Wallet Tracker — 9 Analysis Scenarios
Built-in analysis patterns for tracker tx query data. Each scenario defines detection rules, algorithm, and output format.
Data source: All scenarios operate on the trade records returned by baw tracker tx query --json. Fetch the full dataset first, then apply analysis client-side.
⚠️ Critical: ts field is in seconds (10-digit Unix timestamp), not milliseconds. All time-delta calculations (10min window, <1h round-trip, >72h wake-up) must use seconds. Multiply by 1000 only if converting to JS Date objects.
Common prerequisite: The A3 API does NOT accept ca or address as filter parameters. The Skill must fetch all records via tx query, then filter by matching ca (contract address) or address (wallet) to enable single-token or single-address focus scenarios.
1. Rhythm — per-address trading rhythm
Purpose: Profile each address's trading pattern — frequency, buy/sell ratio, average ticket, active hours.
Algorithm:
Group records by `address`:
count = total records
buys = records where tradeSideCategory ∈ {11, 19}
sells = records where tradeSideCategory ∈ {21, 29}
mean_usd = mean(txUsdValue)
active_hours = histogram of ts by hour-of-day
intervals = sorted diffs between consecutive ts
median_interval = median(intervals)Output:
📊 Group Trading Rhythm
Address: {address} ({label})
Total Trades: {count} | Buy: {buys} | Sell: {sells}
Mean Amount: ${mean_usd} | Median Interval: {median_interval}min
Active Hours: {top3_active_hours}2. Anomaly — Large Order Detection
Purpose: Detect abnormally large transactions.
Detection rule: txUsdValue > 3 × mean(txUsdValue for same address) AND txUsdValue > 500
Algorithm:
1. Group records by `address`
2. Compute mean(txUsdValue) per address
3. Flag records where txUsdValue > 3 × mean AND txUsdValue > 500Output:
⚠️ Anomaly Large Order
{ts} | {address} ({label}) | {tradeSide} {tokenName} | ${txUsdValue}
(Mean ${mean_usd}, {multiplier}×Mean)3. Co-buy — Consensus Buying
Purpose: Detect coordinated buying of the same token within a short window.
Detection rule: Same ca, ≥3 distinct addresses buying within 10-minute window.
Algorithm:
1. Filter records: tradeSideCategory ∈ {11, 19} (buys)
2. Group by `ca`
3. For each token, sort buy records by `ts`
4. Sliding window: find any 10-min window (600s) with ≥3 distinct addressesOutput:
🔥 Consensus Buy
{tokenName} ({ca}) | {n}consensus | Window: {window_start} - {window_end}
Buyers: {addr1} ${v1}, {addr2} ${v2}, {addr3} ${v3}4. Round-trip — Quick In-and-Out
Purpose: Detect addresses that buy and sell the same token within 1 hour.
Detection rule: Same address, same ca, buy followed by sell, time gap < 1h.
Algorithm:
1. Group records by (address, ca)
2. Sort by `ts`
3. Find pairs where: first is buy (11/19), second is sell (21/29), gap < 3600s (1h)
4. Compute profit = (sellPrice - buyPrice) / buyPriceOutput:
⚡ Round-Trip
{address} ({label}) | {tokenName} ({ca})
Buy: {buy_ts} @ ${buy_price}
Sell: {sell_ts} @ ${sell_price}
Holding: {holding_minutes}min | Profit: {profit_pct}%5. Rotation — Sector Rotation
Purpose: Detect capital flowing from one set of tokens to another across time windows.
Detection rule: Compare token top-N by volume between two consecutive time windows.
Algorithm:
1. Split records into two windows: first half and second half (by ts median)
2. For each window, compute token volume ranking (sum txUsdValue by ca)
3. exiting = top tokens in window1 not in window2 top
4. entering = top tokens in window2 not in window1 top
5. intersection = tokens in both windows' topOutput:
🔄 Sector Rotation
Exiting Tokens: {exiting_tokens} (window1 top but not in window2)
Entering Tokens: {entering_tokens} (window2 top but not in window1)
Persistent Hot: {intersection} (in both windows)6. First-mover — Pioneer Buying
Purpose: Identify the earliest buyer of each token.
Detection rule: For each ca, find the address with the earliest buy ts.
Algorithm:
1. Filter records: tradeSideCategory ∈ {11, 19} (buys)
2. Group by `ca`
3. For each token, find record with minimum `ts`
4. That address is the first-mover for that tokenOutput:
🎯 First BuyAddress
{tokenName} ({ca}) | First Buy: {address} ({label}) @ {first_buy_ts}
Buy Price: ${price} | Current Price: ${current_price}7. Leader-follower — Follower Delay
Purpose: Measure how quickly followers copy the first-mover.
Detection rule: After first-mover buys, compute median delay for subsequent buyers.
Algorithm:
1. For each ca, identify first-mover (scenario 6)
2. Find all subsequent buys for same ca (sorted by ts)
3. Delays = [follower_ts - first_mover_ts for each follower]
4. Median delay = median(delays)Output:
👥 Leader + Followers
{tokenName} ({ca})
Leader: {first_mover_address} @ {leader_ts}
Follower Count: {n_followers}
Median Follower Delay: {median_delay}min
Delay Distribution: <5min: {n1} | 5-30min: {n2} | >30min: {n3}8. Wake-up — Dormant Wallet Activation
Purpose: Detect addresses that were inactive for >72h then suddenly trade again.
Detection rule: Gap between consecutive transactions for same address > 72h.
Algorithm:
1. Group records by `address`
2. Sort by `ts`
3. Compute intervals between consecutive records
4. Flag addresses where any interval > 259200s (72h)
5. The record after the gap is the "wake-up" tradeOutput:
😴➡️⚡ Wake-Up
{address} ({label})
Last Trade: {last_ts} ({days_ago}days ago)
Current Trade: {wake_ts} | {tradeSide} {tokenName} | ${txUsdValue}
Sleep Days: {sleep_days}9. Bot-like — Bot Activity Detection
Purpose: Detect addresses exhibiting automated trading patterns.
Detection rule: Three evidence pieces — low interval variance + high night ratio + low amount dispersion.
Algorithm:
1. Group records by `address`
2. Compute:
a. Interval variance: std(intervals) / mean(intervals) — low = bot-like
b. Night ratio: fraction of trades between 00:00-06:00 UTC — high = bot-like
c. Amount dispersion: std(txUsdValue) / mean(txUsdValue) — low = bot-like
3. Bot score = (1 if interval_cv < 0.3 else 0) +
(1 if night_ratio > 0.4 else 0) +
(1 if amount_cv < 0.5 else 0)
4. If bot_score ≥ 2, flag as bot-likeOutput:
🤖 Bot-Like
{address} ({label}) | Bot score: {score}/3
IntervalCV: {interval_cv} (Low if < 0.3, else Normal)
Night Ratio: {night_ratio} (High if > 0.4, else Normal)
AmountCV: {amount_cv} (Low if < 0.5, else Normal)
Evidence: {evidence_list}Scenario Composition
Multiple scenarios can be combined for comprehensive analysis:
- 4h Comprehensive Summary: Run rhythm + anomaly + co-buy + rotation on 4h data window
- Group Profile: Run rhythm for all addresses → summarize group behavior
- Token Heat: Run co-buy + first-mover + leader-follower for a specific token (filter by ca client-side)
Data Fetching Strategy
Since tx query returns all records without pagination, fetch once and apply all scenarios in memory. For large groups, consider filtering by --trade-side to reduce dataset size before analysis.