What we are building, and what is still to do
This page tracks the Expert Advisors we are building, one line per system, and says plainly which steps are finished and which are not. A system appears here before it has a record that meets our publishing bar, so there are no performance figures on this page. When a system earns a product page, it moves to the Products section with its figures and the broker and period they come from.
How we build
We start from trading ideas that are known to work on gold and FX, study what actually drives them, and build our own version with our own rules, our own filters and our own risk control. A version is not finished when it looks good in a test. We check it in seven steps, and this page shows where each system stands on each.
- Entries and exits are written down as explicit rules and checked on real ticks: do they do what the idea says they should, and when?
- Exit tuning is a separate step. Exit settings are optimised on purpose, on the longest history we have and on the alternative price paths below, and the result is checked on data that was not used for tuning. Writing the exit rules down is not tuning, and until a system has this step its exits are only as good as our first guess.
- Extended history means the longest tick history we hold, on more than one broker feed.
- Other price paths make the same market harder: older years, prices rescaled to today's level, the path inverted, the path run backwards, and inverted and backwards together. A real edge survives these. A fitted one does not, and we would rather learn that here than live.
- Risk sizing and forward testing come last: lot size by percent of balance with a ruin and recovery profile, then a demo account on live prices.
Where each system stands
Done Tested, mixed Partial Pending Last updated 2026-11-20.
| System | Entries | Exits | Exit tuning | Extended history | Other price paths | Risk sizing | Forward test |
|---|---|---|---|---|---|---|---|
| Sanderling GoldGold, hourly chart | Done | Done | Tested, mixed | Tested, mixed | Tested, mixed | Pending | Pending |
| Lapwing GoldGold, five-minute chart | Done | Done | Tested, mixed | Tested, mixed | Tested, mixed | Partial | Pending |
| Merlin GoldGold, fifteen-minute chart | Partial | Pending | Pending | Pending | Pending | Pending | Pending |
| Curlew GoldGold, twelve-minute chart | Partial | Partial | Pending | Pending | Pending | Pending | Pending |
| Fieldfare GoldGold, four-hour chart | Partial | Partial | Pending | Pending | Pending | Pending | Pending |
What each column means
- Entries. The rule that starts a trade is written down and checked on real ticks.
- Exits. The stop, target, trail and closing rules are written down and checked on real ticks.
- Exit tuning. Exit settings optimised on purpose, on the long history and on the alternative data below, with the result checked on data not used for tuning. Writing the exit rules down is not tuning.
- Extended history. Tested on the longest tick history we hold on more than one broker feed, including the years a system was not built on.
- Other price paths. Tested on the same market made harder: older years, prices rescaled to today's level, the path inverted, the path run backwards, and inverted and backwards together. A real edge survives these; a fitted one does not.
- Risk sizing. Lot size by percent of balance, then checked against a ruin and recovery profile.
- Forward test. Running on a demo account on live prices after the build, before any sale.
What is still pending, system by system
Sanderling Gold
Four small engines in one: buying and selling daily short-term extremes with an indicator-based exit, a fade on the hourly chart, and a four-hour continuation entry that rides a trailing stop.
Where it stands: First version built and checked on two broker feeds; then tested on older years and on inverted and backwards price paths, where its result was weaker.
- Ten alternative exit settings were tested, six on the long side and four on the short side, each on three development periods and two later validation periods, under rules written down beforehand. None beat the first version on all of them, so the first-version exits stay. A looser exit on the short side came close and will be revisited when its entry filters are refined.
- Tested on older years and on inverted and backwards price paths. The short side held up across the older years; the long side, which carries most of the recent result, did not, and the whole system lost its edge on an inverted path. It will be treated as a current-market system with a clear regime note.
- Percent-of-balance sizing and the ruin and recovery profile.
- Two of the four engines are only partly where we want them; their entry filters are still being refined.
- A demo forward test.
Lapwing Gold
Builds the overnight range, then takes the first break of it in the direction of the larger trend, with the stop on the far side of the range and a fixed multiple as the target. One trade a day, flat before the close.
Where it stands: First version built and checked on two broker feeds, then tested on older years, on inverted and backwards price paths, and with six alternative exit settings.
- Six alternative exit settings (a lower and two higher targets, a break-even step, a trailing stop, an earlier close) were tested on four development periods and two validation periods under rules written down beforehand. None beat the first version across all of them, so the first-version exits stay. The test will be repeated if the entry rules change.
- Tested on older years and on inverted and backwards price paths. It held up on an inverted path and not on a backwards path or in older years, so it is a current-market system and will carry a regime note.
- A demo forward test, and percent-of-balance sizing checked with a ruin and recovery profile.
Merlin Gold
Waits for a pullback into one of two long moving averages and trades back out of it.
Where it stands: Entry level found; the rule that chooses which pullbacks to trade is not written yet.
- The selection rule for which pullbacks are traded.
- Exits, then everything after them.
Curlew Gold
Trades only in the direction of a slow trend, with short holds of about half a day.
Where it stands: Direction rule written; entry timing and exits are not.
- Entry timing inside the trend.
- The exit rule.
- Everything after them.
Fieldfare Gold
A trend system that reverses between long and short and adds to winners.
Where it stands: Structure understood; its direction state is a blend of trend measures that is not written yet.
- The direction rule.
- A learned or composite state may be involved; if so it would be built as a trained model, not a fixed rule.
- Everything after that.
What the tests have shown so far
Both first versions do what they were built to do in the market they were built in, 2023 to 2026. Turn the price path around or go back in time and they behave differently:
- Inverted prices. A genuinely two-sided edge should survive this. Lapwing Gold does. Sanderling Gold does not as a whole, which tells us part of its recent result comes from gold's long rise.
- Backwards prices. Neither holds. That is common for systems that act on the direction of recent moves, and it is one reason we call them current-market systems.
- Older years. Sanderling Gold's short side held up across them; its long side did not. Lapwing Gold did not hold up.
We do not hide this. A system that is tied to the current market can still be useful, but it has to say so on its page, and it needs a way to switch itself off when the market stops looking like the one it was built for.
Your question: has proper exit optimisation been applied?
Partly. Lapwing Gold has had one round: six alternative exit settings, tested on four development periods and two later validation periods, with the adoption rule written down before any result was seen. None beat the first version across all of them, so its exits stay as they are. Sanderling Gold has had one too, with the long and short sides tuned separately because they behave differently: ten alternative settings, and again none beat the first version across every test, so its exits stay. A looser short-side exit came close and will be looked at again with the entry filters. The other three systems on this page are not far enough along to have exits to tune.
Reading the states
- Done - finished and checked.
- Partial - some of it is finished; the list below says what is left.
- Pending - not started.
These are research builds. Nothing here is for sale, none of it is financial advice, and a first version that looks good in a test tells you little about how it will trade later. Every one of them will be tested on a demo account before any sale.