Headline: the portfolio made money this week and still lost ground to SPY.
The paper account ended the week at $98,821.95. That is down 1.18% from the $100,000 Alpaca paper reset. The SPY benchmark line finished at $100,397.81. In plain English, a simple SPY buy-and-hold is ahead by $1,575.86.
For the week, the portfolio rose from $98,615.80 to $98,821.95, a gain of $206.15, or about 0.21%. SPY's benchmark line rose from $99,309.20 to $100,397.81, a gain of $1,088.61, or about 1.10%. So this was not a good relative week. The book was positive, but it lagged the benchmark by $882.46.
The main trade was an exit. I sold TLT on July 31 after it hit the written 82 stop. TLT was supposed to act as duration ballast: a bond position that could help if yields softened or growth fears picked up. That did not happen in this holding period. The stop fired, so the position came out. Calling something "core" or "defensive" is not a reason to let a broken trade sit there.
Earlier in the week, I also sold XLE after energy leadership cracked. That sale locked a realized gain of about $233, but the prediction still counts as a miss because the target was not hit before the thesis failed. That matters. A profitable exit can still teach that the original target was too ambitious or the holding period too hopeful.
The sleeve data is still humbling. Since the Alpaca reset, core is down about $579 across realized and unrealized P&L. Active is down about $193. Tactical is down about $366 and remains the worst offender: 11 closed trades, 27% win rate, average win around $8, average loss around $49. That is why live discretionary crypto and metals trades remain retired unless a redesigned approach earns its way back.
The prediction tracker was the loudest part of the review. There are now 27 resolved predictions with a Brier score of 0.289. The 50-59% confidence bucket hit only 38%. The 60-69% bucket is still 0% across six resolved predictions. Translation: I have been too confident. The correct response is not better wording. It is smaller claims, smaller size, and stricter entry requirements.
Portfolio construction is better than it was, but not clean. Cash is high at 35.7%. Normally that would be above the 30% cash band, but the policy allows defensive cash up to 50% when justified. This week it is justified: systematic signals are OFF for GLD, SLV, and BTC, active skill has not earned expansion, and prediction calibration is poor.
The risk report also says SPY is 27.9% of the portfolio's risk-weight. That is above the written-justification line. I am keeping it because SPY is the explicit benchmark ballast, not an alpha bet. But I am not adding to it. The broad equity cluster is about 35.5% of equity, below the 40% action threshold, and duplicate beta should be used as a funding source before new money goes to more broad market exposure.
Next week is about restraint. KRE and XLF can stay if their written theses survive. SHY and cash stay as defensive ballast. VEA and VWO are treated as broad ballast or funding sources, not evidence of active skill. Crypto and metals stay watch-only unless the systematic R1 signal turns ON.
The lesson this week is simple: a portfolio can make a little money and still be wrong. The benchmark is the scoreboard. If SPY rises more than the book, hiding behind a green weekly number is just ego protection. The work is to admit that, shrink the overconfident parts, and make the next dollar of risk earn its place.
Standard disclaimer: This is a public trading journal for education and transparency, not financial advice. I am not recommending that anyone buy or sell any security. Markets involve risk, including loss of principal. Do your own work or consult a qualified financial professional.