Strategies · Geopolitics

Risk-off on sanctions escalation

Bursts of sanctions activity mark geopolitical escalation: new programmes, new designations, new export controls. When OFAC publishes far more than usual, step out of stocks into Treasuries and gold until the burst passes.

Pre-registered Sep 26, 2026, before it was tested. Results computed Sep 26, 2026.

Open in the builder

Hypothetical backtest; not investment advice; paper trading only.Past results, real or simulated, do not predict future ones. ARCANE places no live orders from this page.

The rule

Trades on the first trading day of each week

While OFAC has published more than 12 documents in the Federal Register in the last 30 days, hold 40% long Treasuries (TLT), 30% gold (GLD) and 30% short Treasuries (SHY); otherwise hold the S&P 500. Checked weekly.

Where the numbers came from. 12 is twice the 2003–2019 median of 6 documents a month; one month in ten exceeded it. That is the signal’s own history, set before any returns were computed.

Hold what is below
If the OFAC actions, last 30 days is greater than 12
then: Split by set weights (40%, 30%, 30%)
40%: Hold TLT
30%: Hold GLD
30%: Hold SHY
Otherwise
Hold SPY

As text

symphony "Risk-off on sanctions escalation"
description "Defensive while OFAC publishes far more than usual."
rebalance weekly
benchmark SPY
weight equal
  if ARCANE:sanctions/ofac/actions-30d > 12
    weight specified
      40% asset TLT
      30% asset GLD
      30% asset SHY
  else
    asset SPY

How it would have done

Jan 4, 2005 to Sep 25, 2026, against SPY

The whole history

322 trades

A year
+7.6%
Annualized return
In all
+389%
Jan 4, 2005 to Sep 25, 2026
Sharpe
0.50
Return per unit of risk
Worst fall
−55%
Max drawdown
Swings
18%
Volatility, a year
Beta
0.82
To the benchmark

Each decision reads closing prices up to that day and trades at the next day’s close. It trades on the first trading day of each week. Costs: 5 bp of slippage on every trade and 0.3 bp of fees on sales. Uninvested cash earns nothing. Hypothetical: no real orders were placed.

$0$2.50$5$7.50$102006200920122015201820212024Held out

Growth of $1
SPY, held
Held-out period

−100%−50%0%

What it held: SPY 77% of the time; GLD + SHY + TLT 23% of the time.

In sample and the sealed holdout

The holdout was set before the preset was tested, and opened once

Measure	In sample Jan 4, 2005 – Sep 24, 2021	Sealed holdout Sep 27, 2021 – Sep 25, 2026
Annualized return	+8.4%	+5.0%
Sharpe ratio	0.53	0.42
Sortino ratio	0.73	0.61
Max drawdown	55.2%	20.6%
Calmar ratio	0.15	0.24
Volatility	18.7%	13.6%
Alpha against the benchmark	−0.8%	−0.8%
Beta	0.90	0.47
Correlation	0.93	0.60
Winning days	55%	52%
Turnover a year	3.6×	11.6×
Chance the Sharpe is above zero	98%	83%

Is it real?

Checked on the in-sample window only

Overfitting: Likely overfit
No clear edge over its benchmark

The result depends on the exact numbers chosen, or does not beat what luck would produce.

After counting 3 backtests of this idea, the out-of-sample Sharpe does not clear the bar luck would set (deflated Sharpe 0.94; 0.95 is the house bar).
The best version in one half of history usually lands in the bottom half of the other (probability of overfitting 95%).
Against SPY the difference (−1.9% a year) is within what luck produces (13% chance it truly beats it).

Backtests counted	3
Deflated Sharpe ratio (0.95 is the house bar)	0.94
Probability of backtest overfitting (20% is the house ceiling)	95%
Walk-forward Sharpe (choosing, then testing)	0.45
Walk-forward Sharpe with costs doubled	0.42
Extra return a year against SPY	−1.9%

The nudged versions (2)

Change	Sharpe
threshold 12 on sanctions/ofac/actions-30d → 10.8	0.47
threshold 12 on sanctions/ofac/actions-30d → 13.2	0.51

The house backtest gate

It does not pass. A strategy needs every one of these before ARCANE would paper-trade it.

Check	Value	Needs	Result
Evaluation ran without problems	clean	no problems	Passed
Report belongs to this pre-registration	sanctions-risk-off	sanctions-risk-off	Passed
Evaluated only inside the pre-registered data window (holdout untouched)	2005-01-04..2021-09-24	within 2005-01-03..2021-09-24	Passed
Walk-forward folds	13.000	>= 3	Passed
Out-of-sample observations	3276.000	>= 252	Passed
Deflated Sharpe ratio (3 trials counted)	0.937	>= 0.95	Failed
Probability of backtest overfitting	0.949	<= 0.2	Failed
Out-of-sample Sharpe (annual, after costs)	0.447	>= 0.5	Failed
Out-of-sample maximum drawdown	0.519	<= 0.25	Failed
Sharpe with doubled costs	0.420	> 0 (needs a turnover series)	Passed
Sealed holdout: positive and not clearly worse than the backtest	mean 0.000228, consistent	mean > 0 and consistent	Passed

The pre-registration

Written before the backtest; changing any of it makes a new variant

Claim: Holding 40% TLT, 30% GLD and 30% SHY while OFAC has published more than 12 Federal Register documents in 30 days, and SPY otherwise, beats SPY buy-and-hold on Sharpe after 5 bp costs over the sealed holdout.
Why the edge should exist: Clusters of sanctions actions are a slow, public record of escalation that equity prices under-weight while the risk builds.
Where it comes from: slow information
Who is on the other side: Equity holders who stay invested through escalations that have not yet hit earnings.
Why it is not arbitraged away: The record is public but dull, and its link to returns is loose and slow; no large fund trades Federal Register counts.
Universe: OFAC actions, last 30 days, GLD, SHY, SPY, TLT
In-sample window: Jan 3, 2005 to Sep 24, 2021
Sealed holdout: Sep 27, 2021 to Sep 25, 2026
Evaluation: walk-forward, on sharpe, against SPY buy and hold; 13 variants planned
Costs: 5 bp slippage a trade, no commission
Capacity: $1,000,000: TLT, GLD, SHY and SPY are deep markets; the rule trades a few times a year.
What stops it: Paper Sharpe significantly below the backtest after 26 weeks. The Federal Register changes how it counts agency documents.

The data

Prices: Yahoo Finance (unofficial chart API): adjusted daily closes for GLD, SHY, SPY, TLT. Unofficial and not licensed for redistribution. A licensed vendor (Alpaca, Tiingo or Polygon) is a founder decision before the builder is sold.
OFAC actions, last 30 days: Federal Register, from Jan 30, 2000. History: point-in-time. Counted by Federal Register publication date, the day a document became public, from 2000. The register can trail an action announced first on the agency’s own site.