In modern portfolio management, the tension between maximising long-term gains and minimising catastrophic losses defines nearly every asset-allocation decision. Traditional approaches — 60/40 stocks/bonds — rely on asset class diversification to lower risk. Yet diversification can falter during crises when stocks and bonds fall simultaneously, as they did in 2022.

Traditional buy-and-hold delivers the best returns in long, secular bull markets with dampened volatility. But strong bull markets do not last forever. Every investor should be aware of hindsight bubbles and lost decades like the 2000s and 1970s. Even worse: so-called “deep risk” exists in equities. From 1929 to 1954, the inflation-adjusted return of the US stock market was zero percent. That is deep risk. No buy-and-hold portfolio or target date fund can protect investors from deep risk and tail events like 2008, 2020, or the Great Depression.

Jack Bogle — the biggest proponent of buy-and-hold investing ever — said “don't peek!” was perhaps the most important part of investing. Most investors come to realise that is simply not possible. Since the internet, investors can check their portfolio performance at any time. With news constantly circulating and information everywhere, investors need a way to mitigate their drawdown anxiety without abandoning equities entirely. The baseline GuardRail models — US GuardRail 1, EU GuardRail, and Global GuardRail — do not aim to beat the market. Their goal is to achieve a very similar long-run return to the underlying index, but with significantly lower risk, lower volatility, and far smaller drawdowns. The US GuardRail 2 and 3 models do aim to beat the market over long time horizons — but they take on noticeably more risk than the underlying S&P 500 to do so. In all cases, the strategy diversifies risk across time rather than trying to time or forecast markets.

GuardRail's signal is built around the S&P 500, chosen for its breadth, liquidity, and centrality to the global equity market. Using global indices as the signal would introduce currency (FX) risk and volatility, greatly reducing reliability. The S&P 500 — the world's largest single-currency index — is the cleanest, most stable benchmark available. It also has an extremely high correlation with global stocks: if European or Asian markets sink, the whole world does. If they rise, they all do.

All signals are evaluated using daily closing prices only, eliminating intraday noise. The signal produces one of two states at the close of every trading day. What those states mean for portfolio allocation — the specific ETFs, the exact percentages, and the precise signal conditions — is proprietary and available to subscribers only.

In long, deep bear markets like the dot-com bubble burst (2000–2002) or the global financial crisis (2007–2009), the S&P 500 experienced 18–30 month downtrending bear markets with declines of more than 50%. GuardRail stepped out near the top and re-entered near the bottom in both cases — because slow recoveries gave the trend signal enough time to confirm before re-entry. This is the reason GuardRail performed exceptionally well in the 2000s, while most equity markets had negative cumulative returns for the entire decade.

In sharp V-shaped recoveries — like March 2020 or December 2018 — GuardRail will underperform buy-and-hold. The signal does not know the recovery will be fast. These short-duration whipsaw events are the cost of the strategy: small, bounded losses that are far preferable to absorbing a sustained 50%+ drawdown.

GuardRail's five models are built around three of the world's most important equity benchmarks, each chosen for breadth, liquidity, and relevance to its regional investor base.

The 200-day SMA remains one of the simplest yet most enduring tools in systematic investing. Its advantage stems not from predicting markets — no indicator can do that — but from filtering structural trends and detecting volatility regimes. Meb Faber's landmark 2007 paper demonstrated that long-term moving-average filters historically improve Sharpe ratios and reduce drawdowns across multiple asset classes.

When the market is below its 200-day SMA, annualised volatility historically more than doubles compared to periods above it — up 105% at the 200D. Nobel Prize recipients Robert Engle and Tim Bollerslev's GARCH and ARCH models confirmed that markets cycle through persistent low-volatility and high-volatility regimes lasting 6–18 months on average. The 200-day SMA is a reliable practical proxy for identifying which regime the market is currently in.

Shorter moving averages like the 20-day or 50-day appear more attractive at first glance — the S&P 500 returns over 16% annualised when above them vs. 13% above the 200-day. But shorter averages generate far more false signals. Those false signals accumulate as small losses that reduce long-run CAGR by 30–50% compared to the 200-day. Any length between 130 and 220 days can work. The 200-day is not chosen because it is definitively optimal — path dependency means no single length is always best across all markets. It is chosen because it is a well-established threshold that balances signal quality with trade frequency.

GuardRail's benefits are reduced in taxable accounts due to realised gains on each trade. As an example: the capital gains tax rate in Germany is 26% across the board, regardless of holding period. From 1975–2026, a German taxable account using GuardRail saw CAGR decrease by approximately 20% from taxes — from 11.2% to about 9.5% after tax. The strategy remains meaningfully positive even in high-tax jurisdictions, but the tax impact is real and should be factored into any decision to use it in a taxable account.

Trading costs are minimal. The ETFs used in GuardRail trade with spreads under 0.01% and carry the lowest possible annual expense ratios. Combined with its low transaction frequency, GuardRail's all-in implementation cost is far below actively managed funds with comparable volatility targets.

A raw moving average crossover without any filter has a win rate of roughly 20% on individual signal transitions. Tolerance bands around the moving average — as well as confirmation delays, time delays, and SMA-SMA crossovers — can all push that win rate up to just under 50%. This sounds like a significant improvement. And for individual trade quality, it is.

But here is the critical finding: none of these tolerance factors have any measurable effect on long-run CAGR or Sharpe ratio. Extensive testing across six major global indices since 1975 shows that the long-run return curve across all band sizes is essentially flat. This is because each increase in band size filters out more false signals, but makes each remaining false signal proportionally larger. The two effects cancel exactly. There is no free lunch from filtering.

The only legitimate reason to use a tolerance band is cost reduction. Fewer signal changes means fewer taxable events and lower transaction costs. GuardRail employs a specific tolerance level that reduces trade count by approximately 60%, bringing average annual trading friction from ~16 basis points down to ~7 basis points, while keeping signal entries and exits as close as possible to the actual trend line.

In rapidly reversing markets, GuardRail can lag on re-entry — especially when rebounds occur within days of sharp selloffs. The signal's emphasis on daily closing data means that intraday crashes are unavoidable until the following session, though historically such events have been exceedingly rare.

A common alternative to daily signals is using monthly moving averages — the 10-month moving average checked once per month. This reduces trade count, but fails in two critical ways. First, the market could trigger a regime switch the day after the monthly check and trend in that direction for the rest of the month — the signal is missed entirely. Second, choosing the first or last day of the month as the check date is arbitrary; checking mid-month would produce completely different decisions. In flash crashes only a few weeks long — like 1987 — a monthly signal leads to disaster: selling at the bottom after the market has already recovered. GuardRail uses daily signals to avoid all of these failure modes.

Most importantly: trend following goes through extended periods of underperformance. In secular bull markets with rapid V-shaped recoveries, GuardRail's whipsaw signals will reduce returns compared to simple buy-and-hold. This is the cost of the insurance. Investors using GuardRail must understand and accept this tradeoff before starting.

Most investors understand asset diversification — mixing stocks and bonds reduces volatility because they don't always move together. GuardRail applies the same mathematical principle across time. Rather than absorbing a catastrophic bear market all at once — the −55% in 2008, the lost decade of the 2000s, the −48% of 1973–74 — the strategy redistributes that risk into a series of smaller, manageable losses distributed across time in the form of whipsaws.

GuardRail is designed to be implemented alongside a buy-and-hold allocation — not as a pure replacement for it. The research analysed the optimal balance extensively. How much of the portfolio to allocate to the trend signal has a non-linear effect on both volatility and market correlation. The two charts below quantify exactly how these tradeoffs play out.

100% buy-and-hold S&P 500 produced ~17.5% volatility historically. 100% GuardRail trend produced ~11.5%. A 67% trend / 33% buy-and-hold portfolio produces ~12% volatility — nearly identical to 100% trend, while preserving meaningful market correlation. This non-linearity is the mathematical basis for GuardRail's two-thirds trend allocation.

The standard GuardRail models manage equity exposure with minimal structural risk. GuardRail 2 and 3 extend the same core framework into a higher-octane structure for investors with long time horizons, strong risk tolerance, and the behavioural capacity to endure amplified swings. The signal logic is identical to the standard models — the only difference is the leverage level applied during risk-on regimes.

Leverage has a bad reputation. In most contexts it deserves it. A mathematical phenomenon called “leverage decay” means that daily-resetting leveraged ETFs underperform in choppy, mean-reverting markets. A market moving +10% one day and −10% the next leaves a 1× fund approximately flat — but a 3× fund is meaningfully down. During the week of August 8–15, 2011, the S&P 500 ended up +0.51% for the week while a 3× leveraged fund ended down −2.02% — purely from daily volatility compounding against itself.

Across 50+ year backtests from 1975–2026, GuardRail demonstrated consistently high risk-adjusted performance. The table below summarises US GuardRail 1 alongside traditional stock/bond portfolios. GuardRail was also tested across 30 random global equity indices: 29/30 saw improved maximum drawdown, 30/30 saw lower volatility, and 25/30 had better Sharpe and Sortino ratios vs. buy-and-hold.

Every subscriber receives a personalised, fully customisable portfolio allocation workbook. Just let me know when you email whether you're a US or EU based investor and I'll send the right version. The spreadsheet lets you build a complete portfolio blending cash, Treasury bonds, global stocks, the S&P 500, and any combination of the GuardRail models — and shows you real historical performance data for your exact allocation.

Portfolio Allocation & Historical Data Section

The personalised allocation section lets you set your own percentages across every asset class and GuardRail model. The historical data table below it shows the full backtest stats for each — so you can build exactly the portfolio that fits your risk tolerance and investment horizon. US and EU versions both included.

Modified Glidepath Chart — Risk Level vs. Years to Target

The glidepath chart shows how your recommended risk level should evolve as you approach your target date (retirement or other financial goal), based on your planned withdrawal rate. Investors with very low withdrawal rates (<2%) can maintain higher risk levels deep into retirement. Those drawing down 5–7%+ should de-risk more aggressively. The chart is automatically personalised to your inputs in the spreadsheet.