Market Microstructure

On CEXes, microstructure means orderbook depth, bid-ask spread, and queue position. On AMMs, liquidity sits in pool curves — there is no spread or queue. But the trade tape (the chronological list of swaps) contains rich signal:

1. Who is trading? — Whale wallets vs retail, smart money vs bots
2. How are they trading? — Large single swaps vs DCA-style splits
3. When are they trading? — Volume clustering around events or time zones
4. What direction? — Net buy vs sell pressure over sliding windows

These signals feed into entry/exit timing, position sizing, and token quality scoring.

Trade Classification

Buy vs Sell Identification

On Solana DEXes, every swap has an input token and output token:

From API Data Sources

Birdeye Trade History (/defi/txs/token):

• Returns side field: "buy" or "sell"
• Includes from (input token) and to (output token) amounts

DexScreener Pair Trades:

• Returns type field indicating swap direction relative to the pair

Helius Parsed Transactions:

• Parse swap instructions to extract input/output mints and amounts
• Classify based on which mint matches your target token

See references/trade_classification.md for detailed classification logic and size buckets.

Volume Profiles

Time-Based Profiles

Aggregate trade volume into fixed time buckets to identify patterns:

# Hourly volume profile
hourly_volume = {}
for trade in trades:
    hour = trade["timestamp"] // 3600 * 3600
    hourly_volume.setdefault(hour, {"buy_vol": 0, "sell_vol": 0})
    if trade["side"] == "buy":
        hourly_volume[hour]["buy_vol"] += trade["volume_usd"]
    else:
        hourly_volume[hour]["sell_vol"] += trade["volume_usd"]

Key metrics from time profiles:

• Peak hours — when is the token most actively traded?
• Volume trend — is volume increasing, decreasing, or stable?
• Volume anomalies — spikes exceeding 3x the rolling average

Size-Based Profiles

Classify trades into size buckets to separate whale activity from retail:

Buyer/Seller Pressure Metrics

Core Ratios

def compute_pressure(trades: list[dict], period_seconds: int = 3600) -> dict:
    """Compute buy/sell pressure metrics over a time period."""
    buy_vol = sum(t["volume_usd"] for t in trades if t["side"] == "buy")
    sell_vol = sum(t["volume_usd"] for t in trades if t["side"] == "sell")
    total_vol = buy_vol + sell_vol

    buy_trades = sum(1 for t in trades if t["side"] == "buy")
    sell_trades = sum(1 for t in trades if t["side"] == "sell")
    total_trades = buy_trades + sell_trades

    return {
        "buy_sell_ratio": buy_vol / sell_vol if sell_vol > 0 else float("inf"),
        "buy_volume_pct": buy_vol / total_vol if total_vol > 0 else 0.5,
        "net_flow_usd": buy_vol - sell_vol,
        "trade_count_ratio": buy_trades / total_trades if total_trades > 0 else 0.5,
    }

Signal Interpretation

See references/flow_signals.md for the full signal catalog and composite scoring.

Trade Size Distribution

Analyzing the distribution of trade sizes reveals market structure:

import statistics

def analyze_trade_sizes(trades: list[dict]) -> dict:
    """Analyze trade size distribution."""
    sizes = [t["volume_usd"] for t in trades]
    if not sizes:
        return {}

    return {
        "mean": statistics.mean(sizes),
        "median": statistics.median(sizes),
        "stdev": statistics.stdev(sizes) if len(sizes) > 1 else 0,
        "skew_indicator": statistics.mean(sizes) / statistics.median(sizes),
        "max_trade": max(sizes),
        "whale_pct": sum(s for s in sizes if s > 5000) / sum(sizes),
    }

Interpreting skew: A skew_indicator (mean/median) well above 1.0 indicates a fat-tailed distribution — a few large trades dominate. This is normal for tokens with whale interest but can also signal manipulation.

Momentum Signals from Trade Flow

Volume Acceleration

Compare current period volume to the previous period:

acceleration = current_volume / previous_volume if previous_volume > 0 else 0

• acceleration > 2.0 — volume surge, potential breakout or dump
• acceleration 0.8–1.2 — stable activity
• acceleration < 0.5 — dying interest

Buy Pressure Acceleration

Track how the buy ratio changes over time:

current_buy_ratio = current_buy_vol / current_total_vol
previous_buy_ratio = prev_buy_vol / prev_total_vol
buy_momentum = current_buy_ratio - previous_buy_ratio

Positive buy_momentum with increasing volume is a strong accumulation signal.

Token Velocity

Token velocity measures how frequently tokens change hands:

velocity = daily_volume / circulating_supply

High velocity combined with low unique trader count is a wash trading red flag.

Wash Trading Detection

Wash trading inflates volume to make a token appear more active than it truly is. Key detection signals:

1. Low unique trader ratio — unique_wallets / trade_count < 0.3
2. Volume/TVL anomaly — daily_volume / tvl > 10 (volume vastly exceeds liquidity)
3. Uniform trade sizes — low entropy in trade size distribution
4. Self-trading — same wallet on both sides within short windows
5. Funded-together clusters — multiple wallets funded from the same source

See references/wash_trading.md for detailed detection methods and scoring.

Data Sources

Birdeye API

Primary source for trade history on Solana tokens:

• GET /defi/txs/token — recent trades for a token
• GET /defi/ohlcv — candle data with volume
• GET /defi/price/volume — aggregated volume data

Requires API key. See the birdeye-api skill for endpoint details.

DexScreener API

Free, no-auth alternative for pair-level data:

• GET /latest/dex/tokens/{address} — token pairs with volume
• GET /latest/dex/pairs/solana/{pairAddress} — pair details

Helius API

For wallet-level trade analysis and parsed transactions:

• Parse swap transactions to extract trade details
• Attribute trades to specific wallets
• See the helius-api skill for transaction parsing.

Composite Momentum Score

Combine multiple flow signals into a single score (range: -100 to +100):

def compute_momentum_score(
    buy_ratio: float,
    volume_accel: float,
    whale_buy_pct: float,
    unique_trader_trend: float,
) -> float:
    """Compute composite momentum score from flow signals.

    Args:
        buy_ratio: Buy volume / total volume (0 to 1).
        volume_accel: Current vol / previous vol.
        whale_buy_pct: Whale buy volume / total whale volume (0 to 1).
        unique_trader_trend: Change in unique traders vs previous period.

    Returns:
        Score from -100 (strong sell pressure) to +100 (strong buy pressure).
    """
    # Buy ratio component: 0.5 = neutral, maps to [-40, +40]
    buy_component = (buy_ratio - 0.5) * 80

    # Volume acceleration: >1 = growing, maps to [-20, +20]
    vol_component = min(max((volume_accel - 1.0) * 20, -20), 20)

    # Whale direction: 0.5 = neutral, maps to [-25, +25]
    whale_component = (whale_buy_pct - 0.5) * 50

    # Unique trader growth: positive = healthy, maps to [-15, +15]
    trader_component = min(max(unique_trader_trend * 15, -15), 15)

    score = buy_component + vol_component + whale_component + trader_component
    return max(-100, min(100, score))

Integration with Other Skills

Files

References

• references/trade_classification.md — Buy/sell classification logic, size buckets, aggregation
• references/flow_signals.md — Complete signal catalog with formulas and interpretation
• references/wash_trading.md — Detection methods, metrics, and risk scoring

Scripts

• scripts/trade_flow_analysis.py — Fetch trades, classify, compute flow signals and momentum
• scripts/volume_profile.py — Hourly volume profiles, trend detection, anomaly identification