What is a seasonal edge, and how to read a TradeWave Pattern Card

Learn what a seasonal edge is and how to read every field of a TradeWave Pattern Card, from edge_score to receipts and the broker-agnostic order ticket.

Lesson 01 · 8 min read

What is a seasonal edge?

Markets keep calendars. The same window of the year shows up again and again - a stretch where a symbol has, across many past years, moved in the same direction more often than not. That recurring, calendar-driven tendency is a seasonal edge. Think "this index has historically firmed up in the back half of October" or "this stock tends to drift lower into mid-summer." The patterns repeat because the real-world cycles behind them repeat - earnings calendars, fiscal-year flows, tax dates, weather and demand cycles, index rebalances, and plain old investor habit.

A seasonal edge is not a prediction and not a price target. It is a statistical lean: a window with a favorable historical hit rate and a measurable average move. TradeWave surfaces those windows honestly, shows you the receipts, and lets you decide.

A few framing rules hold everywhere in the API:

The unit of all of this is the Pattern Card - the object returned by /scan, /analyze, and /daily-pick. Read one card well and you can read the whole API.

Pull a card in one call

You do not need an account to see a real Pattern Card. The public demo token tw_demo_explore works against /daily-pick so you can inspect the live shape before you sign up:

curl -s https://api.tradewave.ai/v1/daily-pick \
  -H "Authorization: Bearer tw_demo_explore"

That returns today's forward-tested pick as a full Pattern Card - the same object the rest of this article dissects. Keep the response open in a second window and follow along.

An illustrative Pattern Card

Here is a realistic (illustrative) card for a long US equity setup. Values are made up to teach the shape - do not trade off them.

{
  "rank": 1,
  "symbol": "AAPL",
  "market": { "id": "0", "name": "DOW 30 STOCKS" },
  "direction": "long",
  "bias": "bullish",
  "setup": {
    "entry_date": "2026-10-12",
    "entry_window": "2026-10-12 to 2026-10-16",
    "hold_days": 21,
    "exit_date": "2026-11-02"
  },
  "edge_score": 78,
  "edge_basis": "0.40*win_rate + 0.30*sharpe + 0.20*ml + 0.10*sample",
  "stats": {
    "historical_win_rate": 0.80,
    "sharpe_ratio": 1.6,
    "avg_return_pct": 3.4,
    "median_return_pct": 3.1,
    "years": "15"
  },
  "ml": {
    "ml_score": 72,
    "ml_win_prob": 0.69,
    "pred_return_pct": 2.8,
    "pred_mfe_pct": 4.5
  },
  "receipts": {
    "years_tested": 15,
    "wins": 12,
    "losses": 3,
    "historical_win_rate": 0.80,
    "avg_return_pct": 3.4,
    "median_return_pct": 3.1,
    "best_year": { "year": 2019, "return_pct": 9.2 },
    "worst_year": { "year": 2018, "return_pct": -4.1 },
    "per_year": [
      { "year": 2024, "return_pct": 4.0, "result": "win" },
      { "year": 2023, "return_pct": 2.2, "result": "win" },
      { "year": 2022, "return_pct": -2.7, "result": "loss" }
    ],
    "curve_summary": {
      "shape": "strengthens into the exit; seasonal index rising +6 pts over the hold",
      "trend": "rising",
      "change_pts": 6.0,
      "peak_day": 19,
      "trough_day": 0
    },
    "source": "tradewave-seasonal-engine",
    "as_of": "2026-06-02"
  },
  "next_step": {
    "order_ticket": {
      "side": "BUY",
      "symbol": "AAPL",
      "type": "MARKET",
      "time_in_force": "DAY",
      "suggested_entry_date": "2026-10-12",
      "suggested_exit_date": "2026-11-02",
      "note": "Seasonal long. Place at your broker; sizing is yours."
    },
    "copy_text": "Long AAPL, enter ~Oct 12, exit ~Nov 2 (21-day seasonal hold).",
    "set_reminder": {
      "type": "entry",
      "fire_on": "2026-10-12",
      "message": "AAPL seasonal long entry window opens"
    },
    "framing": "We show the edge and timing; you place the trade."
  },
  "headline": "AAPL has a strong late-October seasonal lean (long).",
  "verdict": "Strong, consistent seasonal long. Seasonal shape: strengthens into the exit; seasonal index rising +6 pts over the hold.",
  "disclaimer": "Educational seasonal pattern + ML research, not personalized investment advice and not a recommendation to buy or sell. Past performance is not indicative of future results.",
  "tier_notes": "ML score shown."
}

disclaimer is not illustrative wording - it is byte-identical on every single card TradeWave returns, verbatim, regardless of tier or setup. tier_notes is also always one of a small fixed set of strings describing the ML state of this exact card (here, "ML score shown." because an ml block is present); it is never null.

Reading every field

Identity and direction

The setup window

setup is the when. entry_date is the central entry day; entry_window is the few-day band around it (seasonality is fuzzy, not a single magic date). hold_days is the holding period and exit_date is when the window historically closes.

edge_score: the one-number summary

edge_score is a 0-100 blend of the four things that make a seasonal lean trustworthy: hit rate, risk-adjusted return, the ML read, and sample depth.

edge_score = round(100 * clamp01(
    0.40 * win_rate +
    0.30 * min(sharpe / 3, 1) +
    0.20 * ml_or_winrate +
    0.10 * min(years / 15, 1)
))

edge_basis spells out which components actually drove the number for this card. Higher is stronger, but always read the receipts before you trust the score.

stats: the historical track

stats summarizes the backtest of this exact window across years of history:

Field Meaning
historical_win_rate Share of past years that were profitable (0..1).
sharpe_ratio Average move per unit of variability.
avg_return_pct Mean percent move over the hold.
median_return_pct Middle percent move (robust to outliers).
years How many years are in the sample.

ml: the model's read (and a key distinction)

ml is either null or an object. When present it adds the machine-learning view: ml_score (0-100 conviction), ml_win_prob (the model's predicted probability of a winning trade, 0..1), pred_return_pct (predicted percent return), and pred_mfe_pct (predicted maximum favorable excursion - best percent move reached during the hold).

Do not conflate two things called a "win rate." historical_win_rate is a measured fact - the share of past years that were profitable. ml_win_prob is a forward prediction - the model's estimated probability for this trade. One looks backward and is countable; the other looks forward and is a model output.

ML is available on every tier but metered per day: the free tier gets a small daily allowance, Pro and Business are unlimited, and the daily pick's ML is always free. It covers US stocks, indices, and ETFs on shorter seasonal holds (up to about 90 days). For longer holds, ml is null and tier_notes explains why: "ML score not available for this setup - the ML model covers shorter seasonal holds (up to about 90 days)." Run out of allowance and you still get HTTP 200 - a card with ml: null, an ml_remaining_today count, and a gentle upgrade nudge, never an error. See the API pricing page and your console for current limits.

receipts: the trust layer

receipts is what makes TradeWave different. It is the auditable proof behind the card: years_tested, wins, losses, the best_year and worst_year, and - most importantly - per_year, a year-by-year breakdown with each year's return_pct and a result of "win" or "loss".

Scan per_year before you trust any score. Twelve wins in fifteen years with one ugly outlier tells a very different story than a knife-edge 8-of-15. curve_summary is an object describing the Trend Chart shape over the hold window - shape (a one-line description), trend ("rising", "falling", or "flat"), change_pts, peak_day, and trough_day - or null when no curve section is available for this setup. as_of plus source time-stamp the evidence. For the daily pick specifically, this evidence is a public, forward-tested track record (see /daily-pick/track-record) - receipts you can check, not a marketing claim.

neutral bias: honesty as a feature

When the best available setup is weak, TradeWave refuses to manufacture a trade. If edge_score < 40, or historical_win_rate < 0.55, or years_tested < 5, the card comes back with bias: "neutral" and no order_ticket. A no is a real answer. Surfacing it is a feature, not a gap.

next_step: the broker-agnostic order ticket

next_step turns a seasonal pattern into action without touching your money. The order_ticket is deliberately price-free:

There is no limit price or dollar level - timing and direction only; sizing and execution are yours. copy_text is a ready-to-paste summary, set_reminder schedules an entry nudge, and framing restates the deal: we show the edge and timing, you place the trade at any broker.

Putting it together

To read a Pattern Card well:

  1. Check bias. If it is neutral, stop - there is no edge worth trading.
  2. Read edge_score for the headline, then edge_basis for what drove it.
  3. Open receipts.per_year - this is the trust step.
  4. Compare historical_win_rate (measured) against ml_win_prob (predicted); treat them as two different measures.
  5. Use setup and next_step.order_ticket for timing, then size and place the trade yourself.

Everything here is illustrative and educational, not personalized advice. You already pulled a live card above with tw_demo_explore; when you are ready to hit /scan and /analyze with your own symbols and rate limits, grab a key at your account console.