🧪 How we scoreMETHODOLOGY v0.1
Rule #1: every joke has a number behind it. Labels describe the verifiable behaviour or outcome of one call — never a person's character or motives.
⚠️ Preview limits: KOLs, posts and wallet trades are all fictional demos generated from a fixed seed (reproducible). Demo calls only use fictional tokens ($SAMPLE, $SAMPLEDOG…) on a made-up SAMPLECHAIN, with simulated daily prices, so no demo character is ever linked to a real coin. The only real market data on the site is the clearly labelled live ticker (DexScreener / GeckoTerminal). Production will use real calls and the volume-weighted price in the first minute after the call.
1. Definitions
- Call time t₀: when the post / message went out.
- Call price P₀: first price after t₀ (preview: that day's close).
- Follower 7-day result r₇ = P(t₀ + 7d close) ÷ P₀ − 1. We use the close, not the high — no "sold the top in my dreams" bias.
2. Wallet verification levels
| Level | Source | Scored? |
|---|---|---|
| L1 Claimed | KOL signs with the wallet + verifies their X / TG identity | ✅ |
| L2 Self-published | An address the KOL posted themselves in their X bio / posts / TG channel (evidence link saved) | ✅ |
| L3 Third-party tag | Labels from other platforms | ❌ lead only |
| L4 On-chain inference | Funding-flow heuristics | ❌ not used in phase 1 |
3. Four sub-scores per call (0–100)
| Sub-score | Measures | Rule |
|---|---|---|
| T Fair timing | Did they get in cheap in the 72h before calling? | No pre-buy or entry ≥ 0.9×P₀ → 100; 10–100% in profit at call → 60; 100–500% → 30; >500% → 0. Disclosed holding: +20 (max 100) |
| H Holding | % sold in the 24h after the call | 0–20% → 100; 20–50% → 70; 50–80% → 40; >80% → 10 |
| F Follower PnL | Bought at the call price — how is it 7 days later? | Percentile of r₇ among all calls in the same period × 100 |
| S Skin in the game | Did verified wallets hold it at call time? | Yes → 100; No → 40; no verified wallet → not scored |
4. Totals
Call score = 0.30·T + 0.30·H + 0.25·F + 0.15·S (missing parts re-weighted)
Say-Do Score = (n × mean call score + m × μ) ÷ (n + m), m = 10, μ = mean of all calls in the window (30D μ = 77).
No verified wallet → T, H and S can't be measured, so no Say-Do Score; we only show "all talk, no wallet" plus hit rate.
Hit rate is shown separately and never mixed into the Say-Do Score: 7-day median return, and the share of calls whose 7-day close ≥ call price. Translation: you can be very honest and pick terribly, or vice versa.
5. Label dictionary (every label has an objective trigger)
| Label | Trigger | Number shown on card |
|---|---|---|
| 💎 Diamond Hands | Verified wallets still hold ≥ 90% seven days after the call | % still held |
| 🚌 Early Bus | Verified wallet bought within 72h BEFORE the call (we describe the trade, we don't guess why) | hours early, entry vs call price |
| 🏃 Speedrun Exit | Sold ≥ 50% within 24h after the call | % sold, hours after |
| 🗣️ All Talk, No Wallet | The called token never shows up in any verified wallet | number of verified wallets |
| 📸 Hindsight Hero | A brag post that first mentions the token after it already did ≥ 100% (not scored) | gain at first mention |
| 🕯️ Ghost Lamp Call of the Week | Worst 7-day result of all calls this week (tags the call, not the person) | 7-day return |
| 💡 Bright Lamp of the Week | Best median 7-day return this week AND a 30-day Say-Do Score ≥ 70 | median return, sample size |
| 🫡 Disclosed | Labelled the ad (#ad) or said they already hold it | what was disclosed |
5b. One stamp per call
| Stamp | Condition (checked in order) |
|---|---|
| BRIGHT LAMP CALLED IT · PAID OFF | Post and wallet matched, and buyers at the call price were up after 7 days |
| GHOST LAMP CALLED IT · WENT SOUTH | Post and wallet matched, but buyers at the call price were down after 7 days |
| MOUTH ≠ WALLET SAID ONE THING · WALLET DID ANOTHER | The post said buy or hold, but verified wallets sold ≥ 50% within 24h — or never bought at all |
| PENDING WAITING ON 7 DAYS | Not 7 days yet, no result |
| ALL TALK NO VERIFIED WALLET | This KOL has no verified wallet, so we can't compare |
"Mouth ≠ Wallet" is checked first: verified wallets sold ≥ 50% within 24h, or never bought. Otherwise the 7-day result decides: ≥ 0 is a Bright Lamp, < 0 is a Ghost Lamp. A stamp is about that one call, not the person.
6. Wording rules (all languages)
- Labels describe one call, not a person. ✅ "Sold 62% three hours after this call"
- Every number links to evidence (tx hash, original post).
- "Ghost lamp" only ever describes a call's result — a bad-luck joke, not an accusation.
- We never use words that accuse anyone of bad motives or crimes, in any language. An automated check blocks them on every page.
- Any methodology change bumps the version and gets a public changelog; old cards keep the version they were made with.
7. Things we don't do
- No bots auto-replying, @-mentioning or DMing on X
- No scraping X, no buying X data (preview)
- No share rewards, points or giveaways
- No copy-trading, no "KOL just bought" alerts; production delays wallet moves ≥ 30 minutes
- Phase 1: no partnerships, no referrals, no token
8. Data sources
| Data | Source | Preview status |
|---|---|---|
| Live ticker (top of page only) | DexScreener public API + GeckoTerminal daily OHLCV (free, no key) | ✅ real · labelled · 2026-10-17, 09:12:38 Vancouver time (PT) |
| Demo tokens & their prices | $SAMPLE… on SAMPLECHAIN, simulated (fixed seed) | ⚠️ fictional · not a real coin |
| KOL list, posts | Production: KOLs' own public posts / TG channels | ⚠️ fictional demo |
| Wallet trades | Production: Solana history of L1/L2 wallets | ⚠️ fictional demo |
9. Changelog
- v0.1 (2026-10-17): first preview, implements spec §8; P₀ uses the daily close; entry bar 7D ≥ 2, 30D/90D ≥ 5. Site available in English, 简体中文 and 한국어.
For entertainment and research only. Not financial advice.