I Tested the Golden Cross Against a Coin-Flipping Monkey

Hi YXI friends,

From Golden Ratios to Golden Crosses, investors are obsessed with finding the Midas touch among technical analysis concepts.

I spent years market-making the most liquid instruments in finance, FX and short-term rates, while practising a wide range of analytical tools on the side. From that seat, the problem is obvious. Too many preconceived theories go unchallenged, and they lead investors down the wrong rabbit holes.

If you are familiar with my charts on X, you will know that until recently I was pretty into Elliott Waves and Fibonacci ratios. When they work, it feels like magic. How often they actually work is another question.

So naturally, I backtest, and I keep score live. Everything I run goes through robust backtesting across multiple market regimes before deployment, including my better-known Multimodel Signals, which spans machine learning, neural networks, trend and market regime. Even then, I stay cautious. Past performance absolutely does not guarantee future success, because markets are fundamentally reflexive. And no matter how hard you try, some degree of overfitting always creeps in.

Running a systematic portfolio alongside discretionary technical analysis, with a live scoreboard on every call, has taught me one thing above all. Once you keep honest score, it becomes very hard to keep enjoying big, bold, discretionary calls over a well-built process that is consistent and less ego-driven.

Today, let’s break down the myth of the Golden Cross, starting with its most commonly cited combination: the 50-day versus the 200-day moving average.

GOOGL Moving Average Strategy 50/200-Day

A Golden Cross is when the short-term moving average (here, the 50-day) crosses above the long-term one (the 200-day). The strategy we will test only stays long while the price is above the 200-day average and the 50-day is above it too. If the price falls below the 200-day, or the 50-day crosses under it (the “Death Cross”), the strategy steps out.

Here are the results for GOOGL over the past two years.

The strategy made six trades, so three round trips, returning 81% against 111% for simply buying and holding. But there is real downside protection. The maximum drawdown was 21% versus 30% for buy-and-hold. The Sharpe ratios are not far apart. Overall, though, buy-and-hold was the better trade. Notice the two head-fakes along the way, where price briefly dipped below the 200-day and promptly recovered.

There is, however, a hidden psychological benefit. Most investors struggle with two sins: cutting winners early, and holding onto losers. Both are rooted in our loss-averse wiring.

If following a simple 50/200 rule is what lets someone hold a winner this long, then capturing 72% of GOOGL’s upside still beats the index comfortably. And it comes without the stress of sweating entries and exits, FOMO, and loss aversion. A mediocre rule, followed, can beat good judgment, abandoned.

GOOGL Moving Average Strategy 10/50-Day

Over the same period, a better, and only slightly twitchier, combination is the 10/50. Buying GOOGL when the 10-day average crosses above the 50-day, while the price holds above the 50-day, returned 126%. That is more than 100% of the upside captured. More impressively, the maximum drawdown was just 11%, barely a third of buy-and-hold. Eight position changes in two years. On this window, on this stock, it beat everything.

Of course, a sharp observer like you will want to ask the only question that matters. Was it foreseeable that the 10/50 would be the superior combination? Or did I just show you the winner after the race?

This is where most technical analysis content stops, and where the interesting work begins. I recently put exactly this question through the full machinery, across my entire coverage of nearly two hundred tickers.

The honest test looks like this. Take every sensible moving-average pair. I use seventeen of them, from 10/50 out to 100/250. Once a year, pick the “best” pair for each ticker using only the data available up to that day. Then grade that choice on the year that follows, the year the picker never saw. Repeat, year after year, ticker after ticker. No peeking.

Here are a couple of results I would like to share.

Firstly, in hindsight, there is always a better MA pair, but that’s missing the point.

Scored over the full history, the actual best pair per ticker beats the one you would actually have picked by a meaningful margin, over a tenth of a Sharpe point on average. Run the same selection walk-forward and the entire advantage evaporates.

The 10/50 on GOOGL is this trap in miniature. It looks brilliant now, over this window. Two years ago you could not have known. The data says picking it then would have been luck, not skill.

Secondly, the moving average pair barely matters at all, on a meta-level.

I compared the annual best-pair picker against deliberately dumb alternatives. One fixed pair for every ticker. An average of all seventeen pairs. Even choosing pairs at random.

Aggregated over nearly two hundred stocks, they all land within a rounding error of each other. The picker beat the random picker on fewer than half of the names.

Read that again. A coin-flipping monkey chooses moving averages roughly as well as the optimiser does. The Midas touch you are hunting for in the parameters simply is not there.

So is trend-following useless?

No, and that is the point.

Across the whole test, what actually earned its keep was the rule itself: be long in uptrends, stand aside in persistent downtrends. That is really a drawdown-protection device, as the 50/200’s shallower drawdowns showed even while it lost on return.

It was also the patience: a minimum holding period after we trade in or out that stops the strategy from twitching on noise. In my testing, this strategy attribute is nearly free risk reduction.

Plus the psychology. A rule you actually follow beats a judgment you abandon at the worst moment.

If you use moving averages, the practical takeaways are these.

Spend your energy on whether to use a trend filter, not which one. Distrust any backtest where the parameters were chosen with the benefit of hindsight.

And if a rule helps you hold winners and sleep through drawdowns, that behavioural edge is real.

The REAL Golden Cross is Discipline.

PS. If you enjoyed this piece, please check out my latest offering on daily Systematic Portfolio and Multi-model Signals at YX Insights.

DISCLAIMER: This newsletter is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.