Trading Strategy Validation: 72 Tested, 0 Ready to Use

Trading Strategy Validation: 72 Tested, 0 Ready to Use

  • An audit of all 72 strategies found 0 ready to use.
  • A portfolio selected after 1,024 tests lagged index buy-and-hold returns.
  • Rules and pass criteria are now documented before calculations begin.

1. None of the 72 strategies was ready

On July 27, 2026, I audited the entire existing strategy list. None of the 72 was ready to use. Of 35 signal modules, only 12 still had evidence of their performance.

I also tried to reproduce 43 strategies from outside papers and materials. None produced a profit when reproduced.

Trading Strategy Validation: 72 Tested, 0 Ready to Use

2. The portfolio lost to the index after 1,024 tests

The portfolio I spent the most time on had been selected after 1,024 tests. Its return over the evaluation period was +51.53%. KODEX 200 returned +334.44% over the same period.

The last strategy I froze was called F301. It made 190 trades over 250 trading days. It finished down −24.24% under the base cost assumptions and −62.49% under the stress cost assumptions. It failed 9 of the 11 criteria set in advance. Simply holding UPRO over the same period would have returned +26.40%.

What I builtResultBuy-and-hold over the same period
Portfolio selected after 1,024 tests+51.53%KODEX 200 +334.44%
F301 (base costs)−24.24%UPRO +26.40%

3. I changed the order of the research

The problem was choosing a winner after running a large number of tests. When I restarted the lab on November 1, 2026, I changed the process.

Before calculating anything, I document the rules, time period, costs, and pass criteria. This is called preregistration. If I change a rule after seeing the results, I assign it a new number and count it as a new attempt. Failed experiments stay in the record.

I have never placed a real order. Even a rule that passes the criteria is monitored in a paper-trading ledger, where I record its forward performance without investing real money.

Trading Strategy Validation: 72 Tested, 0 Ready to Use

4. What gets tested

The candidates include well-known trading methods and rules introduced in videos. I translate each verbal rule into one a machine can follow, then check whether it clears the criteria on actual price data after costs are deducted.

A failure means the rule fell short of the preset criteria with these data, over this period, and at these costs. It does not prove the rule has no effect. Passing does not guarantee a profit.

On November 4, four days after opening the lab, I tested several rules in one day. The crypto hypothesis scored 0/8, copying U.S. House members' trade disclosures scored 0/5, and a Nasdaq 100 strategy for just after the open scored 0/9. That evening, one rule passed all 5 criteria for the first time. The research continues in Part 2.

댓글