DIARY
Plan A: no win.
Plan B: no win.
A diary of testing ideas on historical data and keeping the results as they came out. Even when we expect to lose, we check how we lose.
RESULTResults so far
- Tests run
- 510+
- Weekly news trades
- 31
- Passed the recheck
- 1
- Proven winners
- 0
The 26 A–Z plans add up to +73.9% before costs and −645.3% after. Most of the “why it loses” is the spread. One idea passed the recheck (+17.7% with swap), but it is not proven either.
CHAPTERSChapters
Plans A–Z
26 famous methods, A to Z. The two that passed the criteria failed the recheck.
Moving averages
Nearly all H1 crosses lost. Daily crosses did not beat random.
MACD
Different settings didn't outrun the costs.
RSI
H1 reversal lost. A daily pullback was the exception.
Bollinger Bands
H1 and daily bars gave opposite results.
Other indicators
ADX, Heikin Ashi, Parabolic SAR: all the same shape.
Original ideas
Reversing losers, switching, pairs, and ranking.
Knowing the trend
A cheating thought experiment. Buy-and-hold beat every technique.
Time
Shorter timeframes lost more, because of the spread.
Volatility
Size can be read, direction can't. Sizing was the useful part.
Weekly news
17 weeks trading only on news. −1.5%, no different from random.
What survived
The one idea that passed the recheck and the unused-data test.
COMMON CONDITIONSCommon conditions
- Data: January 2022 to October 2026, historical data from one broker's demo account (17 symbols), read through the MT5 Analyzer Python environment. No orders were placed.
- Costs: the recorded spread once per round trip. No commission. Swap is excluded unless stated.
- Results are % of capital, split evenly across symbols, sized so one symbol's daily move is about 1% of its capital (no compounding).
- The first half runs to June 2024, the second from July 2024. “Random” keeps symbol, direction and holding time, and changes only the entry time.
UPDATE LOGUpdate log
- October 31, 2026: Added chapters. We also found that some downloaded 2026 data was missing, so we downloaded it again and recalculated every chapter. A–Z figures changed slightly; verdicts did not.
- October 31, 2026: Published plans A to Z.
This diary records tests on historical data. It does not recommend any method or symbol and does not indicate future profit or loss. Please make investment decisions at your own risk.