GOLD DIGGA PRO ULTRA

PRO ULTRA v2.0: PRO’s two sleeves, as a strategy

ULTRA is no longer a preset workshop. It is GOLD DIGGA PRO ver2.0 in strategy mode, so the Strategy Tester shows every number we publish, on your chart.

One sleeve at a time (🔬 Lane under tuning, because TradingView nets positions):

• Scalps (default): the measured Gold 15m Scalps
• Regulars: the overnight breakout, 15m only
• Magic: retired, no orders

Want both? Add ULTRA twice.

The defaults ARE the measurement. No paste:

XAUUSD 15m  2022-01 → 2025-05   167 tr  +$42,803.67  PF 1.37
MGC1!  15m  2023-01 → 2026-10   199 tr  same list as measured
Regulars    2010 → 2026-10    4,262 tr  record PF 1.14 · DD 23.7R

Regulars is a trend-year sleeve: before 2010 the rule lost. Its record is in R, net of $0.30/oz, so the tester’s PF will read differently.

Preset strings are retired. “Tune a preset for your asset, paste it into PRO” is withdrawn: our own test showed that the best setting on one window does not predict the next. The preset inputs are removed. To keep your own setups, use Settings → Defaults → Save as… (sharing as strings is gone, on purpose). The tuning menu stays. If you experiment, tune on one window, then test once on a later one you never touched.

Also new

• Runs on every symbol. Off gold, a red UNMEASURED box (last setting hides it)
• 🕓 Regular clock: Etc/GMT-3 by default
• Compounding now defaults OFF
• PRO ver2.0 visuals; build stamp ULTRA-2.0 b3 at the top centre

After updating

1. Reset is required. The removed inputs shifted every later setting up four places. Settings → Defaults → Reset settings (or remove ULTRA and add it again)
2. Tick the terms box
3. Delete and recreate your alerts

PRO ULTRA v1.6.1 — three fixes, one clock

A maintenance release. Registry unchanged (84 keys); every existing string loads.

The session clock follows daylight saving. Default zone is now Europe/Athens instead of the fixed “UTC+3”. It also drives the daily trade counter, which used to parse a fixed number out of the “UTC±N” text and could not follow the calendar.

“Enable trade manager” is only a display switch now. With it off, the strategy received no trades at all — the hand-off to the strategy sat behind that switch — and limit diamonds jumped back to the order-block bar. The trade is always computed now; the switch hides the tooltip.

Measured. Three 15m scalp presets (gold, BTC, ETH), two disjoint windows each:

v1.6                     +99.9k
v1.6.1                  +104.9k   gold identical to the cent
all 5m bars on 15m        +8.7k   built, measured, not shipped

That last line is the one worth reading. On a 15m chart the 5m engine reads only the last 5m bar of each candle. Reading all three looks like the obvious fix — it was better in both windows nowhere. The presets were tuned on the current behaviour and it is part of what they are. It is now written into the header, so nobody “fixes” it again without re-tuning the 15m lanes on it.

After updating: switch 🕓 Session timezone from “UTC+3” to Europe/Athens once. Winter months of a backtest move by that hour — that is the fix working.

PRO ULTRA v1.6 — one opt-in gate, and what it measured

v1.6 changes exactly one thing against v1.5: it adds the Δγ absorption gate. The registry grows from 81 keys to 84 — dgOn, dgPushBars, dgDecay.

With the gate off, v1.6 is byte-identical to v1.5. Not “close enough”: the request for 1-minute data sits inside the toggle, so with dgOn=0 nothing is requested and no code path differs. Any v1.5 measurement reproduces exactly.

Old strings still parse. 88-, 85-, 81- and 80-key strings all load; new exports mint at 84, and a string from an older registry arrives with dgOn=0.

No sleeve ships with the gate on, and PRO has never seen it. Here is why, because the reasoning is the more useful half.

Δγ asks whether the flow that pushed price into the order block is still running, and drops the diamond if it is. Four measurements:

BTCUSDT.P  5m   +5,061   5 events
MGC1!      5m   −3,657   5 events
XAUUSD     5m      +64   0 events
ETHUSDT.P 15m       −4   0 events

BTC came back positive with the sign holding in both halves — the first order-flow filter in this project to measure positive at all. Then MGC, an equally fair second sample on the same timeframe class, came back the other way, and split by its own halves the sign flips.

Mechanism plausible, edge unproven. The BTC result rests on five events, and three correct vetoes in a row is not rare when two-thirds of candidate trades lose anyway. Once a fair second sample came back negative, that number stopped being distinguishable from luck.

So the gate stays default-off and opt-in on the bench, and its forward-validation plan is downgraded from confirm before shipping to watch, expecting nothing.

It joins a long line of entry vetoes measured honestly and declined — CVD twice, the cooldown family 0 for 8, the session block. The standing lesson: filtering entries has never paid here. When you are exposed and how much you risk has.

#manual-ultra ch. 44.