•Mars now holds two stocks from each of the top seven sectors instead of the top three — fourteen positions rather than six. Nothing else about the strategy changes: the same momentum ranking, the same universe filters, the same equal weighting and the same next-open execution.
•The reason is what we found when we examined Mars’s worst drawdown. Between 13 May and 29 July 2026 it fell 44%, while the S&P 500 fell only 1.7%. That was not a market decline — it was a momentum unwind, and Mars was far more exposed to it than its own sector limits suggested.
•At the peak, Mars believed it was diversified: two Semiconductors, two Industrial/Manufacturing, two Mining/Metals. But Lam Research was filed under Industrial and Corning under Mining, and those two moved together with a correlation of 0.78. Together with the two semiconductor names, four of the six positions were effectively the same datacenter and semiconductor trade wearing three different labels. The two-per-sector rule was working exactly as designed and still produced a concentrated bet.
•Widening to seven sectors forces the portfolio into genuinely different parts of the market. Mars’s worst drawdown falls from -44.4% to -28.8% — roughly a third less — and the Calmar ratio, which measures return against drawdown, improves from 1.70 to 1.77 versus the same rules recalculated on three sectors.
•We tested four other approaches first and none of them worked. Taking profit at a fixed gain, stopping out at a fixed loss, capping how correlated two holdings may be, and holding more stocks within the same sectors all left the drawdown essentially unchanged. Only spreading across more sectors moved it, and it did so consistently: every step from three sectors to seven reduced the drawdown further.
•This is a deliberate trade, not a free improvement. Returns fall as well — a more diversified portfolio holds more moderate performers alongside the strongest ones. If you hold Mars, expect steadier but smaller gains than before.
Mars Performance Figures Have Been Restated
•Changing the portfolio structure required recalculating Mars’s history from the beginning rather than extending it. In doing so we found that Mars’s previously published record could no longer be reproduced by the current code: recalculating even the OLD three-sector rules from scratch produces a materially lower result than the figures we had been showing.
•We have replaced the published record rather than continue showing numbers we cannot regenerate. Mars’s stated annualised return falls from 100.2% to 51.0% and its final value falls about 64%. Roughly 60% of that reduction comes from the recalculation itself and is unrelated to the diversification change; the remainder is the cost of the wider portfolio.
•What improves is the risk. Mars’s worst drawdown falls from -44.4% to -28.8%. In the Combined Portfolio, where Mars is one of six sleeves, the effect is clearly positive: worst drawdown improves from -19.9% to -16.1% and the Calmar ratio rises from 2.82 to 3.01.
•We would rather publish a smaller figure we can stand behind than a larger one we cannot. No other strategy is affected.
New Strategy: Sirius
Sirius — Trend Quality
•Launched Sirius, a long-only strategy that asks not just whether a stock has risen, but whether it rose well.
•Two stocks can post the same six-month return by completely different routes. One gapped 40% on a single announcement and drifted sideways for months; the other ground higher week after week, closing at new highs dozens of times. A plain return ranking cannot tell them apart. Sirius scores three separate measures of the same window — how far the stock climbed, how many days went its way, and how often it closed at a new high — and blends them equally, so excelling at one while failing the others does not rank.
•The day-balance measure is taken from each session’s own open to its close rather than from the previous close. That distinction is deliberate: it captures whether buyers controlled the day itself, rather than whether the stock happened to open higher on overnight news.
•Positions are volatility-scaled rather than equal-weighted, so a calmer name gets a larger allocation and each position contributes a more comparable amount of risk. A cap stops any single quiet name from dominating. A stock must trade above its own short-term average to be bought, though a position already held is never force-sold on that average — it leaves on rank alone.
•Exposure is reduced when short-run market volatility rises above its longer-run level, and halved again while the broad market trades below its long-term trend, where no new positions are opened at all. Exits keep working normally in those periods.
•Sirius’s record begins in February 2022 and covers 4.5 years: 27.1% annualised with a worst drawdown of -21.4%. That drawdown is real. The overlay reduces risk in falling markets but does not remove it, and it can also leave the strategy underinvested through the early part of a recovery.
•Full methodology, portfolio, performance, trades, and stock universe pages are now live under the Sirius section, and Sirius joins the Combined Portfolio as a sixth sleeve.
A Note on How Sirius Was Built
•Sirius is a port of a strategy researched in our sister project, and how its tradable universe is defined turned out to matter more than any other single choice.
•The original design screened on fixed numbers: a minimum share price and a minimum daily volume. We did not adopt that, because a fixed threshold means something different in every market — the same cut-off admitted a few hundred names in one year and more than twice as many in another, quietly changing the strategy without anyone deciding to. It showed the better historical record, but that record was substantially bought by the thresholds rather than earned by the rules.
•What Sirius uses instead is those same requirements expressed as relative positions: a stock must sit above a given percentile of price, and above a given percentile of liquidity, measured against every other candidate that day. The screen then means the same thing whether the market has doubled or halved.
•The liquidity test is a minimum, not an average — a stock must have traded heavily on every one of the last thirty sessions, not merely on average across them. That distinction is load-bearing rather than pedantic: an average lets a name qualify on a few enormous days surrounded by thin ones, which is exactly the stock you cannot get out of when the strategy wants to sell.
Taurus & Jupiter Retired, Orion Launched
Taurus and Jupiter — Deprecated
•We have retired the Taurus and Jupiter strategies. Both underperformed materially in 2026: Jupiter returned -4.6% and Taurus +1.2% year-to-date, while every other live strategy returned at least 16.9% over the same period.
•Jupiter scored stocks on bullish bar patterns, new-high frequency, and low realized volatility. Taurus was our only strategy that shorted individual stocks, targeting the weakest bar structure from an easy-to-borrow universe.
•With Taurus gone, no strategy shorts individual stocks. Mercury still holds short TQQQ in two of its three regimes and continues to require a margin account. If you were running Taurus as a hedge, that exposure is no longer offered on the platform — worth planning around deliberately rather than leaving the slot empty.
•Taurus and Jupiter pages, signals, and data have been removed from the site, the mobile API, and the combined portfolio, and their URLs now redirect to the strategies index. Gemini, Mars, Vega, Mercury, Scorpio, and the newly launched Orion remain fully live.
Orion — Residual Momentum
•Launched Orion, a long-only strategy that ranks stocks on residual momentum: the return left after each stock’s own sector move is stripped out.
•Ordinary momentum has a hidden problem. When you buy the strongest stocks you are often just buying whichever sector is running, which works until the sector turns and every position turns with it. Orion measures each stock against its own sector and ranks on the remainder. A stock up 12% while its sector fell 5% scores above one up 30% alongside a 30% sector.
•Each score is divided by how variable that stock’s leftover return has been, so a quiet, persistent trend ranks above a violent, erratic one. Without that adjustment the ranking collapses onto the most volatile names on the board. No single sector may fill more than a set number of portfolio slots, and overall exposure is scaled back when short-run market volatility rises above its longer-run level.
•Each stock is matched to a sector from its own price behaviour rather than from a filed classification, which can be stale or debatable. Since the strategy depends on removing the right sector from each stock, that match has to be right.
•Orion’s record begins in February 2022 and covers 4.5 years: 21.1% annualised with a worst drawdown of -21.7%. That drawdown is real and deeper than several of our other strategies — sector-neutral is not market-neutral, and the portfolio remains fully long. The backtest assumes execution at open prices with no commission or slippage, and Orion trades more often than most of our strategies, so real-world costs will matter more here than elsewhere.
•Full methodology, portfolio, performance, trades, and stock universe pages are now live under the Orion section.
Combined Portfolio — Rebuilt
•The Combined Portfolio is now an equal-weighted blend of five strategies: Gemini, Mars, Mercury, Orion, and Scorpio, rebalanced weekly. Jupiter and Taurus have been removed and Orion added. Vega continues to be excluded because it shares Gemini’s signals.
•Because the sleeve set changed, the combined portfolio’s historical curve has been recalculated from the new set. Individual strategy records are unaffected.
Scorpio Turnover Reduction
Scorpio — Fewer, Longer-Held Rotations
•Scorpio now requires a challenger ETF to beat a position you already hold by a clear margin before it takes that slot. Previously the strategy held the top 3 scores with a hard cutoff, so a holding could be replaced by a fund scoring a hair above it.
•That produced a lot of trading with little economic purpose. Two-thirds of all Scorpio round trips were opened and closed within three sessions, and the average holding lasted under four days. On August 31 the old rule signalled an exit from GDX (Gold Miners) to enter GDXJ (Junior Gold Miners) on a score difference of 0.009 — swapping one gold-miner fund for a nearly identical one, having bought GDX that same morning.
•Under the new rule, Scorpio makes 993 trades across its full history instead of 4,149 — a 76% reduction. Round trips closed within three sessions fall from 66% to 16%, and the average holding period rises from 3.7 days to about 16 days.
•The margin was chosen on an out-of-sample test rather than by fitting the full history: the rule was selected on data through June 2021 and then measured on July 2021 onward, where it produced a higher return (26.3% vs 25.9% annualised) and a shallower worst drawdown (-17.8% vs -23.9%) than the old rule.
•Measured across the entire backtest since 2016, the change is close to neutral on return and modestly worse on drawdown: 19.7% annualised versus 20.2%, with a worst drawdown of -31.6% versus -27.8%. We think that is a fair price for cutting trading by three quarters, because the historical results assume every trade fills at the open with no commission or spread — an assumption that flatters a high-turnover strategy and that none of us gets in a real account.
•Scorpio's scoring, its 19-ETF universe, the 3-position equal-weight structure, and next-open execution are all unchanged. Only the replacement rule is different. Performance, trades, and portfolio pages have been recalculated and now reflect the new rule throughout.
Saturn Retired
Saturn — Deprecated
•After several consecutive months of underperformance and poor overall results in 2026, we have retired the Saturn short-selling strategy.
•Saturn shorted the weakest momentum stocks from an easy-to-borrow universe using a 4-factor composite weakness score. In 2026, the approach failed to generate the risk-adjusted returns we expect from a live strategy, and the short weak-momentum edge deteriorated across multiple evaluation windows.
•Short exposure on the platform is now handled exclusively by Taurus, which targets stocks with the most bearish bar-level price structure and has continued to perform in line with expectations.
•Saturn pages, signals, and data have been removed from the site, the mobile API, and the combined portfolio. Existing subscribers are unaffected — Gemini, Mars, Jupiter, Mercury, Taurus, and Scorpio remain fully live.
New Strategy: Jupiter
Jupiter — Bullish Momentum
•Launched Jupiter, a long-only stock strategy that identifies stocks exhibiting bullish momentum characteristics: strong price bars, new high patterns, and low volatility.
•Each qualifying stock receives a composite score combining bar pattern analysis, new high detection, and volatility assessment to find stocks with the strongest upward momentum.
•Unlike sector rotation strategies, Jupiter evaluates each stock individually — every qualifying stock competes directly on its composite score with no sector grouping.
•Full methodology, portfolio, performance, trades, and stock universe pages are now live under the Jupiter section.
New Strategy: Vega
Vega — No-Rebalance Momentum
•Launched Vega, a new long-only momentum strategy that shares Gemini's stock universe and 7-factor scoring but never rebalances open positions.
•Positions enter at equal weight and are held at whatever size until they fall out of the top ranks. Winners are allowed to run freely, letting position drift work in your favor.
•Many subscribers requested a version of Gemini without daily rebalancing, as manually resizing positions each day can be impractical. Vega addresses that directly.
•Full methodology, portfolio, performance, trades, and stock universe pages are now live under the Vega section.
Mars & Gemini Strategy Updates
Mars — Universe & Parameter Refinements
•Removed GOOGL from the stock universe — GOOG (Class C) is already in the universe, so holding both share classes was redundant.
•Implemented a trade reduction optimization in the rebalancing engine. Positions within a tolerance band of their target weight are no longer trimmed and re-bought, significantly cutting unnecessary turnover.
•Adjusted strategy parameters (Skip 5→8, MinPrice $55→$60, MaxPrice $400→$380) following a full parameter sweep. The updated backtest shows a Calmar ratio of 4.62 with +1,104% total return and 26.2% max drawdown.
•Current portfolio positions have shifted slightly as a result of these changes.
Gemini — Weekend Data Bug Fix
•Fixed a bug where weekend momentum data files were being processed as trading days. 216 phantom Saturday/Sunday snapshots were included in the backtest, causing incorrect overnight change calculations.
•The symptom: on certain Mondays, the "Overnight" change would show the exact negative of the "Intraday" change, with "Daily Change" at $0 — a mathematical impossibility under normal conditions.
•Root cause: a Sunday momentum file created a snapshot with stale Friday prices, then Monday's overnight was calculated against that phantom Sunday entry instead of Friday's actual close.
•The fix filters weekend dates when loading momentum files. No portfolio changes — this was a display/calculation bug only.