The most common accusation against an expert advisor is that it holds many positions at once, and it is close to useless on its own. Across the 26 default audits in our catalog, 22 hold more than one position at the same time. Only 7 of them get a red structure lamp. The count of open tickets is not what separates a breakout system from a grid. The geometry is, and the geometry is three different things that look identical in a screenshot of the terminal.
Our structure pass walks the trades in entry order and, for each new entry, looks only at positions already open on the same side of the same symbol. It then sorts that entry into one of three buckets.
The red lamp needs the third bucket to be both substantial and dominant, at least 20 adverse adds and more adverse adds than adds with the move. Mere intensity, meaning many simultaneous positions with a high stacking share, earns a caution and not a red. That split was not free. It came from a measurement in August 2026 where the older intensity rule handed the same red to a breakout system and to a genuine averaging system, which is a rule that stops meaning anything.
Put four audits side by side and the buckets do the separating on their own.
Gold House logs 4,620 split tickets against 813 adds with the move and 136 against it, with at most 4 positions open at a time. More than half of all its entries are stacked, and the structure lamp is informational. The stack is bookkeeping, four sub strategies writing their own tickets.
Gold Atlas logs 4,057 adds with the move against 1,364 against it, at most 5 open, with a stacking share of 0.527. Thirteen hundred adverse adds sound like a grid until you see that three times as many entries went the other way. Its structure flag reads up to 5 simultaneous positions and says in as many words that this is not a grid by itself.
Lizard logs 3,827 split tickets, 4,572 adds with the move and 695 against, with up to 10 open at once. That is a caution, driven by intensity rather than geometry, and the audit says so in the flag text.
Quantum Queen X logs 1,346 adds against the move and 3 with it, up to 17 positions open, and 40.1% of all its entries are adverse adds. That is a red, and it is the only one of the four where the word grid is earned. The four systems all hold multiple positions. One of them averages down.
The extreme case is Scalping Robot Pro with 42,534 adverse adds against 32,744 with the move. It clears the red rule, but only by a ratio of about 1.3 to 1, which is a far narrower margin than the raw count suggests. Pulse Engine, which we hold unlisted because its source report predates our own data world, sits on the other side of the same boundary at 542 adverse against 648 with the move, which leaves it 107 entries short of the red rule. Two audits, one caution and one red, and about a hundred entries between the two verdicts. A threshold is a threshold and we would rather name it than hide it.
There is a second, older structure test that compares lot sizes after a loss with lot sizes after a win. It works on closed trades, and on the systems it was built for it works well. Logan raises size after a loss in 28.8% of cases against 4.5% after a win, which is a caution. Waka Waka, also unlisted for the same reason as Pulse Engine, reads 31.9% against 5.7%.
Now read Quantum Queen X on the same test. It raises size after a loss in 0.0% of cases, after a win in 0.0% of cases, and its median size ratio after a loss is exactly 1.0. The martingale lamp is ok. A closed trade test cannot see escalation that happens inside a basket that never closes, and Queen X escalates inside the basket. If we had only that test, the clearest averaging system in the catalog would read clean on structure.
That is the whole reason the entry geometry pass exists. It reads the basket while it is open.
Honesty about our own instrument. The detector also records the median size of an adverse add relative to the average size already open, which is meant to catch a grid that doubles as it goes. Across the 26 default audits that number is 1.0 in 18 of them, absent in 7 because those systems never make an adverse add at all, and above 1.0 in exactly one. Quantum Emperor reads 1.25, and it is also unlisted.
So on the listed catalog this field has never once separated anything. The work in every red lamp we publish is done by the direction count, not by the size ladder. That may say something true about the current generation of gold expert advisors, which tends to add at flat size rather than double. It may equally mean the measure needs a longer or a different sample before it earns its place. We are not going to claim the first reading when the second is just as consistent with 26 audits.
Structure lamps and the underlying counts are published for every audit we run. The grid category page lists the ones that earned the label, and the free browser check reads your own report.