x) happen; also counts initials-out moves.
Holding horizon
Median time from first buy to final exit. Under 10 minutes is scalping; over 3 days is swing.
Disposition
Median hold of losing trades divided by winning ones (≥3 of each). Above 1 means it rides losers and cuts winners.
Tilt
After a loss of 50% or more, how often the next position opens within an hour at ≥1.5× its usual size.
Win rate
Closed positions with known cost - bought and sold for SOL or stablecoins - that returned at least their cost.
Profit factor
Gross SOL won divided by gross SOL lost on those positions (capped at 99).
Bot-likeness
Weighted: no ≥4h sleep gap on busy days (≥6 txs), transaction rate, sub-30-second round trips, same-slot flips. Needs ≥30 transactions.
Wallet age
Days since the first transaction seen. A lower bound when history is capped. Fresh wallets are typical of farms and bundles.
Personas
| Persona | Rule (params v006) |
|---|---|
| Bot | Bot-likeness ≥ 0.6. |
| Fresh Wallet | A flag, not a persona: first seen ≤ 3 days ago. |
| Smart Money | Win rate confidently ≥ 50% over ≥10 trades, and profit factor ≥ 1.5. |
| Paper Hands | Capitulation confidently ≥ 50% over ≥5 drawdowns. |
| FOMO Chaser | Whipsaw rate confidently ≥ 35% over ≥5 re-entries. |
| Revenge Trader | Tilt confidently ≥ 35% over ≥4 big losses. |
| Diamond Hands | Capitulation confidently ≤ 45% over ≥5 priced drawdowns. |
| Bag Holder | Disposition ≥ 3× (holds losers 3× longer than winners). |
| Scalper | Median hold under 10 minutes over ≥5 positions. |
| Clipper | ≥75% of ≥6 exits done in one shot. |
| Scaler | ≤30% of ≥6 exits done in one shot. |
| Dip Buyer | Adds below average cost - chase rate confidently ≤ 40%. |
| Chaser | Adds above average cost - chase rate confidently ≥ 60%. |
| Unclassified | Has history, but no rule is met with enough confidence. Shown grey, at the bottom. |
Charts
Panic curve / take-profit wall
Each holder is simulated from their own sell history, conditional on where price is today versus their cost; blue is history, grey is the population prior. A tile turns grey when the prior is most of its number.
Holder map
The 60 largest holders; circle area is share of circulating supply, colour is a shared non-hub funder (up to three clusters, the rest folded).
Cost basis vs price
Supply by snapshot price ÷ the holder's average cost. Red is underwater (likely sellers into strength), blue is in profit.
When holders got in
Supply by first buy, in time after the earliest current holder's first buy - snipes and bundles cluster at the left.
Concentration
Top-N share of circulating supply and the Gini coefficient across snapshot holders.
Wallet charts
Cumulative realized PnL (closed, known-cost positions), outcomes by proceeds ÷ cost, hold times of sold positions, and activity by UTC hour.
Proof links
Every evidence line links the transaction it came from; token logos link to that token's report.
Known limits
- No price paths yet: “sold the bottom” and chart-momentum FOMO are not measured, only entry-relative behaviour.
- Token-for-token swaps and transfers carry no SOL price, so positions touching them have unknown cost and are left out of PnL, win rate and multiples.
- Wallet history is capped per wallet; very old or very busy wallets are read from their most recent activity (marked “capped”).
- Funding links use each wallet's first incoming SOL or stablecoin; exchange-like hubs (busy and ≥30 days old) are reported but never merged.
- Bundle research reads the launch: same-slot and same-transaction bundles, snipers, distributions and the funding trail (ADR-0013). Jito tip fingerprints are not used yet.
- Curves are models, not forecasts; they have not yet been backtested against realised sell-offs.
All thresholds live in versioned parameter files (v006). Every score is stamped with the params and engine version that produced it and can be replayed exactly.