The 100-Year-Old GOLD Strategy (Institutions Won't Trade It)

Summary

Matteo Conti tests a mechanical gold-futures strategy based on a false breakdown below a prior candle’s low, then compares fixed-contract trading with volatility-based position sizing. He describes the setup as a Wyckoff spring, a version coded by Linda Bradford and Larry Connors in the 1990s as Turtle Soup, and a pattern also associated with the ICT trading community.

Strategy rules

• Use gold futures on a 4-hour chart.
• Select a reference candle and track its low.
• Wait for a later candle to break below that low, followed by a bullish candle that closes back above the reference low.
• Place a buy-stop order at the bullish candle’s high.
• Set the stop-loss at the low of the range and the take-profit at a 1:1 reward-to-risk ratio.
• If the buy-stop is not triggered and price moves back toward the range low, cancel the order and look for a new setup.

Backtest and risk management

The test covered January 2020 through “June 26”; the year is not specified in the subtitles. It included about 600 trades.

With fixed sizing of one contract per trade, the strategy produced:

• 57% win rate and a 1.26 profit factor
• About $149,000 net profit
• A 0.44 Sharpe ratio
• Maximum drawdown reported as 11% on a $500,000 account, or $79,000. These figures do not reconcile arithmetically.

The average win and average loss were approximately equal, so profitability depended on winning more trades than losing. Gold buy-and-hold reportedly returned $456,000 over the comparison period, outperforming the fixed-contract strategy.

The largest losing trade was said to be 18 times the average losing trade, and nearly 8% of the account. This highlighted the danger of sizing every trade identically despite changing volatility.

Conti then applied equal-dollar-risk sizing, targeting $10,000 risk per trade by adjusting the number of contracts according to the distance between entry and stop. Wider stops meant fewer contracts; tighter stops meant more. He says this approach:

• Kept the trade count and entry/exit logic broadly unchanged, with about 600 trades, a 57% win rate, and a 1.26 profit factor
• Increased net profit to $644,000, above the stated buy-and-hold result, while being in the market 34% of the time
• Raised the Sharpe ratio from about 0.4 to 1
• Reduced the largest losing trade to $19,000, versus an average loss of about $9,000

The presenter attributes the sizing approach to research on volatility-managed portfolios and volatility targeting. He argues that volatility clustering can support scaling exposure down when volatility rises and up when it falls.

Takeaways and cautions

• Conti’s main concern is that the entry pattern lacks a demonstrated causal or research-backed “why” that would explain a persistent edge. A plausible narrative about stops below obvious lows being triggered before a reversal is not, in his view, proof of an edge.
• He says he would not trade the strategy live, even after the improved backtest, because the underlying signal still lacks that justification.
• He presents volatility-based sizing as a way to improve risk-adjusted performance and limit tail losses—not as evidence that the entry signal itself is robust.
• He gives a hypothetical example of central-bank buying near recent gold lows as the kind of evidence that could provide a stronger rationale, but does not claim to have established that behavior.
• No formal “not financial advice” disclaimer appears in the subtitles.

Assets, instruments, and sources mentioned

• Gold futures, with gold buy-and-hold as a benchmark
• Wyckoff spring, Turtle Soup, and the ICT school’s related pattern terminology
• Research sources: Moreira and Muir, “Volatility Managed Portfolios,” and a paper titled “The Impact of Volatility Targeting”

Presenter and sources

Matteo Conti presents the analysis. He says he previously worked seven years as a market maker for a large investment bank, generating more than €30 million for the bank, and now runs his own fund. The initial strategy is attributed to Ruben Villar Mozar in the subtitles.

Other named sources include Linda Bradford, Larry Connors, Wyckoff, the ICT school, Moreira, and Muir.