Monthly list from a trained model
AI picks: 20 stocks a month
Every month a trained model ranks the S&P 500 stocks we cover by 20 metrics. The list helps you find stocks – the decision is yours.
List for November 2026
Launch list- Reference date
- Nov 20, 2026
- Data as of
- Close of Nov 19, 2026
- Model
- 2026.1
- Created
- Nov 20, 2026 at 10:55
Launch list: the first list was created on the launch day instead of the second trading day of the month.
502S&P 500 stocks we cover
498scored with enough data
20with the highest model score on the list
The 20 stocks on the list
The 20 stocks with the highest model score among 498 rated.
- Rank 1: hidden
- Rank 2: hidden
- Rank 3: hidden
- Rank 4: hidden
- Rank 5: hidden
- Rank 6: hidden
- Rank 7: hidden
- Rank 8: hidden
- Rank 9: hidden
- Rank 10: hidden
- Rank 11: hidden
- Rank 12: hidden
- Rank 13: hidden
- Rank 14: hidden
- Rank 15: hidden
- Rank 16: hidden
- Rank 17: hidden
- Rank 18: hidden
- Rank 19: hidden
- Rank 20: hidden
Free account Which stocks they are – with three metrics each – you can see with a free account.
By sector
Number of stocks per sector on the current list.
How the model works
Metrics, training and open questions
On the second trading day of each month the model scores every S&P 500 stock we cover that has enough data. It ranks them by what, in the past, came before a higher rank of the price change in the following month, and shows the 20 with the highest score. It predicts no percentage, only an order.
20 metrics in six groups
- Quality Net margin, return on equity, debt relative to equity, free cash flow margin and Piotroski score.
- Growth Revenue and earnings per share, each over 3 and 5 years.
- Valuation against its own history P/E and the P/E relative to its own median of the last 5 years.
- Position and technicals Distance to the 50- and 200-day lines, distance to the 52-week high, daily volatility (ATR) and RSI.
- Price change over 1 and 3 months Longer periods are not used.
- 13F How many of 30 tracked large investors hold the stock according to their mandatory filing – and how that changed from the prior quarter.
Gradient boosting, simply explained
The model consists of 200 small decision trees. Each tree asks at most three yes-or-no questions, such as “is the margin in the upper half?”, and adds a small contribution. Each new tree corrects what the previous ones still placed wrongly; the sum of all contributions is the model score.
Each metric enters as a rank among all scored stocks, not as a raw value. For each stock we show the three metrics with the largest upward contribution; other metrics may have lowered the score. The position shows where the value sits among all scored stocks: “in the top quarter” means the value is higher than for three quarters of the stocks.
Trained on the past only
Model 2026.1 learned from 160 months up to November 2025 and creates the lists for 2026. Every January it is retrained with data up to the previous year; the settings stay the same.
The reference date is the second trading day of the month. All data runs to the close of the previous trading day; prices from the reference date itself are not used.
What is open
An edge over the market is not proven. The list is not a forecast: it shows how the model ranks, not how the stocks will move. That is why we keep an unchangeable log from launch – with good and bad months.
Log
For each list, the average change of the 20 stocks, the S&P 500 (SPY) and all scored stocks (equal weight) – from the open on the reference date to the open on the next month’s reference date. Before costs; unchangeable log since launch on Nov 20, 2026.
The first result is available after the first month has ended.
-
November 2026 Nov 20, 2026 to Dec 2, 2026 Launch list
Result after month end (reference date Dec 2, 2026)
- 4 covered stocks not scored
- List checksum
e6f1ee167d67
Required disclosures
- Created by: TickerEagle. The list is generated automatically by the model’s fixed rules, without manual intervention.
- Created on Nov 20, 2026 at 10:55 with model 2026.1; data as of the close on Nov 19, 2026.
- Method: see “How the model works”. We deliberately show no backtest; only the log since launch counts.
- The operator holds no own positions in the stocks on the list.
- No investment advice, no forecast. Changes before costs; past changes say nothing about future ones.