FINAXIONResearch
Finaxion Research · ICT / Smart Money: the order block · indices and ETFs · Recomputed 2026-09-10

Does ICT / Smart Money: the order block work on GLD?

ICT / Smart Money: the order block does not beat buy-and-hold on GLD: -10.9 points a year over 22 years.

Total return
-2.7%
Buy and hold
820.3%
CAGR
-0.1%
Sharpe
-5.60
Max drawdown
-4.1%
Trades
26
Out of sample
overfit
Years
22
1×6.1×20052007200920112013201520172019202120232025
StrategyGLD

Findings

  1. ICT / Smart Money: the order block does not beat buy-and-hold on GLD: -10.9 points a year over 22 years.
  2. Against the S&P 500 over the same window: -11.1 points a year.
  3. Total return -2.7% versus +820.3% for buy-and-hold, with 26 trades.
  4. The result does not hold out of sample: it is overfit.
  5. Maximum drawdown -4.1%.
  6. It beats in neither half of the history (split at 2015).
  7. It ranks #128 of 210 markets for this strategy by annual excess over the S&P 500.

Findings derived from the full panel. Each carries its number and its denominator; the computations are reproducible with artifact 543c64263612.

Exactly what was tested

Mechanical approximation: after a 16-bar bullish impulse we mark the last prior bearish candle; a return of price to that zone → entry; exit after 20 bars. ICT is discretionary: this tests the most-cited version, not any specific trader.

How to read it

The verdict compares the strategy with buying and holding GLD over the same window, with round-trip costs per asset class. “Out of sample” is walk-forward: the rule is fit on one stretch and judged on the next. A result that beats in the full sample but not out of it is overfit. Full methodology in the protocol. How we test.

Open this backtest with its trades   Trades (CSV)

Every strategy on GLD → · Full page for ICT / Smart Money: the order block →

How to cite

Finaxion Research (2026). Does ICT / Smart Money: the order block work on GLD?. Daily backtest 1962-01-01–2026-08-31, artifact 543c64263612. https://finaxion.app/en/research/ict-order-block/gld

Not a recommendation. Past performance does not guarantee future results.