Tactical Defensive flips sides: from +0.45 pp to -1.01 pp against the benchmark

The week's event is a crossing. The Tactical Defensive bot closed 16 October at 3.19% since seeding against a benchmark at 2.74%. This week it reads 3.17% against a benchmark at 4.18%. That is 1.01 percentage points below the yardstick. It got there without moving: its week was -0.02%. NAV stands at 103,151.

When a portfolio changes sides without moving, the move is in the other term. Our house benchmark is the equal-weighted basket of the 500 S&P 500 members as of the start date, computed over a common window for every bot, 18 June to 23 October: 4.18%. A week earlier it stood at 2.74%. Separately, the benchmark calculation was fixed this week so it only counts index members on the start date: a reused ticker (POM) was inflating it from 4.12% to 8.36%. The fix lowers the bar: previous notes compared the portfolios against an inflated benchmark, which made them look worse than they were.

Two portfolios are still ahead. Core S&P 500 sits at 4.78% since seeding, +0.6 pp over the benchmark, with a -0.22% week and NAV 104,758. Diversified Basket is at 4.37%, +0.19 pp over the benchmark, with a +0.56% week and NAV 104,353. Those are thin margins. A couple of weeks like Tactical Defensive's would be enough to flip them too.

Momentum + Core remains the worst result in the catalogue by a wide margin: -13.10% since seeding, 17.28 percentage points below the benchmark, NAV 86,886. It had the best week of any bot, +1.19%, recovering from -14.12% a week ago. That does not change the picture. After 89 days in simulation, the gap to the yardstick is the same order of magnitude as its own drawdown.

Two bots sit outside the equity benchmark comparison. Trend BTC+ETH is at 19.68% over 125 days, with a -0.53% week and NAV 119,678. Multi-asset 6 classes has been running 45 days since 21 August and is at 0.41%, up 0.17% on the week. With that little track record, the number says little.

In the event study, the updated forward read: baseline 0.878 over 11,167 cases; earnings 1.142 over 670 cases; guidance 1.053 over 72 cases; executive changes 0.906 over 277 cases. The event's effect is the gap to the baseline, not the raw ratio. Earnings and guidance stay above their baseline forward, as they did in backtest (1.089 and 1.154 against a 0.964 baseline), although guidance, with 72 cases, is not a sample to conclude anything from. Executive changes do not repeat: they printed 1.115 in backtest and sit almost at their baseline forward.

Work shipped this week: a summary of all 55 in-house strategies tested through 20 October, of which 18 came out robust, and the removal of the 7 bots running non-robust strategies, keeping their history and their page. The bots of the 18 robust strategies and the strategy-of-strategies portfolio were published, with a review at 6 months, period by period. The strategy-of-strategies portfolio also went live, weighting the 10 robust ones inversely to their volatility; its 12.0% and -18.3% figures are backtest, not live trading.

None of this is investment advice. Every bot listed runs on simulated money, and anything labelled backtest is a simulation over past data that describes no real trading.