The models

Indices in an uptrend tend to recover from short pullbacks. The swing model waits for the lowest close of the last ten days while the index sits above its 200-day average, then holds for a few days.

A multi-day low is a real dip rather than a one-day wobble, and the exit on a fresh five-day high gives the bounce room without riding it into the next fall.

• In on a 10-day closing low while above the 200-day moving average.
• Out on a 5-day closing high, or after 15 trading days.
• Waits five trading days after a position closes before the next one.
• About eight positions a year on each index, typically held four to ten days.

A confluence model buys a dip only when the backdrop agrees. The RSI pullback is required; then enough independent measures must line up: the long-term trend, trend strength, MACD, the volatility regime, the calendar and a cross-asset momentum rank.

Requiring the dip matters. Without it the backdrop checks alone agree in any calm bull market, and a model would buy every day rather than wait for a pullback.

• RSI 14 under 40 is required (under 30 on the Nasdaq 100).
• The S&P 500 and Nasdaq 100 need two more of the six other checks; the JSE Top 40 needs all five of its checks.
• Out when RSI recovers (over 65 on the JSE, 70 on the S&P 500, 60 on the Nasdaq) or after 25 trading days.
• A few positions a year, typically held three to five weeks.

Extremes in the gold to silver ratio tend to normalise. When the ratio reaches a one-year extreme, silver has historically outperformed gold on the way back.

The edge is in defining the extreme precisely, and in accepting that extremes can persist for a while.

• The ratio at a one-year z-score extreme flags a reversion candidate.
• Out when the ratio normalises rather than at a fixed target.
• Rare by design, about two positions a year.

Bitcoin is read as a trend-follower. The model goes long on a breakout of a Donchian channel, a new 20-day high, and is out when the trend turns.

A funding-rate filter skips breakouts that occur when perpetual-futures funding is stretched, which historically marked crowded, reversal-prone conditions.

• A close above the 20-day high opens a long position.
• A close below the 10-day low closes it.
• Stretched funding rates skip the entry.
• Each position also carries a 90-day time limit.

Prices that have risen over the past months tend to keep rising, across markets and over a century of data. The trend model reads four horizons at once rather than betting on one.

It aims to sit out sell-offs, which roughly halved the drawdown in the backtest, and in return it lags the index in steady rallies. Its value shows in the bad years, not every year.

• Checked every Friday against the 1, 3, 6 and 12-month returns.
• Long while at least three of the four are positive; out when two or fewer are.
• No time limit: a position lasts as long as the trend.
• Two or three positions a year, held for months.

Sitting behind every read is an eleven-indicator recession model: six coincident indicators answering "are we in it now", and five leading indicators answering "is one coming". Each month it counts how many are flashing and reads the result on a single dial.

US downturns drive global risk-off, so this reading frames every other read on the platform. The full rule set is on the recession diffusion index page.

• 0 to 2 active indicators: healthy expansion.
• 3 to 4 active indicators: warning, a risk-off tilt.
• 5 or more: the threshold that has marked every US recession since the 1970s in the backtest.
• Built entirely from free, public Federal Reserve (FRED) data.