Exploration's DeGrussa-Hall Problem

I was going to call it the Monty Hall Problem because it was analogous, but a) the Monty Hall Problem is already a thing, b) the pun might have been strong because of the whole Monty following DeGrussa, c) anyway, you get what you get.

This treatise on exploration follows on from an intriguing discussion after Jacquie Coombs posted an image of costs per metre drilled (inflation adjusted to CPI) and asked what was driving up costs. The second dataset was the brownfields vs greenfields discoveries.

Obviously, the first influence is the cost drillers charge to drill a metre of rock. But in terms of total exploration spend the above graph shows costs per metre going up, and total metres down since about 2022.

The discussion digressed from these simple facts into a wide range of topics, both on-topic, and some random grab-bag of the usual gripes. The one that intrigued me, and for which took a little while to formulate an answer is the one about 'de-risking' projects, and my claim that a lot of explorers tend to 'up-risk' their projects and sink in more money (upping overall spend per metre drilled) without really delivering more mineralisation (discoveries being a contentious issue about when, who, and how they are made).

The Monty Hall Problem

Suppose you're on a game show, and you're given the choice of three doors: Behind one door is a car; behind the others, goats. You pick a door, say No. 1, and the host, who knows what's behind the doors, opens another door, say No. 3, which has a goat. He then says to you, "Do you want to pick door No. 2?" Is it to your advantage to switch your choice?

It is counterintuitive, but the probabilities say one should change your choice, and increase your odds of winning the car. Monty Hall problem - Wikipedia

Let's reframe this for mineral exploration, without stretching the analogies so thin they render the whole thing too absurd, and i pose the DeGrussa Hall Problem;

Suppose you have a mineral exploration project, with three prospects. They are ostensibly equally likely to contain mineralisation. You can choose which prospect to drill, but you only have enough money for one deep drillhole. Let's assume you choose Prospect 1. But before you decide to commit, you undertake a geophysical survey which reveals that Prospect 3 does not have the likelihood of success in finding mineralisation. Do you change your drill choice to Prospect 2 or do you maintain your original choice and drill Prospect 1?

If this was a true game of random chance (as we suppose it might be given the initial conditions assigned equal probability of success at drilling any of the conductors) at striking it rich, then the similarity to the Monty Hall problem would mean we should not switch. The central options are highlighted where the host randomly chooses a door and reveals the goat. In this case, switching drill choice after a target is eliminated by further information coming to light would, all things being equal, result in worse outcomes (ie; 1/3 chance to 1/6 chance).

eg;

Here the analogy is stretched, because the Monty Hall Problem relies on a host which knows where the goat is, and reveals it on purpose as part of the game (hence, where a lot of interpretation as to what the motivation of the host is arises). Given no one knows where a mineral deposit lies, if a target is revealed to not be likely to carry ore, one may presume that switching your drill target would be a losing proposition.

However, let's analyse the situation a bit further. We must remember that we undertake a geophysical survey to increase our chances of finding mineralisation by increasing information around the suite of targets, devaluing some and increasing others. We presume that this reduction in search space will result in improved outcomes - "de-risking" - but it may not actually do so. The key here is the concept of randomness in mineral deposit location.

Non-random Distributions

We should all agree that mineral deposits are not strictly randomly dispersed throughout the crust. They form due to (generally) unobservable processes in deep time, resulting from "mineral systems", the components of which we can measure via their effects upon rocks that result in measurable changes (alteration, in the very broadest sense, to encompass physical, mechanical, chemical and isotopic changes).

Given mineralisation is non-randomly distributed, there is a case that selection of drill holes should not occur according to the process, above, whereby new information is akin to a game host randomly selecting a door to reveal the barren prospect. We may envisage this as a game host which is a godlike being, revealing a structured result to us when we ask a question of the rocks with our science. If we see the revelation, of one of the prospects being barren, as random revelation then we are misconstruing the game, and in fact, misconstruing the data itself, as revealing the component information space of the mineral system (assuming there is one, but that's a separate issue).

Therefore, I would say, it is appropriate to consider the probability space herein as being one where a choice is revealed to us in a structured manner, and we should consider our choices of which prospect to drill as being closer to a Monty Hall problem (switching wins) versus not (switching loses).

It is by this process that the addition of additional geophysics will be seen to 'de-risk' exploration. But here we have to ask whether probability of success is actually changed by adding new information, and whether probability of success can therefore be massaged by adding ever more information prior to drilling

We consider that adding information de-risks the project. By the analogy of the DeGrussa Hall Problem, conditional probability means you should roll dice without being wedded to your original choices made with old information, and should instead reassess ostensibly equal prospects after gaining new information. However - and this is crucial - by only making a choice to drill before undertaking the survey, and then changing that choice after we should see increased success. It would be bold of anyone to do this in reality, right?

Layering Data and Cost

My point that explorers act in a manner they believe 'de-risks' projects, but 'up-risk' projects is that the DeGrussa Hall Problem exists in a potted example. In reality, what tends to happen is that geophysical information expands search spaces, not contracts it.

The DeGrussa Hall Problem envisages a world where geophysical tools are deployed to quantify the physical properties of a known set of prospects, and divine amongst them those that fit the model and expected features of mineralising systems, and discard those whose properties (signatures) do not. In reality, geophysical surveys find multiple anomalies, which we tend to see as good. In reality we might consider this the Kalmbalda Type Problem;

Suppose you have a mineral exploration project, with four prospects. They are ostensibly equally likely to contain mineralisation. You can choose which prospect to drill, but you only have enough money for one deep drillhole. Let's assume you choose Prospect 1. But before you decide to commit, you undertake a geophysical survey which discounts two targets, but reveals six more anomalies, ostensibly similar to the original four. Your third and fourth prospect are now revealed to not be prospective, and can be eliminated. Do you change your original choice?

So, you now have eight prospects from the application of one geophysical method (in this case, surface EM). We then add another geophysical method, that reveals another set of new information and changes the selection of prospects, adding some and discounting others. Do you change your choice of drill target?

One can repeat this with a plethora of data layers. We intuitively, and certainly we rhetorically maintain that adding more data de-risks our project (increases likelihood of success). But what is true is that by adding EM, we turned our initial 1/3 choice into a 1/8 choice. That's added risk, counterintuitively, because it has added targets.

Again, we must consider is this the application of randomness, implying we must stick to our guns to resolve the best probabilistic outcome, or the result of the revelation of structured information, meaning we should change our choices to increase our odds of success? I'll let the conditional probability analysts argue over this thorny conundrum for a few decades, but this is the reality of mineral exploration.

Mineral exploration adds layers of data, like a torte cake, believing that more information equals less risk, but it can perversely result in a multiplex of choices, that are additive and cumulative, resulting in additional risk.

Decision Paralysis

One consequence of proliferating one's prospects is that of decision paralysis.

The DeGrussa Hall problem is predicated upon one ostensibly false premise: that we have only enough resources for one hole. This is true for a junior explorer on its last legs. An example we lionise is Sirius Resources, prior to the Nova-Bollinger discovery. We are led to believe that the company had no more money, and it was a do or die RC drilling campaign, and they chucked a handful of RC holes in to their best handful of targets and the rest is history. A story of pulling victory from the jaws of defeat, a tale of science, and the ultimate prospectors luck, etc etc. Let's set that porky pie aside to cool off for a minute.

However, with 500 junior explorers listed on the ASX, many with sub-$1M, this is a very true position many tiddlers and penny dreadfuls now find themselves in. Enough money for one moonshot hole, or a slow, debilitating slide into delisting or phoenixing as a crypto mining / biotech / fintech / dotcom shell.

We have explored the devilish conundrum of de-risking a project with geophysics, remote sensing, space divination (soil helium anyone?), or DGPR dowsing, and the multiplication of choice.

One consequence of multiplying targets is that, with only enough cash for a moonshot Hail Mary, delist-or-DeGrussa hole, how do you gain the risk appetite to just pin your ears back and go for it? Bearing in mind the Directors' responbilities is to keep the shell going, odds are that intuitively it will be easier to motivate a Hail Mary on 1 in 3 versus one in 15.

We must now circle back to the apocryphal tale of Sirius. Run by Messers Bennett et al., renowned nickel geologists, with no money in the bank, were they really about to run out of money, or were they just spending the last couple of shekels before raising more money? Let's be honest, there's always more money. Lack of success wouldn't have been the end of Sirius. They had nickel prospective ground at Narndee, gold at Polar Bear under a lake. Lets not fool ourselves with mythology here.

And let's not fool ourselves about these 500 listed shells, that risking most of the money in the tin on a Haily Mary is the end of the company in question - it rarely is. Let's ask ourselves a better question: is it worth risking a Haily Mary, or taking the certainty of a slow death by listing fees and bleeding out publicly on the bourse?

If we're in a risky profession, our job is to take risks, not be parlaysed with fear about our choices. I say, take a stab. There's always an antimony project.

Quality of Information

The remedies for increasing risk via multiplying targets include ensuring that the new information is high quality.

For example, if you have 8 different conductors (all else being equal as to the quality of that layer of information) there is an argument to be made that the quality of the other layers of information needs to be made equal, to eliminate the situation where one perceives that the targets are equal in quality when they may not be, with some are comprised of poor quality data or interpretations.

If two targets have different levels of data quality of, say, drilling assay reliability (particularly with old assays), quality of logging, or location accuracy and veracity, then it is clearly riskier to drill the target with the less-verified information because it may be false, misleading, erroneous or wrong.

QAQC is an important aspect in assessing quality of targets and of actually de-risking exploration. This goes well beyond simple assay ranking, which is what we generally understand in QAQC.

I would argue that not all interpretations (which includes geology mapping and drill hole logging) is equal throughout the history of a project, and understanding the quality of these datasets is critically important.

For mature, well-drilled areas with lots of logging throughout time, it is particularly important to consider the heterogeneity of logging information and the subjective errors which that can introduce. Verifying, re-logging where appropriate, and maintaining a coherent database (and logging code appropriate to the rocks!) is important to eliminate these sources of error - and risk.

Ranking Information

One method that is frequently used to parse a large variety of targets which become available due to exploration data collection is a ranking exercise. This may be prosecuted in any number of ways, these days also including the use of Machine Learning (otherwise portrayed as Artificial Intelligence though there is no reasoning behind it at all) which is a computer-facilitated 'Weights of Evidence' based method.

An illustration of an hypothetical process is given below. Here, additional data layers are added (new information) and thresholds applied. In this example, thresholds are lower as the process continues because the value of a target based on only one layer of data is lower but it remains a target in its own right. One can see that the more layers a target has, the greater the score it attains. But, crucially, it doesn't discount other targets.

The targets that fail the hurdles failed the individual gates of that method for defining a quality target. For example, it might be a low-conductance EM plate, a structure which isn't coherent with the mineralisation-causative deformation set, etcetera.

A ranking process should be a process of structured analysis and reasoning undertaken by a geologist, or at least these days a ML campaign supervised by a geologist, in order to prevent a complete GiGo (garbage in/garbage out).

The use of such a process derives a better risk weighted set of choices based on the information available. However, the difference in rankings is up to the scores assigned, which comes down to either a calculated choice by the human based on their knowledge base, tastes, and feelings/instinct, or a set of algorithms defined by the machine, set by a human in the first place.

Note also that if one defines a drill target after the process is wholly complete according to a certain score, some of the original targets which were discarded on the first set of information, might sneak back in. This is the Monty Hall problem again; do you maintain a score of 50 as the drill threshold, even if it causes you to change your original choices?

Pattern Drilling

Does pattern drilling or testing every target really address the risks? Or is just drilling everything a way of leaning into the risks? If we consider the hypothetical original problem, which was 1 in 3, and we 'de-risked' the project by adding targets, does just raising money to drill everything result in a better outcome, or a worse outcome?

Arguably, by drilling everything to eliminate the hypothesis that any single conductor might be nickel, and eliminating the prospectivity of the ground for competitors once you have done the job (which is a thing), have we not just caused an increase in total spend, in metres drilled, and reduced our success rate?

The logic of mechanised and production line exploration is solid and clear, but industrialised mechanistic exploration might be a reason for current poor performance because it is agnostic of risk.

This exploration machine is composed of several cogs;

• overly thorough exploration strategy
• overly-resourced exploration budgets
• brownfields thinking
• big data driven philosphies (assay all elements, do all spectral techniques)
• no stone unturned, no anomaly undrilled philosophies
• leave nothing behind for the next holder

Hot Money Doesn't Care - Should We?

We must now turn to another human aspect of the exploration industry to assess risks versus rewards, but from a purely financial aspect, having now confused ourselves about purely probabilistic and methodological issues.

As an industry (and as commentators, thought-leaders and analysts) we wonder why drill metres (and drill costs per metre) are up and discoveries are down. We talk about risk, search-space, and ontology.

However, in a boom, where money is easy to come by as it chases any return in any dog-arse project, there is money around that alters the balance of risk versus reward, and incentivises poor risk taking and over-drilling.

Taking the Kambalda Type Problem, we have 8 prospects and 2 discounted prospects, for 10 in total. We start 2024 with poor nickel prices and an atrophied budget capable of one glory shot at the stars. Then 2025 rolls around, nickel goes for a tear, and the board raises hot money worth a small Carribean nation's GDP, because that's the way of the world.

Suddenly, we're drilling all 10 prospects, beyond the original 4, in order to leave no stone unturned and no conductor undrilled (and for good reason; nickel is a tricky beast and hides well amongst black shales, so you kind of have to drill everything).

Would we debate the declining discovery rates and lack of success, the poor performance in drilling metres and drilling costs, when it is clear we should be doing is what juniors do: raise licks of cash when the market is hot, hire idiots, pattern drill everything, and not worry about the probability. Open every door in the Monty Hall problem, you're bound to win the car.

Right?

We also have to consider that a lot of exploration is gambling, and gambling is for the big win. Mature, well-run mining companies which make a lot of money from producing do not explore, they develop, because the gamblers rely on selling out a winning hand to those capable of financing mines.

A key question we need to ask is this: Is a poor success rate really a problem with the system, or is it the actual system?

Majors aren't putting smart money into risky gambling and spray-and-pray, and they have the money to burn on taking these risks. So who cares if tax drivers burn their play money?

The system is operating as intended: non-financial companies exist as conduits for risk-taking by high nett worth individuals whose money lacks decent returns and has nowhere to go (boom cycle) and would otherwise be lost to taxes (oh noes) or diminishing returns (over-leveraged industrials). If wealthy people burn chunks of cash having a punt, and in the meantime create jobs for drillers, geologists, and parasitic service ecosystems, is this not pure neoliberal trickle-down in action? Isn't capitalism grand?

From a commie governmental point of view (to paraphrase ur-capitalists), and indeed from a geologists' and drillers' point of view, the worse the discovery rates the better, because we get paid more money to find the same amount of copper, and the funding of mines occurs outside of the exploration ecosystem anyway. Literally who cares.

Bad Behaviours

I think by now I have demonstrated a very poor understanding of probability, which puts me right in the middle of the bell curve for exploration geologists.

What I wanted to get across is that the exploration industry exists to spend money. We must accept this as the ultimate premise. We exist to spend shareholders' money.

What we must do as an industry is, yes, try to spend that money wisely. However, we often don't spend it wisely, and the worst thing we can do, often, is to misapprehend what risk is, and what risk-taking is.

For example, I believe the worst possible thing we can do as an industry is to hold onto dud tenure for too long. Every sub-block or square kilometre of dud ground you hold onto is wasted money, wasted effort, and increased risk. Why do we hold onto ground? That's another whole post, but one aspect is not rigorously understanding where mineralisation is certainly not. This is rooted in poor thinking as much as anything, and is exacerbated by decision paralysis: not admitting your ground fails the prospectivity hurdle and should be dropped is, to my mind, a cardinal sin. You want to see industry wide inefficiency in expenditure per discovery? Look no further than this.

The next worst thing we can do is to take money from sophisticated investors and their brokers to have a gamble, and not put it all on red. Not drilling is the worst possible outcome, even worse than drilling a dud hole. Why? because if you're in theDeGrussa Hall Problem and you have a 1/3 or even a 1/33 chance with the companys last $150K for a moon shot, and you say "I'm not going to open any doors" then you've lost, 100% of the time. All we have to do is ensure we drill cheaply, but effectively.

Another bad behaviour we do is deploy geophysics to solve problems that can be solved via cheaper and alternative means. Often simple, basic means, like mapping, regolith mapping, thinking deeply about mineral systems.

We also tend to sink too many holes in to chasing insufficient metal. We don't discount targets based on low grade, insufficient metal accumulation along strike or per vertical metre. We drill shoots that are too short on strike and grade accumulation down plunge, and hope to get better.

We drill metallurgically challenging ores into a resource without solving the metallurgy first. This I lay squarely at the clay REE industry's feet. You guys are bloody hopeless. We've got billions of tonnes of clay with, to date, no way home. Tens if not a couple hundred million wasted drilling up resources and never applying metallurgy and logic to determine if it's ever extractable (and quibbling RPEEE about it). You can get in and out of a whole project for $50K.

We drill targets that will never make a company, just to satisfy a bourse or a PR statement made by a director. It's a brave company that just flat-out admits their project will never be a mine and pulls the funding, and goes and does something else. Instead, apples roll off the table quietly and rot under a rug for the next boom to come along.

So, yes, in conclusion, we are here to clip a ticket and get our trickle downs. In a way, we are performers in a macabre theatre of pantomime and kabuki theatre, pretending to be doing things and finding mines, when in reality we are just eking out a survival RPG in end-stage capitalism. We should not be ashamed that we're inefficient, because if we were perfectly efficient in some daft AI utopia where EarthAI and other ML startups just drill a perfectly minimal amount of holes and find everything in 5 years, the industry will be gone and metal prices will be nothing and we'll all be watching holovids in space. But until then, we have to take risk-weighted gambles.

Let's just try to do a good job while we still have them.