Rules-based XAUUSD trend-following
The gold trading system that shows you when it is losing, and why.
Finemark is a transparency-first, rules-based gold (XAUUSD) trend-following system you run on your own broker or demo account. The dashboard tracks a demo account (no real money), live-updating, on a validated Supertrend H4 strategy, and it reports that result honestly, even when it is behind.
The Strategy Lab publishes honest verdicts on 7 systematic approaches to trading gold. Three of them, the mean-reversion strategies, are labelled dead on gold, because the evidence says so. That is not a bug. That is the entire point.
Backtest results are hypothetical and do not predict future performance. Live track record is demo-account only, under 2 weeks, small sample.
The Strategy Lab leaderboard, ranked by 22-year backtest Sharpe (2004-2026). Strategies without a documented full-cycle Sharpe show a qualitative verdict. Past results are not future results.
The problem
Every gold signal seller promises 90% accuracy. Most have two months of data from a bull run.
Sharpe 1.7 on a one-year bull-run window. Sharpe 0.89 on 22 years that include the 2013 crash. Same strategy. Every seller shows you the first number. We run the second.
You have probably seen the pattern. A provider launches a gold signal channel. The first six months look extraordinary. Then gold has one bad month, the drawdown arrives, the provider says to trust the process, and half the account is gone before the channel goes quiet.
The problem is not that these systems never work. The problem is that nobody shows you what they look like across a full market cycle: the 2013 crash, the 2022 sideways grind that ate positions alive, the years where a trend-following system simply sat in cash and did nothing. No one labels their own failures. We built Finemark because we wanted the longer window, with the honest number, and all the failure modes labelled.
Short track record
Most gold signal providers launched after 2022. Their track record is a single bull cycle, not evidence of a robust system.
Hidden failures
No competitor publishes a leaderboard that includes the strategies that do not work on gold and explains exactly why.
Recency bias by design
Showing a 1-year M15 window instead of a 22-year H4 window is not an accident. It is a selection. We show you both.
How it is different
Radical transparency is the architecture, not the marketing.
Every other gold system puts its best number in the headline and buries the caveats. Finemark does the opposite. The Strategy Lab ranks seven approaches by 22-year out-of-sample Sharpe. Three are labelled dead on gold because the evidence is there. The live dashboard says BEHIND when the live result is behind the backtest, and a small-sample banner is always on while the live record is one week old.
Strategy Lab labels its own losers. RSI, Bollinger, Z-score - dead on gold. Shown plainly, not buried.
Live vs backtest comparison, updated every 5 minutes. The BEHIND verdict is visible by design.
Small-sample warning on every live metric when the track record is under 90 days. Today that warning is always on.
22-year parameter plateau chart. Broad Sharpe stability across 126 parameter combinations is robustness, not luck.
Risk kill-switch visibility. When the daily-loss or drawdown breaker halts new entries, you see it immediately.
How it works
Three steps. No black box.
The system is a deterministic rules-based strategy. No machine learning, no inference layer, no hidden signal. You can read the logic in plain English in the Strategy Lab drawer.
The strategy watches every 4-hour bar on gold (XAUUSD). When the Supertrend indicator (period 10, multiplier 3) flips bullish and no long position is open, it places a long entry at 1% position risk. When Supertrend flips bearish, it exits. Long-only. Worst case: flat in cash. Never short.
The dashboard collects the result every 5 minutes from the live demo account and writes it to the state file. You see the open position, the entry price, the unrealised P&L, and the active signal. The Supertrend line updates on the H4 chart with trade markers.
Telegram alerts fire on every fill and every morning at 08:00 WIB (UTC+7) with a daily P&L summary. If the bridge goes down, a heartbeat alert fires within the next cron cycle. You know what the system is doing, and when it stops.
The risk kill-switch (core/risk_guard.py) halts new entries when daily loss or total drawdown exceeds defined thresholds. It does not block exits - the system can always close a position, never only open one.
The Strategy Lab
A backtest leaderboard that includes the ones that fail.
Most algo tools show you one strategy and call it their best. The Strategy Lab shows you seven, ranked honestly by 22-year Sharpe, with a click-through drawer for each: the plain-English explanation, the formula, a per-year Sharpe heatmap from 2004 to 2026, and the known failure modes.
The Lab also shows the recency-bias trap on the live strategy itself: Supertrend scores about Sharpe 1.7 on a one-year intraday window and 0.89 on the honest 22-year H4 window. The intraday window is almost entirely a 2023–2024 bull run. The 22-year window contains the 2013 crash, the 2022 grind, and fourteen years of mixed regimes. We show you both, and explain why we run the lower one.
Open the full Strategy Lab - freeSharpe ratios are from a 22-year backtest (2004–2026, 33,616 H4 bars). Backtest results are hypothetical and do not predict future performance. Live track record is demo-account only.
Backtest leaderboard, 2004–2026. Past results are not future results.
22-year backtest proof
22 years. Three regimes. One honest number.
The backtest runs from 2004 to 2026 on 33,616 four-hour bars of XAUUSD. It covers the 2004–2011 bull market, the 2013 crash (gold fell from roughly $1,700 to $1,200), the 2014–2018 range, and the 2019–2026 period including the 2022 sideways grind. The live system has run for approximately one week on a demo account, and is currently slightly negative - a number with no statistical meaning at this sample size. We show it anyway.
Backtest equity curve vs buy-and-hold
Long-only Supertrend (gold) sits in cash through downtrends, cutting the deep drawdowns that buy-and-hold (grey) takes in full. Backtest, 2004–2026 - not a prediction.
How the signal reads the chart
XAUUSD H4 candles with the Supertrend line beneath price in an uptrend (gold) and above price after the flip (red). A long entry opens on the bullish flip; the position exits on the bearish flip. Illustrative, not a live quote.
Parameter robustness plateau
Sharpe across 126 combinations of period (7–20) and multiplier (2.0–3.5). No cell falls below 0.75. A narrow peak would be curve-fitting; a plateau this broad is robustness. The live parameter (period 10, multiplier 3) is circled.
Transparency manifesto
We built this for traders who have been burned by black-box systems. So we show you everything.
This is self-directed analytics software. You run it on your own broker or demo account. We never take your money, never pool it, and never trade on your behalf with discretion. In the worst 22-year backtest period the maximum drawdown was −20% - on a $10,000 account, a drop to roughly $8,000. That −20% is a historical backtest figure, not a limit on losses: real trading can be worse, and you can lose all of your capital. The strategy sits in cash during downtrends - long-only means the worst case is flat, not leveraged against you.
We show you the strategies that fail. Including the one we run, when it does.
We display the live result even when it is behind the backtest. Especially when it is behind.
We label every performance figure with its source and its mandatory caveat, in the same sentence.
We never accept, hold, pool, or manage your money. You keep custody of your own funds.
We never promise a return. The subscription fee is flat. Your results are your own.
We never fabricate a user count, a testimonial, or a track record. Today Finemark has one user: its builder.
Every number on this page is labelled with its source and its caveat. Every losing strategy is named, including the years our own would have sat flat in cash. That is not our marketing. That is the product.
In the dashboard
Everything you can watch. Including the numbers that look bad.
Every feature below already runs on the live demo system today. The multi-user SaaS is in development; founding members get early access as it ships.
Entry price, current lot size, unrealised P&L, and active Supertrend signal - updated every 5 minutes from the live demo account.
Gold on four-hour bars with the Supertrend line and green/red trade markers, time axis labelled. Updates every 5 minutes from state.json.
Timestamp, direction, entry/exit price, and realised P&L for every recent trade. All fills from a demo account.
Balance and equity over time since the bot started around 2026-05-25. Small sample. Shown honestly.
Seven strategies ranked by 22-year Sharpe. Three labelled dead on gold. Click any strategy to open the detail drawer.
Plain-English explainer, the exact formula, a per-year Sharpe heatmap 2004–2026, and the known failure modes on gold.
AHEAD / IN-LINE / BEHIND verdict updated continuously. Small-sample warning always visible during the first 90 days.
When the daily-loss or max-drawdown breaker is active, a banner shows it clearly. New entries halted. Exits still execute.
Fill notification on every open and close. Daily P&L summary at 08:00 WIB (UTC+7). Bot-down heartbeat alert if the system stops responding.
Cron-driven status check every 10 minutes. If the MT5 bridge is unreachable, you know within one cycle.
Not yet shipped (roadmap): multi-user accounts, on-demand parameter Lab runs, State API for the Studio tier. The roadmap is not a promise of delivery dates.
Pricing
Three tiers. One honest price each.
The Observer tier is free, and the Strategy Lab, recency-bias explainer, 22-year summary, and failure verdicts will always be public. There are no profit-share fees, no performance fees, no deposit. If gold goes down this month, you pay the same $9. If it goes up, you pay the same $9.
- Full Strategy Lab - read-only, including the verdicts labelled dead on gold
- Recency-bias explainer: 1-year M15 (Sharpe ~1.7) vs 22-year H4 (Sharpe 0.89)
- Parameter plateau chart: Sharpe 0.75–0.97 across 126 combinations
- Static 22-year backtest summary with caveats front-and-centre
- Bear-test results: 2013 and 2022 drawdowns vs buy-and-hold (backtest)
- No live dashboard
- No Telegram, no API
- Everything in Observer
- Live dashboard: equity, open position, entry price, lot size, unrealised P&L, last 30 fills
- Price + Supertrend H4 chart with time axis, live-updating
- Per-strategy deep-dive drawer: explainer, formula, per-year Sharpe heatmap
- Telegram alerts: fills + daily P&L summary (1 chat ID)
- Risk kill-switch visibility
- Small-sample warning banner under 90 days, plus heartbeat status
- Everything in Analyst
- State API endpoint - raw analytics JSON, rate-limited (not a signal service)
- Up to 3 Telegram destinations
- Up to 3 named user seats on one subscription
- Priority heartbeat: bot-down alert within 5 minutes
- On-demand parameter Lab runs (roadmap, not yet shipped)
USD payments via Polar (card, Apple Pay). IDR payments via Midtrans (QRIS, GoPay, OVO, DANA, bank transfer). Prices shown as of 2026-06-02 (USD/IDR rate Rp17,880); the IDR figure is re-checked monthly and finalised at checkout. Studio ships only after the API disclaimer language is reviewed by counsel, or after crossing $5k MRR, whichever comes first. Paid Indonesian sign-ups open once PSE registration is complete.
Subscription fees are a fixed-price software licence. They are not an investment deposit and give you no claim to any profit. We make no guarantee of return. Cancel any time; access runs to the end of the paid period. No refund is issued for a period already started once dashboard access has been granted.
FAQ
Honest answers to the questions you should be asking.
Is this a scam?
Will I make money?
Is this legal in Indonesia?
The live track record is one week and slightly negative. Why should I pay?
How is this different from a signal service?
Why is the live track record on a demo account and not real money?
When will Studio launch?
Founding members
Join while the track record is still building.
Finemark is in its founding-member phase. The live demo system has been running since approximately 2026-05-25, and the multi-user SaaS is in active development.
Founding members who join the Analyst tier now get the $9/month rate locked for 12 months from first payment. If pricing changes as the track record matures, founding-member rates hold. The Strategy Lab will be free for everyone at launch; you will not need to pay to read it. We are not running a countdown timer. We would rather have 30 founders who understand what they are looking at than 300 who do not.
No profit guarantee. No lock-in. Cancel any time. Fixed-price software licence, not an investment.
0 founding members so far. Be the first - the live count appears here when the waitlist opens.
Founding-member rate locked for 12 months from date of first payment. After 12 months, rate reverts to prevailing Analyst pricing. No auto-upgrade. Cancel any time. Subscription is a software licence; it is not an investment deposit and gives you no claim to any profit.