Start here, About, How a Monday works, FAQ, The backtest
The result that counts is the live one. Since January 12, 2026 the portfolio has run with my own money on Dub, in public, where I cannot edit it. As of October 9, 2026 it is up 109%, against about 13% for the broad market. On the way it had a 30% decline, from June 22 to July 17. The record is here.
The test on this page is how I decide what the rules should be before my money depends on them. It applies the stock picking to past prices and holds the list every week. It is a simulation, and never a promise.
If you want to watch it live before deciding anything, subscribe free and the weekend review arrives every Saturday. The full Monday list is $39 a month or $350 a year.
Why I test it this way
I start from the view that my own system is wrong.
I always go from a point of view that the system is wrong, and that any positive results are likely due to some look-ahead bias or other innocent error. Then I try to prove it wrong. If the results hold up against scrutiny, I get more comfortable with the portfolio.
Look-ahead bias is the mistake that ruins most tests. The test lets the system see something it could not have known at the time, such as earnings released after the close, and the result looks better than it could ever be live. I check for it every week, and I keep finding ways to make the test better.
How the test works
Dates. January 6, 2020 to October 2, 2026, starting with $1,000. The window covers the crash of 2020, the decline of 2022 and the growth stretches in between.
Rules. Today's stock picking, run over the whole period, with the list held every week. A change is adopted only when it improves results across the whole period.
Trades. Every decision uses Friday's closing prices, and every trade fills at Monday's opening price. A stop fills at its price, or at the open when a stock opens below it.
Costs. Every stock trade costs 0.30% of its value and every fund trade 0.10%. Live trades at the open have filled at about the opening price, so this is cautious.
Companies. Companies that were later bought or delisted are included.
Left out. Tax, borrowing and the dividends on the stocks held, which would add a little. The S&P 500 is counted with its dividends.
The results
The stock picking, against the S&P 500 with dividends, from January 2020 to October 2, 2026.
In the test The stock picking S&P 500 with dividends
Average growth a year about 73% about 15%
Worst decline 38% 34%
Calendar years that lost money 2 of 7 1 of 7Three things to know about it.
Three years carry most of it. 2024, 2025 and 2026 supply most of the growth.
The live record is close to it. Over the same months as Dub, January 12 to October 2, 2026, the test shows +122.1% and Dub shows +106.9%. Dub is what happened.
It holds stocks every week. The live portfolio also steps out of stocks into one fund when conditions turn against them. Those rules went live in September 2026, so they are not in this test. I will add them here once they have held up live for longer.
What the test says to expect
It can fall hard. The worst decline in the test was 38%, from February to April 2025. Live, the portfolio fell 30% in June and July 2026.
Some years lose. It lost 3.9% in 2021, a year the S&P 500 gained 28.7%, and 10.4% in 2022.
Waits can be long. The longest stretch without a new high was 419 days.
Where it is weakest
When last year's fastest risers turn and fall. This is the worst case. In 2021 the stock picking lost 28.6% from January to August while the S&P 500 gained 15.4%. June and July 2026 was the same thing live.
Long, quiet markets. It is built for growth, and a flat market gives it little to work with.
One company's bad news. Up to 30% sits in one stock, so one bad earnings report is a bad week. A stop does not help when a stock opens far below it.
Choppy markets. Stops get hit, and then the stocks recover.
How a change gets in
The rules are refined over time, and more changes are likely. Since September 2026 there is a written set of checks for a proposed change, and I decide. A change has to:
improve the return for the risk taken, on the same dates and costs;
show better picking, and not just bigger bets;
still win with the most profitable stocks and the best days removed;
hold up on data that was not used to design it;
hold up at higher trading costs;
use no information from the future;
be tradable live exactly as tested.
Since January 2026 I also save the data the system saw each Monday, before it decides, and never change it afterwards. Going forward, that is a record any test can be checked against.
No AI makes the weekly decisions. They come from fixed rules I can test. I use AI to research and build improvements to those rules, and I decide what goes in. I watch over the process and can step in.
Does it work with more money?
It buys large, heavily traded US companies. A capacity study on the spring 2026 version of the rules found the full return held to about $120 million in the portfolio, and fell after that. It has not been repeated on today's rules.
What the test cannot tell you
What happens next. A test shows how the stock picking would have handled the years since 2020. It cannot show the years ahead. The live record on Dub is the result that counts, and it adds a week every Monday.
The full list goes to paying readers every Monday before the market opens. It is $39 a month or $350 a year, and you can cancel in one click. The weekend review is free for everyone.
Lionshare is a publication, not financial advice. What you do with your own money is your decision. Past performance does not guarantee future returns.