What it shows: a systematic timing model that votes two signals (a trend filter and a 200-day momentum filter) into one long/cash decision — and how the blend compares to its parts and to buy-and-hold.
The price path is simulated and calibrated to realistic statistics. Illustrative — not a backtest of a specific market, and not investment advice. The point is the method and its overfitting risk, not any specific result.
Build your model
Blend two timing signals — a trend filter (fast vs slow average) and a long-term momentum filter (price above its 200-day average) — with weights and a conviction threshold. Be long only when the weighted vote clears the bar.
Every knob is a chance to curve-fit
Your blended model returned +300% with a worst drawdown of -20%, versus buy-and-hold at +298% / -37%. It rode out the drops more gently — but notice how easily you can tweak the weights and threshold until the backtest looks great. That is the trap: the more dials you turn, the better the past looks and the less the result tends to repeat. A simple rule you trust beats a complex one you have overfit.
Blended vs the parts vs buy-and-hold
Blended
+300%
dd -20% · 33 trades
Buy & hold
+298%
dd -37%
Trend only
+239%
dd -21%
Momentum only
+251%
dd -23%
Equity curves
Reading this
Combining signals can genuinely smooth a strategy's ride — but it multiplies the parameters you can tune, and a backtest tuned on history is the easiest thing in finance to fool yourself with. Favour few, robust rules over many fragile ones. The full method (and the overfitting warning) is in Building a systematic market-timing model.
Frequently asked
- What is a systematic market-timing model?
- A rules-based approach that combines one or more signals (trend, momentum, valuation, volatility, etc.) into a mechanical buy/sell or in/out decision, removing discretion and emotion. This sandbox blends a trend filter (fast vs slow average) and a momentum filter (price above its 200-day average) by weight, going long only when the weighted vote clears a conviction threshold.
- Does combining signals make a strategy better?
- Sometimes — blending uncorrelated signals can smooth returns and reduce drawdowns. But it rarely beats simply staying invested on raw return, and it comes at a cost: every extra weight and threshold is another parameter you can tune until the backtest looks great by accident.
- What is overfitting and why does it matter here?
- Overfitting is tuning a strategy so well to past data that it captures noise instead of a real edge — so it looks brilliant in the backtest and fails live. The more knobs a model has (weights, thresholds, windows), the easier it is to overfit. The honest takeaway: prefer a few robust rules you trust over many fragile ones you have optimised.
- Is this a real backtest?
- No. The price path is simulated and calibrated to realistic statistics; the point is to show how combining signals behaves and how readily extra parameters invite curve-fitting — not to validate any specific model. Not investment advice.