I am still waiting on the top 20

Screening Filters

1. user_watchlist: True

   • Purpose: Limit the search universe strictly to the stocks you originally provided.
   • Rationale:
     • Your request was specifically: “I need the following stocks ranked by price prediction” and then you listed a custom set of tickers.
     • This filter ensures we do not introduce any new stocks outside that list and only work with your chosen names.
     • It keeps the analysis aligned with your existing ideas/holdings, rather than doing a broad market scan.

2. user_id: CUSTOM_LIST_KNTNF_MU_EGO_..._MUFG

   • Purpose: Technically define which “watchlist” to use when user_watchlist = True.
   • Rationale:
     • This ID encodes your exact custom list: KNTNF, MU, EGO, WPM, FNV, CENX, FTAI, STRL, AMD, VRT, …, MUFG.
     • It tells the screener: “When you see user_watchlist=True, use this specific list of tickers for the search.”
     • Together with user_watchlist, it guarantees that the screener only evaluates and ranks the stocks you named.

3. one_month_rise_prob: { min: "50" }

   • Purpose: Require at least a 50% probability (i.e., better than a coin flip) that the stock’s price will be higher one month from now.
   • Rationale:
     • Your goal is to get the top 20 stocks by price prediction. It’s not just about the magnitude of predicted return, but also whether the model thinks the price is more likely to go up than down.
     • This filter removes names where the model sees the odds as worse than 50/50 in the next month, even if some may have high but very uncertain upside.
     • Outcome: you’re left with candidates that have both decent return potential and a reasonable probability of actually rising in the near term.

4. one_month_predict_return: { min: "0.0001" }

   • Purpose: Screen out stocks with a non-positive or effectively zero predicted 1‑month return.
   • Rationale:
     • Your original instruction was to rank by price prediction, which is naturally interpreted as ranking by predicted return.
     • By requiring a minimum positive predicted return (even very small), the screener removes any stocks that the model expects to be flat or negative over the next month.
     • This refines your list to stocks with some expected upside, then lets the ranking (by predicted return) highlight the most attractive among them.

Why These Results Match Your Request

• The combination of user_watchlist: True + your specific user_id ensures the screener only considers the exact tickers you gave, directly matching your request.
• The one_month_predict_return and one_month_rise_prob filters align with “top 20 by price prediction” by:
  • Forcing positive expected 1‑month returns, and
  • Requiring at least 50% odds of a price increase in that period.
• After filtering, the system can rank the remaining names by predicted 1‑month return, giving you the top 20 most promising stocks within your own list based on the model’s price forecasts.