My performance tracker currently shows a total return of 25.36% across roughly nine months of active tracking. August was the worst month in that record, at −7.10%. I want to talk about it openly, because the losses are as much a part of the record as the gains.
Before August, the system had recovered from dips and continued to grow. After trading resumed on a new setup, July was profitable, then August went badly. Market conditions were quiet and difficult for several of the signals I was using. But I cannot put the result entirely down to the market. I was also changing strategies and filters too often, and I was still interfering with decisions that should have been made by a consistent process.
For much of the earlier period, I was working with weeks or months of history on individual signals. A strategy could have a flat spell and I would start looking for something better. I now have around a year of history on some of the signals I developed. That longer record has changed how I judge them. A flat month, or even a losing month, is not enough on its own to tell me a strategy has stopped working.
I made another mistake after the relaunch. I took alert sets with hundreds of signals and filtered some of them down to only 20 or 30 historical trades that looked exceptionally good. Those results were attractive, but the samples were too small to give me much confidence about what would happen next. I was fitting my choices too closely to the past.
I am using a much smaller set of strategies now. Three have extensive history. A fourth is based on a longer-running profitable signal, but that specific version has only a few months of data, so I treat it with more caution. I am also applying a higher standard to filters and keeping enough trades in each sample to see the losses as well as the winners.
The biggest shift is in how I think about those losses. I used to pursue very high accuracy, which made each losing trade feel like evidence that something needed changing. The current approach expects losses. Its historical win rate is lower, but the risk and exits are designed with losing trades in mind. Whether that holds up live still has to be demonstrated.
I was too confident when trading resumed. I expected too much too quickly and used more exposure than I should have. I am running the revised setup at smaller allocation while I watch the live signals, execution, slippage and results. Backtests help me decide what deserves a live trial; they cannot tell me what the next few months will deliver.
August was a poor month. I made mistakes that contributed to it, and I expect there are still lessons ahead. What has changed is that I have more evidence, fewer moving parts, and a clearer standard for deciding when a strategy has earned my confidence. I will keep publishing the results, including the difficult months, as that work continues.