Market Cap Filter

Root causes of the two “missing” trades

SBET (2025-05-27)

• The universe filter checks for c.volume > 1_000_000, but c.volume is the previous day's regular session volume, not today's premarket volume.
• On 5/26, SBET traded only ~210K shares (from your note), so it fails this filter.
• It never gets into your gap ranking or trade logic. _log_symbol() is accurate when it logs “SBET NOT included: failed price/volume screen”—this means SBET never enters the candidate universe that day, despite a large gap premarket.

KNW (2025-06-06)

• On the prior day, KNW did meet the 1 million volume check, so it passes coarse selection.
• However, your _build_gap_universe() runs at 9:30:10am, before the first 9:30–9:31 bar (today's open) exists in the minute data for most symbols.
• As a result, sec.open at that moment still points to the previous day's open—not the intended 9:30am open price.
• The gap is incorrectly computed as (yesterday_open − yesterday_close) / yesterday_close, often slightly negative.
• KNW is never included in the "top-50 gap list" because the calculated gap does not reflect reality. Even if manually added, the trading logic would only trigger if the price rallied 30% above the true 9:30 open—a rare event.

Why your volume screen isn't working as intended

The expression c.volume > 1_000,000 uses CoarseFundamental.volume, which represents the last full regular session volume—not today's premarket or combined session volume.
This means that premarket block trades or spikes are ignored, and the filter isn’t actually screening for overnight liquidity.

How to properly filter for premarket volume >= 1,000,000 from midnight to 9:30am

1. Move the gap universe build schedule: Run it at 9:31am (not 9:30:10) to ensure minute bar data for 9:30 is available.

   self.schedule.on(self.date_rules.every_day(), self.time_rules.at(9, 31), self._build_gap_universe)
   

2. Loosen the coarse filter: In _coarse_select, keep only the price-check (not volume). This keeps the candidate list feasible but doesn't drop stocks for weak prior-day volume.

   selected = [c for c in coarse if c.price >= 0.5]
   

3. Add the true premarket volume filter in the universe build: For each symbol in your candidate set:

   pre_hist = self.history(symbol, 570, Resolution.MINUTE, extended_market_hours=True)
   pre_volume = pre_hist['volume'].sum()  # sum from 00:00 → 09:30
   if pre_volume < 1_000_000:
       continue
   

4. Fix the gap calculation:

   today_open = pre_hist.iloc[-1]['open']  # This is the 9:30am bar, i.e., today's session open
   prev_close = self.history(symbol, 1, Resolution.DAILY).iloc[0]['close']
   gap_pct = (today_open - prev_close) / prev_close
   

With these changes, your actual premarket volume filter will work as you intend—and SBET (5/27) and KNW (6/6) will both pass the premarket volume test, rank by actual gap, and be eligible for trading based on true opening conditions.

Summary:

• The current code screens on yesterday's volume and yesterday's open, not today's premarket and today's open.
• You must gather today's premarket volume using minute bar data with extended hours, and wait until the market has actually opened (after the 9:30 bar is available) to read the real open and compute the gap.
• This change will align your universe filter with your real-world expectations and screenshots.

Sources: Opening Range Breakout for Stocks in Play, Momentum And State Of Market Filters, Piotroski F-Score Investing

Thanks!  Could you please provide the newly coded algorithm with your 4 suggested changes?