A honey hole is a prediction, not a discovery. Club 49 reads public federal records, scores the ground where several independent lines of evidence land in the same place, and hands you the shortlist. Nobody has confirmed a deposit at any of them. Below is every input the United States instrument uses, how far each one reaches, what it cannot see, and then the shallower six-input model that covers the rest of the world.
Each row links to the agency that publishes the data. The last column is how many of the 1,270 holes that input actually applies to, which is the honest version: some of these are rare, and a rare signal counts for more precisely because it is rare.
| Input | What it tells the model | Applies to |
|---|---|---|
| Recorded placer workings | Ground that has produced before, and how much came out of it. | 1,250 of 1,270 (98%) |
| Lode and hard-rock records | The source upstream that shed metal into the drainage in the first place. | 656 of 1,270 (52%) |
| State geologic map | The host rock underfoot, and whether it is one of the favorable ones. | 1,239 of 1,270 (98%) |
| Tertiary channel gravels | Buried ancient river courses. The richest ground the model knows how to find. | 50 of 1,270 (4%) |
| Serpentinite and greenstone belts | Belt rock that gold systems in the West keep turning up alongside. | 23 of 1,270 (2%) |
| Quaternary faults | Structural control, the plumbing mineralizing fluids traveled through. | 367 of 1,270 (29%) |
| 3DEP LiDAR terrain | Bench height and distance above the modern stream, measured not guessed. | 941 of 1,270 (74%) |
| Floodplain width and gradient | Where a river loses its grip and drops the heavy material it was carrying. | 1,150 of 1,270 (91%) |
| Stream bend geometry | Inside bends, where the current slows on the short side of the curve. | 1,227 of 1,270 (97%) |
| Confluences | Where a second drainage arrives and everything it carried has to settle. | 249 of 1,270 (20%) |
| Elevation profile | Five points across the site, so the model reads the shape and not one number. | 1,236 of 1,270 (97%) |
| HUC12 watershed | The catchment that feeds the site, and everything upstream inside it. | 1,239 of 1,270 (98%) |
| Live stream gauges | Current flow, so you know the water before you drive out to stand in it. | 1,165 of 1,270 (92%) |
| Burn history | Post-fire seasons move more material downstream than quiet ones do. | 78 of 1,270 (6%) |
| Surface management agency | Who owns the ground you would be standing on. | 1,270 of 1,270 (100%) |
| Mining claims | Whether somebody already holds the mineral rights where you are looking. | 437 of 1,270 (34%) |
Every hole carries a score and a geology score. The geology score reads the rock, the water and the terrain alone. The overall score also weighs access and claim status. The two never agree: access pulls the score below the geology on 914 holes and lifts it above on 356. That gap is the useful part, and so is its direction: the ground is promising and the paperwork is not, or the other way around.
No overall score anywhere, in this set or the worldwide one below, is higher than 69. There is no 100 because the model does not claim certainty it has not earned, and a number that tops out at 100 invites you to read the top of the list as a guarantee. The geology score is read before access is weighed and is not held to that ceiling: 8 holes carry a geology score above 69, the highest being 76. The top-ranked hole is one of them, and its card says so: geology 76, and 69 once legal access is taken into account.
Surface ownership comes from the federal surface management layer, checked per hole.
| Surface owner | Holes |
|---|---|
| US Forest Service | 390 |
| Private | 342 |
| Undetermined | 295 |
| BLM | 118 |
| State | 54 |
| Bureau of Indian Affairs | 19 |
| National Park Service | 19 |
| Bureau of Reclamation | 16 |
| USACE | 8 |
| LG | 6 |
| Fish and Wildlife | 2 |
| Department of Defense | 1 |
Where the layer had no clear answer, the app says so rather than guessing. That is what the check-before-you-go class below is for.
| Status | Holes |
|---|---|
| Open public ground You may legally prospect here, inside the usual rules. | 356 |
| Claimed An active mining claim covers it. The minerals belong to the holder. | 223 |
| Private land Shown so you know to stay off it, not so you can go. | 568 |
| Check before you go The ownership layer had no clear answer. Confirm it yourself. | 58 |
| Restricted A park, monument or withdrawal where collecting is closed. | 65 |
It cannot tell you there is gold there. It tells you the evidence stacks up and the setting is right. Pan-test first, every single time.
It is not a claim search of record. Claim data is a snapshot of a database that changes weekly, and the only authority is the BLM. Check the claim before you break ground.
It does not know what happened last season. A road washes out, a gate goes up, a parcel trades hands. The records catch up months later.
It is not a substitute for the rules where you are standing. Motorized dredging, camping, and what you may take home vary by state, by forest, and sometimes by creek.
Everything above is the United States instrument: fourteen weighted inputs, up to 37 measured fields, surface ownership on every hole and a claim check against the BLM. It is the deep version, and it reaches exactly as far as the federal services behind it do, which is the United States and nowhere else.
The workings and the host rock are real everywhere. Most of those services are not. So ground outside the United States is scored by a second, shallower model built from the six inputs that exist worldwide, and every one of those holes states how many of the six it actually got, right next to its score. The two are never presented as the same measurement.
| Worldwide input | What it tells the model, and where it comes from | Measured on |
|---|---|---|
| Recorded placer workings | How many catalogued placer workings sit within 5 km, and how large the biggest recorded producer among them was. USGS MRDS and OpenStreetMap | 5,479 |
| Recorded lode workings | The hardrock occurrences within 5 km. A lode is the source that sheds metal into the drainage below it. USGS MRDS and OpenStreetMap | 5,479 |
| Host rock | Whether the bedrock under the site is a lithology that carries gold, read from the mapped geologic unit. Macrostrat | 5,420 |
| Local slope (500 m) | How flat the ground is across a 500 m cross, because gold drops where moving water slows down. AWS global terrain tiles | 5,440 |
| Local relief (500 m) | The height range across that same cross, which is where a bench or a break in the slope sits. AWS global terrain tiles | 5,440 |
| Elevation band | Whether the site sits in the band most placer ground occupies, roughly 200 m to 1,100 m. AWS global terrain tiles | 5,451 |
Two of those are a coarse 500 m terrain cross and are deliberately not called the grade and slope-break the United States model uses. Those are measured along the stream from NHD flowlines and off 1 m LiDAR. Same idea, different instrument, so they never share a name.
Ten of the fourteen inputs have no global equivalent, and one of them is the one the United States feature is built on. They are named on every hole rather than quietly dropped:
land ownership · active mineral claims · live streamflow · burn history · mapped bedrock faults · upstream catchment · stream bend geometry · floodplain width · confluences · buried tertiary channels
There is no global cadastre. Nothing outside the United States can tell you who owns the parcel you are standing on, and there is no worldwide register of active mineral claims. Live streamflow is United States gauges only. So a worldwide score describes the ground and says nothing at all about the paperwork.
Prospecting law contradicts itself across borders, which is why one generic line would have been wrong almost everywhere. A sluice makes you non-recreational in the Yukon and is ordinary practice in New Zealand. A metal detector is banned inside Tasmania’s fossicking areas. In several countries small-scale mining is reserved to nationals. So access is researched by hand, per jurisdiction, against a named and dated source: 32 of the 110 countries are done.
Ground in a country that is not finished yet carries ACCESS NOT CHECKED in the most severe styling the app has, never green, saying in words that the law has not been verified. That is deliberately not a soft “check local rules”, which reads as reassurance. A Researched access only toggle hides everything unverified, and the build refuses to ship a site that would stay silent or claim a permissive rule with no source behind it.
Club 49 carries sourced places in far more countries than it models holes in. Legends, fossil ground, caves and ghost towns reach 133 countries. Modelled honey holes reach 110. Those are two different counts of two different things, and the app never adds them together or quotes the larger one to describe the smaller.
Every United States hole opens a References and sources panel that links the exact records behind that specific piece of ground. A worldwide hole has no such panel, because it has fewer records to show: its popup names how many USGS MRDS records and OpenStreetMap workings sit within 5 km, the mapped geologic unit it is standing on, how many of the six inputs it actually got, and the dated source behind its access rule. Either way, nothing in the model is a number we ask you to trust. If a hole is wrong, the source that made it wrong is one tap away and you can say so.
Coordinates stay inside the app. Counts, methods and sources are public, locations are not, so a hole cannot be scraped off this page and sold back to you by somebody else.
Open the map