← Back to the bots
XAU-PF01

Six-system portfolio · Gold

DATA SHEET · VERSION 1.0 · 26/08/2026
EH-TRADE
Trading bots
ehtradebots.com

IDENTIFICATION

MarketGold (XAUUSD)
Timeframe4 hours
DirectionLong and short · six single-direction systems, combined into a portfolio
TypeMulti-strategy portfolio · market entry, exits in ATR multiples
Period analyzed11/06/2015 – 06/06/2025 (10,0 years)
Simulation engineMetaTrader 5, hedging account
Costs includedSpread 15 points · slippage 5 points · commission of USD 0,70 per trade. Check your broker's commission: the usual practice on gold is to charge per lot and not per trade, and that can be quite a bit more expensive than what is loaded here
Reference capitalAll the published numbers are calculated on an initial capital of USD 100.000, with a fixed position size of 0,23 lot per USD 100.000 of account. The run was done at 2 lots and the amounts are rescaled to 0,23 in direct proportion, which is exact with fixed position size The lot is measured on the instrument’s standard contract: if your broker uses a different contract size, write to us and we will give you the equivalent.

RESULT OF THE COMPLETE BACKTEST

All the numbers in this block come from a single run, over the ten years the portfolio was built with. The fourteen months that followed were measured on a different data series and are further down, in their own block: nothing in this document adds the two runs together.

ANNUAL RETURN
11,77 %
annual average on initial capital
COMPOUND RETURN
8,10 %
CAGR — the conservative reading
MAXIMUM DRAWDOWN
3,74 %
$ 7.255 measured against the account peak
RETURN / DRAWDOWN
2,17
CAGR divided by maximum drawdown %
METRICFULL PERIOD
Total profit$ 117.715,52
Average annual profit$ 11.771,55
Average monthly profit$ 989,21
Trades2.237 (≈ 18,7 per month)
Winning trades1.025 · 45,82 %
Losing trades1.212 · 54,18 %
Profit Factor1,37
Sharpe ratio1,49
SQN (Van Tharp)2,33
Average win$ 429,03
Average loss$ 265,71
Payoff ratio1,61
Average result per trade$ 52,62
Average duration of a trade8,6 4-hour candles · ≈ 1,4 days
Market exposure68,61 %

ACCOUNT EVOLUTION

100k150k200k2016201720182019202020212022202320242025

Simulated account day by day, from $ 100.000 to $ 217.716 in ten years. Dotted line: the initial capital.

RISK — THE UNCOMFORTABLE NUMBERS

Maximum drawdown$ 7.2557,26 % of initial capital · 3,74 % measured against the account high, which is how the simulator calculates it
Longest losing streak14 tradesin a row, without a single winning one in between
Without recovering the peak286 days7,84 % of the period
Worst trade− $ 1.722,17stop executed
It is right less than half the time45,82 %1.025 winning trades against 1.212 losing trades
Market exposure68,61 %up to six positions open at the same time
Losing components2 de 6in the forward stretch · see note 3
Significance of the forward segmentt = 0,99219 trades are not enough for the test to be conclusive · see note 2

RESULT BY YEAR

YEARRESULTPROFIT FACTORWINS
2015+ $ 1.396since 11/06
2016+ $ 11.128
2017+ $ 4.617
2018+ $ 4.390
2019+ $ 10.124
2020+ $ 30.386
2021+ $ 6.514
2022+ $ 15.208
2023+ $ 4.019
2024+ $ 8.464
2025+ $ 20.772up to 06/06

Eleven calendar years, none in the red. The amounts come from the daily equity curve of the same run; the total differs from the overall profit by around $ 700 because of the position that was still open at the end of the period. Gold went through four regimes in those ten years —sideways between 2015 and 2018, two rallies and the 22 % correction at the end— and the portfolio closed positive in all of them.

OUT-OF-SAMPLE VALIDATION

Two different tests, and I keep them apart because they do not come from the same place. The first is the stretch of the run above that the system did not see while it was being built. The second is fourteen later months, measured on a different data series. Note 1 has the detail of how this portfolio was chosen out of the hundred candidates.

Out-of-sample segment of the run above
Out-of-sample trades689 of 2.237 · 30,8 % of the total
Profit Factor1,30 · against 1,37 over the complete period
Sharpe ratio1,29 · against 1,49
SQN (Van Tharp)0,59 · against 2,33 — the sharpest degradation
Winning trades44,56 % · against 45,82 %
Payoff ratio1,61 · identical to the full period's
Where the maximum drawdown happenedin here · the $ 7.255 were made on data the system did not see
Forward segment, on a different data series
Trades219 · Profit Factor 1,21 · + 24,08 % annual
Losing components2 de 6
Statistical significancet = 0,99 · 219 trades are a short sample for the test to be conclusive
Tests on the selection, against the other 99 candidates
Out-of-sample Return/Drawdown retention46 % · median of the 100 finalists: 21 %
Out-of-sample equity curve stability0,67 · the highest of the 100
Monte Carlo · risk of ruin6,1 % · the lowest of the 100 finalists
Trendless regime test+ 5,89 % per year · 2015–2018, with gold flat

WHAT YOU NEED TO KNOW BEFORE BUYING IT

I would rather you buy knowing this. If something here stops you, it was not for you — and that is worth more to me than a sale.

It won exactly where it could not be tuned

Out of a hundred candidate portfolios, this one won none of the in-sample rankings: it sat mid-table. It won the three tests that measure conditions it had not seen — out-of-sample return/drawdown retention, curve stability and risk of ruin. That is the exact opposite of what happens to an overfitted strategy, which shines inside and goes dark outside. Strong evidence, not a guarantee.

The forward stretch is fourteen months, not ten years

219 trades with a t-statistic of 0,99: it is a short sample and on its own it neither confirms nor refutes anything. The weight of this data sheet is not there, it is in the ten full years and the four different gold regimes above, all of them positive.

Two of the six components lose money in the forward segment

Of the six systems, two closed in the red over those fourteen months. I did not take them out of the portfolio on purpose: dropping the worst one after seeing the result is picking the winner with Monday's newspaper. If one of them is genuinely broken, it will show up in the next window and this data sheet gets updated.

Gold rose fourfold over the period analysed

From $ 1.180 to over $ 4.400. Part of the result comes from trading an asset that went up. What offsets that reading is the 2015–2018 stretch: gold was flat for three and a half years and the portfolio still made 5,89 % a year.

The fixed position size falls short when the asset rises

With stops calculated as multiples of ATR, the dollar risk per trade follows the price of gold. In ten years gold multiplied by four, so the 0,23 lots stay conservative during the cheap-gold years and are tight in the last ones. It is a deliberate decision: I prefer the size that survives the worst stretch over the one that optimizes the average.

The worst drawdown happened on data the system had not seen

The $ 7.255 maximum drawdown did not happen in the stretch the portfolio was built on, but inside the out-of-sample cut. It is what you want to see: the system's worst moment showed up where it could not have been fitted to anything. Even so it is a backtest drawdown, and the real one can be larger.

The two blocks are not added together

The ten-year backtest and the fourteen-month forward stretch were measured on separate data series, so I do not add them up and I do not average them. Each block says where it comes from and is read on its own.

REQUIREMENTS TO TRADE IT

Hedging accountMANDATORY. The six systems trade in parallel and there are longs and shorts open at the same time
PlatformMetaTrader 5 with algorithmic trading enabled
InstrumentGold (XAUUSD). Check your broker's contract size and tick value before trading
Simultaneous positionsUp to six. Market exposure of 68,61 %
AvailabilityA machine switched on all week, or a VPS (USD 12,80–15/month)
Six different identifiersOne per system. They are delivered already configured
Reference capitalAll the published numbers are calculated on an initial capital of USD 100.000, with a fixed position size of 0,23 lot per USD 100.000 of account. The run was done at 2 lots and the amounts are rescaled to 0,23 in direct proportion, which is exact with fixed position size

Price: USD 700, one-off payment, launch price. Six independent systems trading the same account at once. No subscription, no renewal. The file comes tied to your account number and includes the installation guide and support until it is up and running.

On leverage. All the numbers in this data sheet are measured without leverage, with a fixed position size. The bot can be traded with leverage, or with a different lot size, from the Expert Advisor parameters in MetaTrader 5: that multiplies the result and the maximum drawdown in the same proportion, and the numbers in this document stop applying.

Risk warning. Leveraged trading can make you lose all of your invested capital. Nothing in this document guarantees profits. Trade only with money you can afford to lose.

Nature of the results. All the results in this document are simulations on historical data and do not represent live trading. Simulated results have known limitations: they are built knowing the past and do not necessarily reflect the impact of liquidity, slippage or real execution conditions. No result, real or simulated, guarantees future returns. The costs each simulation was run with are stated in the body of this data sheet, in the «Costs included» row; your broker's costs may be different and change the result.

Disclaimer. This document describes a software product. It does not constitute financial, investment, legal or tax advice, nor an offer or recommendation to buy or sell any instrument. The issuer is not registered as an investment adviser, does not manage third-party funds and has no access to its clients' accounts.

XAU-PF01 · Data sheet · 26/08/2026Internal reference Portfolio 800 · 0.76418 + 0.194715 + 0.318343 + 0.455636 + 0.457413 + 0.2478909