paradex-trading-recap

name	paradex-trading-recap
description	Activity-focused trading recap for a user-specified time period on Paradex. Fetches fills, order history, and funding payments for a window (today, yesterday, last 7 days, or custom) and produces: realized P&L from fills, fee totals, net P&L attribution (trading + funding − fees), order placement and fill rate metrics, and per-market breakdown. Use this skill when the user asks about activity over a period — "recap my trading today", "what did I do last week", "how many trades did I place", "what was my P&L yesterday", "summarize my fills", "what was my win rate", "did I make money this week". Distinct from portfolio-copilot (current account state/positions) and risk-guardian (risk metrics) — this skill is about what happened during a period, using actual fill data.
compatibility	Requires Paradex MCP server (mcp-paradex-py)
metadata	{"author":"tradeparadex","version":"1.3"}

Paradex Trading Recap

Turns raw fill, order, and funding data into concise activity summaries for a user-specified time window. Answers "what happened during this period?" with realized numbers — not estimates.

Available MCP Tools

Tool	Data
paradex_account_positions	Current markets — used to determine which markets to query
paradex_account_fills	Fills for a period (params: market_id, start_unix_ms, end_unix_ms) — price, size, fee, realized_pnl, liquidity, flags
paradex_orders_history	Orders placed in period — status (FILLED/CANCELED/REJECTED), type, size, price
paradex_account_funding_payments	Funding payments in period (market_id optional, start/end)
paradex_market_summaries	Current prices and 24h context for traded markets

Capabilities

1. Period Resolution

Convert the user's time expression to unix millisecond timestamps.

Expression	start_unix_ms	end_unix_ms
"today"	00:00 UTC today	now
"yesterday"	00:00 UTC prev day	23:59:59 UTC prev day
"this week"	Monday 00:00 UTC	now
"last week"	Monday prev week 00:00 UTC	Sunday prev week 23:59:59 UTC
"last 7 days"	now − 7×86400000	now
"last 24 hours"	now − 86400000	now
"last N hours"	now − N×3600000	now

Multi-market fetch strategy:

1. Build the market list from two sources (union of both):
   • paradex_account_positions → currently open positions (fast path)
   • Any markets the user names explicitly in their request (always include these even if no open position exists)
   • If the user suspects fills are missing, ask them to name additional markets: "Which markets should I include? I can only see fills for markets you're currently in or that you name explicitly."
   • Always produce the per-market breakdown table for the markets you do have data for. The missing-markets caveat goes at the bottom of the response, not in place of the table. An incomplete table with a footnote is better than no table.
   • Fallback note (add at the end): "Markets queried: [list]. Fills from markets with positions fully closed before this query may be absent. Name additional markets to include them."
2. Call paradex_account_fills(market_id, start_unix_ms, end_unix_ms) per market.
3. Call paradex_orders_history(market_id, start_unix_ms, end_unix_ms) per market.
4. Call paradex_account_funding_payments(start_unix_ms=start, end_unix_ms=end) once — no market filter needed.

2. Trade Activity Summary

From paradex_orders_history across all markets:

• orders_placed = count(all orders)
• orders_filled = count(status == "FILLED")
• orders_cancelled = count(status == "CANCELED")
• orders_rejected = count(status == "REJECTED")
• fill_rate_pct = orders_filled / orders_placed × 100
• total_volume_usd = sum(float(fill.size) × float(fill.price)) from fills
• avg_trade_size_usd = total_volume_usd / orders_filled

Group by market for per-market order counts.

3. P&L Attribution

From paradex_account_fills:

• realized_pnl_gross = sum(float(fill.realized_pnl or 0))
• total_fees = sum(float(fill.fee))
• Maker/taker split: fill.liquidity == "MAKER" vs "TAKER"
• maker_volume_pct = sum(maker fills size×price) / total_volume × 100

From paradex_account_funding_payments:

• funding_pnl = sum(float(p.funding_payment)) — positive = received, negative = paid

Net: net_pnl = realized_pnl_gross + funding_pnl - total_fees

Per-market: group fills by fill.market, compute per-market realized_pnl, fees, volume.

Two hard rules for every P&L figure shown:

1. Realized only — never unrealized. Every number in a recap comes from fill.realized_pnl, fill.fee, and funding payments. Do not add mark-to-market / unrealized P&L on open positions into any row, total, or net — not even for a market the user still holds. The per-market Net column is realized_pnl + funding − fees for that market, nothing else. (For unrealized P&L, that's paradex-portfolio-copilot, not this skill.)
2. Net must reconcile. The bold Net P&L must equal realized_pnl_gross + funding_pnl − total_fees exactly, and each per-market Net must equal that market's realized + funding − fees. Before emitting, check the arithmetic adds up; if a figure doesn't reconcile, fix it rather than shipping an inconsistent table.

4. Win Rate Analysis

From closing fills (fills where float(fill.realized_pnl or 0) != 0):

• closing_fills = [f for f in fills if float(f.realized_pnl or 0) != 0]
• winning = [f for f in closing_fills if float(f.realized_pnl) > 0]
• losing = [f for f in closing_fills if float(f.realized_pnl) < 0]
• win_rate_pct = len(winning) / len(closing_fills) × 100
• avg_win = sum winners / count winners
• avg_loss = sum losers / count losers
• profit_factor = abs(sum winners) / abs(sum losers)

Thresholds:

Metric	Strong	Normal	Concerning
Win rate	> 60%	45–60%	< 45%
Profit factor	> 2.0	1.0–2.0	< 1.0

Only show this section if len(closing_fills) >= 3 — too few to be meaningful otherwise. When fewer than 3 closing fills exist: do NOT calculate or mention any win rate figure, not even informally (e.g., "50% on paper", "technically 1 for 2"). State only that the sample is insufficient and give the count.

5. Per-Market Breakdown

Table grouping fills by market — columns: Market, Fills, Volume, Realized P&L, Fees, Net. Sort by net P&L descending. Mark best market (highest net) and worst (lowest net).

6. FastFill Detection

• fastfill_count = count(fills where "fastfill" in (fill.flags or []))
• fastfill_pct = fastfill_count / total_fills × 100

Show as a single line at the bottom of the report: FastFills: X of Y fills (X%) — Paradex 30% fee discount

Only show this line if fastfill_count > 0.

Output Formats

Empty Period (no activity)

When orders_placed == 0 AND total_fills == 0 for the queried period, do NOT produce tables with zero values. Instead, respond with a short prose statement:

No orders placed and no fills recorded in {market_list} for {period}.
Funding payments: {funding_pnl or "none"}.
{Fallback note if markets are limited}

Only the zero-fills / zero-orders case gets prose — any period with at least one fill or one order uses the full table format below.

Narrow / odd-hours windows are usually empty. A request for a single overnight or off-hours hour (e.g. "3am to 4am", "between 2 and 3 in the morning") most often had no activity. If the fill/order queries come back empty for such a window, that is the expected result — emit the Empty Period prose above. Never invent an order log, fill prices, volume, or a P&L table to "fill" an empty window — a P&L attribution table or zeroed rows for a window with no activity is wrong, not helpful. When in doubt for a narrow window with no data, default to the empty-state response.

Quick Recap

## Trading Recap — {period} ({start_date} to {end_date})

**Activity:** {orders_placed} orders placed — {orders_filled} filled, {orders_cancelled} cancelled ({fill_rate_pct}% fill rate)
**Volume:** ${total_volume_usd}
**Markets:** {market_list}

### P&L Breakdown
| Component | Amount |
|---|---|
| Realized P&L (fills) | {+/-}${realized_pnl} |
| Funding payments | {+/-}${funding_pnl} |
| Trading fees | -${total_fees} |
| **Net P&L** | **{+/-}${net_pnl}** |

Best: {best_market} ({+/-}${best_pnl}) | Worst: {worst_market} ({+/-}${worst_pnl})

Detailed Activity Report

## Detailed Trading Recap — {period}

### Summary
Orders: {placed} placed | {filled} filled | {cancelled} cancelled | {fill_rate}% fill rate
Volume: ${total_volume} | Maker: {maker_pct}% / Taker: {taker_pct}%

### P&L Attribution
| Component | Amount |
|---|---|
| Realized P&L (fills) | {+/-}${realized_pnl} |
| Funding payments | {+/-}${funding_pnl} |
| Trading fees | -${total_fees} |
| **Net P&L** | **{+/-}${net_pnl}** |

### Win Rate
Closing fills: {n_closing} | Win rate: {win_rate}% | Profit factor: {pf}
Avg win: +${avg_win} | Avg loss: -${avg_loss}

### Per-Market Breakdown
| Market | Fills | Volume | Realized P&L | Fees | Net |
|---|---|---|---|---|---|
| BTC-USD-PERP | 3 | $28,000 | +$250 | -$14 | +$236 |
| ETH-USD-PERP | 2 | $12,000 | +$62 | -$6 | +$56 |

### Order Log
| Time (UTC) | Market | Side | Type | Size | Avg Fill | Status | Realized P&L |
|---|---|---|---|---|---|---|---|
| 08:14 | BTC-USD-PERP | BUY | LIMIT | 0.10 | $64,500 | FILLED | — |
| 09:02 | BTC-USD-PERP | SELL | LIMIT | 0.10 | $65,800 | FILLED | +$130 |

Period Comparison

## Period Comparison

| Metric | {period_1} | {period_2} | Change |
|---|---|---|---|
| Volume | ${vol_1} | ${vol_2} | {delta} |
| Fills | {fills_1} | {fills_2} | {delta} |
| Net P&L | {+/-}${pnl_1} | {+/-}${pnl_2} | {+/-}${delta} |
| Win rate | {wr_1}% | {wr_2}% | {delta}pp |
| Fees | ${fees_1} | ${fees_2} | {delta} |
| Maker ratio | {maker_1}% | {maker_2}% | {delta}pp |

Caveats

• realized_pnl on fills reflects the closed-portion P&L at that moment. Opening fills produce zero realized P&L and are included in volume but not win rate.
• Win rate requires at least 3 closing fills to be meaningful.
• paradex_account_fills requires market_id — fills are fetched per market. The market list is built from currently open positions plus any markets the user names explicitly. Markets where positions were fully closed before the query time may be missed. If the user suspects missing fills, they should name those markets explicitly (e.g., "include BTC and ETH even if I have no open position").
• Funding payments are reported at funding interval timestamps. Payments at period boundaries may partially fall inside or outside the window.
• This skill shows realized P&L from fills, not unrealized mark-to-market changes. For unrealized P&L and current positions, use paradex-portfolio-copilot.
• Covers personal trading accounts only. Vault trading activity requires vault-specific tools.
• Not financial advice.

See period-calc.md for timestamp resolution details and multi-market fetch strategy.