Failure to rebuild universe with alpaca iex feed

[Jierr]

Trying to start edge-scanner for the first time: it tries to rebuild the universe. It fails building with following build error. I just used iex for testing purposes. Do you think iex access is too limited to even fetch data for all symbols? Could you recommend eased filter settings, reduced resolutiuons to may make this work? If yes, what files would I need to patch?

I did start it from a docker container.

----------------------------------------------------------------
  Edge Scanner v1.1.6  -- 2026-09-24   data: Alpaca (IEX feed)
----------------------------------------------------------------
[1/6] Universe
       Universe is inf days old --rebuilding (takes ~10s) ...
       ERROR rebuilding universe:
2026-09-24 19:51:26,677 INFO     Alpaca assets (https://paper-api.alpaca.markets): 14379 active US equities
2026-09-24 19:51:26,699 INFO     Symbols from alpaca after eligibility filter: 6802
2026-09-24 19:51:29,583 INFO     After price filter (>= $15.00): 2850 symbols
2026-09-24 19:51:52,682 INFO     Bar metrics computed for 2844 / 2850 symbols
2026-09-24 19:51:52,698 INFO     Filter results: {'price': 3633, 'avg_vol': 2844, 'no_price': 319, 'no_history': 6}
2026-09-24 19:51:52,698 ERROR    Universe too small: 0 symbols (minimum 100). Check network access and loosen filters.

Thanks for the work you do. I am impressed you managed to produce such complex UI's with claude (as per your youtube video).

[Arthur031221]

Your log identifies the immediate blocker: all 2,844 symbols with bar metrics fail the 20-day average share-volume filter ('avg_vol': 2844). Alpaca describes IEX as a single-exchange feed, so that 5,000,000-share default in Edge Scanner v1.1.6 is too restrictive for this run. You do not need to patch the code to change it.

I compared the released screen() function and defaults in v1.1.6 and v1.3.1 with a fixed sample symbol (price $200, 1,000,000 average daily shares, $200 million average daily dollar volume, ATR 2%):

$ python3 - <<'PY'
import ast, subprocess
from types import SimpleNamespace
metrics = {'ABC': {'last_price': 200.0, 'avg_vol_20d': 1_000_000, 'avg_dollar_vol_20d': 200_000_000, 'atr_pct': 2.0}}
for tag, min_avg_vol, min_dollar_vol_m in [('v1.1.6', 5_000_000, 50.0), ('v1.3.1', 0, 150.0)]:
    source = subprocess.check_output(['git', 'show', f'{tag}:scripts/build_universe.py'], text=True)
    fn = next(x for x in ast.parse(source).body if isinstance(x, ast.FunctionDef) and x.name == 'screen')
    ns = {}
    exec(compile(ast.Module(body=[fn], type_ignores=[]), '<screen>', 'exec'), ns)
    rows, diag = ns['screen'](['ABC'], {'ABC': 200.0}, metrics, SimpleNamespace(min_price=15.0, min_avg_vol=min_avg_vol, min_dollar_vol_m=min_dollar_vol_m, min_atr_pct=1.0))
    print(tag, 'kept=', len(rows), 'reason=', diag[0]['result'])
PY
v1.1.6 kept= 0 reason= avg_vol
v1.3.1 kept= 1 reason= kept

The released v1.3.1 builder turns off the share-volume floor by default (--min-avg-vol 0) and instead uses a $150 million daily dollar-volume floor. Upgrade first and rebuild; that removes the precise filter shown in your log. If you need to stay on v1.1.6, build a separate universe with python scripts/build_universe.py --out data/universe_iex.csv --min-avg-vol 0 --diagnostics data/universe_iex_diag.csv, then run python scripts/run_live.py --universe data/universe_iex.csv. Inspect the diagnostics for the next failing filter before changing another threshold. An explicit universe file is used as-is, so rebuild it manually when needed. IEX's partial-market volume can still make volume-based screens sparse; the new default does not guarantee 100 passing symbols in your account.

[Jierr]

Thanks a lot, this gives me some pointers. The filter still fails, but now I got the universe_iex_diag.csv file with which I could test the dashboard and I can look deeper into filter failures. If I am having a successful filter, which produces a sufficient amount of symbols I'll post it here.

[Arthur031221]

Thanks for the update. The diagnostics CSV is written before the builder checks whether 100 symbols passed, so the file can exist even when the build fails. My earlier suggestion addressed the avg_vol blocker in your original log, but the remaining filters can still leave too few symbols.

If you upgraded to v1.3.1, its default --min-avg-vol is 0, while its default --min-dollar-vol-m is 150 rather than 50 in v1.1.6. Please post the new Filter results: line, the version in the startup banner, and the exact builder command. The CSV's result column records the first failed filter for each symbol; those details will show which threshold to examine next. The builder writes universe_iex.csv only after at least 100 symbols pass.

[simonro]

Thanks @Arthur031221 for the thorough answer, that's exactly right for the first failure.

@Jierr the remaining blocker is almost certainly the dollar-volume floor. IEX is a single exchange with only about 2–3% of US volume, so every volume number is roughly 40x lower than on the full tape, and the $150M default filters out nearly everything. Scale it down to match:

python scripts/build_universe.py --min-dollar-vol-m 4 --diagnostics data/universe_diag.csv

If that still leaves fewer than 100, the diagnostics CSV's result column will show which filter is next. Keep in mind that relative volume and volume-based setups will also behave differently on IEX, since all the defaults were tuned on full-market data.