gold multi account trading

Multi-Account Gold Trading: Does Splitting Capital Diversify Anything?

Splitting gold trading capital across several accounts is standard practice among retail traders. We test whether it actually diversifies, using 66 years of real gold prices and six strategy sleeves.

By Auriatech Research Desk · · Last updated 2026-11-12 · 3 min read
risk managementportfolioposition sizinggrid trading Contains our own backtest data

Multi-account trading is one of the most common habits in retail gold. One account runs a grid, one runs a trend follower, maybe one holds a small speculative position. The stated reason is diversification: separate accounts, separate drawdowns, separate psychology.

This guide tests that claim against 66 years of real gold prices. The short version: splitting capital across accounts does not diversify anything if every account trades the same asset. It changes the layout of your dashboard, not the shape of your risk.

All numbers below come from 802 monthly gold averages (1960-01 to 2026-10) published by the World Bank’s Pink Sheet, and from a published simulation script. Every figure can be recomputed from the data files linked at the end.

Correction, 12 November 2026. The first version of this article reported the grid sleeves at 7.8–10.0% a year and drew the multi-account conclusion from those numbers. That was wrong. The grid simulator credited a closed level's proceeds to cash and left the same quantity in the position, so sold lots were counted twice and every grid sleeve was inflated by roughly a factor of ten. The simulator has been fixed, every number on this page has been recomputed from the corrected data files, and the conclusion has changed: the grids do not lag buy-and-hold modestly, they lag it by a factor of six or more. The bug and how it was caught are documented at the end of this article and in our grid versus trend-following evidence guide.

What we simulated

Six sleeves, all long spot gold, all on monthly average prices:

Sleeve Rule
BH Buy and hold one unit of gold throughout
TSMOM Long when the trailing 12-month return is positive, otherwise cash
G1 Grid, 1.00% spacing, 20 levels, 10% of equity per level
G2 Grid, 2.00% spacing, 10 levels, 10% of equity per level
G3 Grid, 3.00% spacing, 10 levels, 10% of equity per level
G4 Grid, 0.50% spacing, 40 levels, 10% of equity per level

The grid simulator uses a static anchor per cycle (never a trailing one), each level takes profit at its own entry price times one plus the spacing, and a level is never filled twice. Each grid fill is charged a round-turn cost of $0.30 per troy ounce. Buy-and-hold and the trend sleeve are charged nothing: spot gold has no listed commission, and bid/ask and financing are not modelled.

Read this before the numbers. These are monthly average prices, not prices you could trade. An entry is assumed at the month average, so every result is illustrative rather than executable. There is no margin, no financing cost, no tax and no slippage beyond the stated grid round turn. Every sleeve is long the same asset: no shorts, no other instruments. The grid cost is a flat dollar amount per ounce and does not model the wider spread you would actually meet in a stress month.

The six sleeves on their own

Sleeve CAGR Volatility Max drawdown Sharpe Total multiple Longest underwater
BH 7.399% 15.833% −62.074% 0.215 117.29× 315 months
TSMOM 8.186% 13.861% −31.774% 0.302 191.00× 191 months
G1 1.450% 8.902% −49.681% −0.286 2.61× 306 months
G2 1.117% 6.011% −36.264% −0.480 2.10× 306 months
G3 0.904% 5.599% −34.714% −0.553 1.82× 304 months
G4 1.633% 10.894% −57.486% −0.217 2.95× 307 months

Three things stand out and all of them matter for the multi-account question.

The grids did not reduce drawdown in any way you could use. G2’s maximum drawdown is −36.26% against buy-and-hold’s −62.07%, which looks like a 25.8 percentage point improvement across sixty-six years. It is not one. G2 simply held far less of the account than buy-and-hold did: its average notional exposure was 34.73% of equity and its peak was 64.40%. Measured against the capital it actually had at work, the grid was not more defensive than owning gold — it earned 1.117% a year while buy-and-hold earned 7.399%, and its Sharpe ratio was negative (−0.480 against buy-and-hold’s +0.215). This is the same result we documented separately in our grid versus trend-following evidence guide, including the bookkeeping bug that had to be fixed to get it right. The mechanics of why a ladder fills up and stays loaded are in our grid trading mechanics guide.

The trend sleeve is the only one that genuinely cut the drawdown, and it did so by holding cash: −31.774% against −62.074%, and the shortest underwater stretch of the six at 191 months.

All four grids lost to buy-and-hold by a wide margin — between 0.904% and 1.633% a year against 7.399%. Over 66 years, G3 finished at 1.82× and buy-and-hold finished at 117.29×. The gap is not a rounding difference in favour of one system; it is a factor of sixty-four.

That is already the shape of the answer. But the interesting part is what happens when you combine them.

How correlated are the sleeves?

Monthly return correlation across all 801 months:

BH TSMOM G1 G2 G3 G4
BH 1.0000 0.8769 0.7283 0.7184 0.6993 0.7391
TSMOM 0.8769 1.0000 0.4696 0.4625 0.4213 0.4865
G1 0.7283 0.4696 1.0000 0.9967 0.9831 0.9963
G2 0.7184 0.4625 0.9967 1.0000 0.9880 0.9881
G3 0.6993 0.4213 0.9831 0.9880 1.0000 0.9697
G4 0.7391 0.4865 0.9963 0.9881 0.9697 1.0000

Read the grid block carefully, because it kills the most common version of this idea. G1 against G2 is 0.9967. G1 against G3 is 0.9831. G2 against G3 is 0.9880. Four grids running at different spacings on four accounts are essentially one strategy split four ways. Their monthly returns moved down together 36.33% of the time, which is almost exactly the same figure as buy-and-hold against any of them.

The lowest correlation in the entire matrix is TSMOM against G3, at 0.4213. That is the only pair in the table that is meaningfully less than one, and it is still high enough that the two lose money together most of the time.

What diversification ratio actually came out

The standard measure: the average volatility of the parts divided by the volatility of the whole. A value of 1.0 means no diversification at all; 1.5 would mean a genuinely different mixture.

Portfolio Members Portfolio vol Diversification ratio Avg pairwise corr CAGR Max drawdown
BH+TSMOM BH, TSMOM 14.385% 1.0321 0.8769 7.868% −40.513%
BH+G2 BH, G2 10.290% 1.0614 0.7184 4.389% −50.311%
G1+G2+G3 G1, G2, G3 6.814% 1.0034 0.9893 1.169% −40.459%
TSMOM+G1+G2+G3 TSMOM, G1, G2, G3 7.367% 1.1665 0.7202 3.016% −33.388%
TSMOM+G1+G2+G3 (no rebalance) TSMOM, G1, G2, G3 11.120% 0.7728 0.7202 6.016% −31.584%
All five BH, G1, G2, G3, TSMOM 8.877% 1.1311 0.7344 3.948% −39.885%

The best ratio in the table is 1.1665, for the trend sleeve plus three grids. In plain terms: mixing a trend account with three grid accounts reduced volatility by about 17% compared with the average of the individual sleeves. That is a real reduction, and it comes from the cash months, not from correlation going negative.

The three-grid portfolio, which is what most multi-account books actually look like, scores 1.0034. Essentially none — and its CAGR is 1.169%, because three grids are one grid with extra steps.

Note the “no rebalance” row, which scores 0.7728 — worse than one. That is the honest face of the “just split it across accounts and don’t touch it” plan: a static split of a trend sleeve and three grids holds more than the rebalanced version does, and it earns less for it. The rebalanced version wins on volatility, the static version wins on CAGR (6.016% against 3.016%), and neither wins on the thing you actually wanted, which was protection.

The trend sleeve earns its keep in a different way. BH+TSMOM has a maximum drawdown of −40.513% against buy-and-hold’s −62.074%. That is not because the two accounts offset each other month by month. It is because the trend account was simply in cash for much of the worst period.

The 1980-1999 episode is the real test

A drawdown statistic computed over 66 years tells you what happened once. The question that matters for capital allocation is what happened in the one episode that actually mattered, and how the split behaved there.

From the January 1980 peak to the July 1999 trough — the run in which gold took nineteen and a half years to find a bottom:

Sleeve Change over the episode Worst drawdown within it
BH −62.07% −62.07%
TSMOM +2.01% −23.85%
G1 −42.59% −49.68%
G2 −29.15% −36.26%
G3 −28.89% −34.71%
G4 −50.70% −57.49%

Now split the capital between a grid account and a trend account and see what the split buys:

Grid sleeve weight Trend sleeve weight Result over the episode
0% 100% +2.01%
25% 75% +1.33%
50% 50% +0.04%
75% 25% −3.23%
100% 0% −29.15%

The result is monotone and unsurprising once you see it. Every dollar you move from the trend account to the grid account costs you on the order of 0.3 percentage points of that episode’s outcome, and there is no threshold effect and no sweet spot.

But notice how different this looks from the previous version of this article. The grid lost less than buy-and-hold over those nineteen and a half years (−29.15% against −62.07%), which is the opposite of what the corrected numbers say about the grid’s long-run edge. That is not a contradiction. A grid that is mostly out of the market loses less in a long bear market and also earns almost nothing in a long bull market — G2 returned 1.117% a year across the full sample for exactly that reason. A strategy that holds cash for most of a decline is not protecting you; it is simply not participating.

The trend sleeve’s +2.01% over nineteen and a half years is not a return anyone should expect from a strategy. It is what sitting in cash looks like during a twenty-year gold bear market, and it is the reason the sleeve has a −31.774% drawdown instead of −62.074%.

Do the accounts go under water together?

This is the part most people never check, and it is the one that decides whether a split is worth the operational cost.

Measure Months (of 802)
Grid (G2) under water 529
Trend (TSMOM) under water 592
Buy and hold under water 602
Grid and trend under water together 509
All three under water together 509
Grid and buy-and-hold together 529

The grid and the trend sleeve were simultaneously below their own previous peaks in 509 of 802 months, which is 63.5% of the entire sample. The overlap is not partial. There is no stretch of history where one account was comfortably in profit while the other was bleeding.

If you had run the split through 1980 to 1999, the two accounts were underwater at the same time almost the entire time. Two dashboards, two statements, one position.

The rolling 36-month correlation between the grid and the trend sleeve tells the same story: 700 windows, median 0.5413, range −0.1552 to 0.9988, and only 4 of the 700 windows were negative at all — all four in late 2025 and early 2026. Whatever the regime, these two almost never moved in opposite directions. There is no “diversification” regime to be found by waiting.

The arithmetic that no amount of account splitting changes

The most basic version of the claim is that spreading capital over several accounts reduces the damage from a bad month. It does not, because the number of accounts is not a variable in the outcome.

In the worst month in the sample, March 1980, gold fell 16.69%:

Arrangement March 1980
One account −16.692%
Two accounts, half the capital each −16.692%

Identical to three decimal places, and not by coincidence: two half-size positions in the same asset are one full-size position. The only thing the split changes is where the loss appears on the statement.

This is worth stating plainly because it is the part that gets lost when the idea is repeated. A capital split across accounts reduces nothing unless the accounts hold different things. If they all hold gold, they are one account with extra steps.

So what does actually help?

Looking at the results, the things that changed the outcome were specific and measurable:

  1. Holding cash. The trend sleeve’s contribution was that it was flat in cash 45.3% of months, at its longest for 96 consecutive months. Nothing else in the test produced a materially different drawdown path.

  2. Trading something genuinely different. The best mixture we found paired the trend sleeve with the grids for a diversification ratio of 1.1665, and the drawdown of the combined book was −33.388% against a grid-only book at −40.459%. That is a real improvement, and it comes from the cash months, not from correlation going negative.

  3. Reducing the size of the whole book. The volatility target, the position size rule, the lot count: these change the arithmetic of how much you lose when wrong. They work on any number of accounts, and they are the subject of our position sizing guide and our risk-of-ruin and drawdown tools. How to read the resulting equity curve is covered in our equity curve guide.

  4. Making sure the accounts are actually different bets. If you want the split to do work, the accounts need to be uncorrelated. If you want to measure it, compute the monthly return correlation between your own accounts. Above roughly 0.8 the split buys you nothing but complexity.

What we cannot tell you from this data

  • Whether any of these strategies is worth trading at all. These are long-only monthly-average simulations. They measure the shape of the risk, not the quality of an entry. A real grid on real tick data with a real spread behaves differently, and we do not publish a claim we cannot reproduce.
  • What happens with leverage. Every sleeve here is unlevered. Leverage scales the equity curve and the drawdown together and does not change the correlations, so the diversification numbers carry over while the drawdown numbers do not.
  • Whether your accounts are more diversified than ours. Ours trade one asset. If you trade gold and, say, a currency pair with no shared driver, you may well have a lower correlation than anything in the table above. That is a measurement you can make on your own data.
  • The cost of running several broker relationships. Spreads, commissions, financing and the operational load of monitoring more than one account are all real and none of them are in these numbers.

The bottom line

Running three grid accounts instead of one does not diversify anything. Their monthly returns correlated between 0.970 and 0.997, and the diversification ratio of the three-grid portfolio was 1.0034.

The only sleeve in the test that changed the drawdown was the one that held cash, and it changed it by 30 percentage points. The grid and the trend sleeve were under water at the same time in 509 of 802 months, so splitting capital between them gave you two statements showing the same loss.

And the corrected grid numbers are the part worth carrying away: a grid on monthly average gold prices returned between 0.904% and 1.633% a year over 66 years, against buy-and-hold’s 7.399%. If your account’s grid looks better than that, the difference is very likely in the simulation rather than in the strategy.

If your accounts are all trading gold, the number of accounts is not a risk decision. If they trade genuinely different things, measure the correlation before you trust the arrangement.

Sources and data

The monthly gold price series used here is the World Bank Pink Sheet monthly average for gold, updated November 3, 2026, covering 1960-01 to 2026-10. We publish the underlying files so the results can be checked:

Both files are published under CC BY 4.0. The underlying price data and the references for the time-series momentum and trend-following literature are listed in the sources section below.


All figures on this page are computed from the published data files by a published script, and independently re-derived from those files by a second implementation. Results are illustrative, based on monthly average prices, and do not represent the performance of any live account. Past results do not indicate future results. Nothing here is financial advice, a recommendation, or an offer to manage money on your behalf.

Frequently asked questions

Does splitting trading capital across multiple accounts reduce risk?

Only if the accounts trade genuinely different things. When every account holds the same underlying asset, the capital split is arithmetically identical to one larger account: in our test the worst month, March 1980, produced exactly -16.69% on one account and exactly -16.69% on two half-sized accounts. The number of accounts changes the paperwork, not the exposure.

How much does diversification help when the accounts trade the same asset?

Very little. Across 802 months of gold prices, a monthly-rebalanced five-strategy portfolio had a diversification ratio of 1.131, meaning portfolio volatility was 13.1% below the average of its parts. Running three different grid spacings together gave a ratio of only 1.003, because those grids had a monthly return correlation of 0.983 to 0.997 with each other.

What actually reduced the drawdown in our test?

Holding cash, not holding more accounts. A 12-month time-series momentum sleeve was flat in cash for 45.3% of months and cut the maximum drawdown from -62.07% to -31.77%. The trend sleeve and the grid sleeve were under water at the same time in 509 of 802 months, so they did not offset each other.

Should I run a grid account and a trend account at the same broker?

From a risk standpoint the broker does not matter, because both accounts carry the same underlying gold exposure. What matters is that the two strategies are not the same bet. If both accounts are grids at different spacings, you are running one strategy with extra steps: their monthly returns correlated between 0.970 and 0.997.

How do I check whether my own accounts are actually diversifying?

Compute the correlation of the monthly returns of each account. If it is above about 0.8, the accounts will lose money together and the split buys you nothing but complexity. If it is near zero or negative, the split is doing real work. You can work through the method on our grid mechanics guide and the position sizing guide.

Sources

  1. World Bank — Commodity Price Data (The Pink Sheet), monthly, updated November 3, 2026 — accessed 2026-11-10
  2. World Bank — CMO Historical Data Monthly (direct download used for this analysis) — accessed 2026-11-10
  3. Moskowitz, Ooi & Pedersen (2012) — Time Series Momentum, Journal of Financial Economics 104(2), 228-250 — accessed 2026-11-10
  4. Hurst, Ooi & Pedersen (2017) — A Century of Evidence on Trend-Following Investing, Journal of Portfolio Management 44(1), 15-29 — accessed 2026-11-10
  5. CME Group — Gold Futures Contract Specifications (100 troy oz, $0.10 minimum price fluctuation) — accessed 2026-11-10

Use the free tools

Open-source calculators — no signup required.

Use the free tools
Risk disclosure — read before acting on anything above

All content on this site is provided for educational and informational purposes only. Nothing published here constitutes investment advice, a recommendation, or an offer to buy or sell any financial instrument. Trading foreign exchange, CFDs, and precious metals involves a high level of risk and may result in the loss of your entire capital. Past performance of any strategy, robot, or expert advisor (EA) — whether simulated, backtested, or live — is not indicative of future results. You are solely responsible for your trading decisions. If you are unsure whether a strategy is suitable for you, seek independent advice from a licensed financial professional in your jurisdiction.

Educational content only. Not investment advice. Trading forex and CFDs carries substantial risk of loss.