How We Bet, Part IV: Is the Odds Engine a Cure for Common Estimation Errors?

In our experience, yes: It protects us from asymmetric skewing.

It is of a union between projections of future value and bias that asymmetric skewing is born.

What is an estimation error?

An estimation error is the difference between an estimated value and the actual probabilistic value.

Most estimation errors that we run into are products of handicapping. Let’s use Denison Mines as an example.

We use a period of 23 Years in order to calculate Benchmark Odds against which to measure Denison’s Implied Odds. Without the Benchmarks, we would not be able to determine the presence of value.


As you can see, we don’t have any long value bets staring us in the face, but even so, the odds of rising to 0.46 in 2020 are quite good — 66.67%.

Interestingly, that’s not far off the mark of Denison’s intrinsic value at $24.70/lb. U308, which we’ve estimated at 0.41. So it looks like, from a fundamental standpoint, shares can be picked up today at a 11% discount.

That’s great, but where are the estimation errors?

The big estimation errors occur when an estimate of future value is calculated. For instance, take Denison’s valuation at $30/lb. U308. At $30, we think Denison is worth 0.49/share. And if somehow, by some miracle, spot uranium popped to that level, Denison would have a 33.33% chance of rising to its calculated valuation.

But the market is stubborn. It simply will not revalue a stock in concordance with your projections. And it is of a union between projections of future value and bias that asymmetric skewing is born. In other words, your own estimation errors have convinced you that your betting odds are considerably better than they actually are.

The odds matter…


Here’s another consequence of estimation errors: You miss out on the value bets* that intersect with good odds.

*The 0.37 price point was so tantalizingly close to representing a short value bet at great odds, that it would have been a shame for a value bettor to pass up (Note: It wasn’t passed up, obviously.)

How We Bet — Part III: The Value Bet Demystified

In Parts I and II, we covered our internal approach to Value Betting stocks, but perhaps an understanding of ‘Value’ as it pertains to stock betting still eludes you.

What is Value Betting?

Value betting is essentially a bet where your own calculated odds of an outcome are better than those calculated by the bookmaker.

In stock betting, we simulate bookmaker odds with Benchmark Implied Probabilities. If the Implied Probability of an outcome is greater than the Benchmark, we are able to say value is present, which doesn’t necessarily make the bet a good one. If one’s own calculated odds are 4.55 and the bookmaker’s are 9.00, is this a great bet? No. Although there is value, the odds are still poor. We’d take a pass on this bet. Again, we are looking for an intersection between value and good odds. That’s the sweet spot.

The Bookmaker’s Advantage

If you’ve dialed in your Benchmark Implied Probabilities, you will find that true value bets are rare, especially after the Vig. After all, as you well know, the house doesn’t like to lose.

So how is one to win? More importantly, how does one find and exploit value bets?

For starters, you’ve got to develop an algorithm with which to appraise stock prices that is not utilized by bookmakers. Remember, an algorithm is just a set of rules to follow. It doesn’t have to be complex. Our own algorithm is comprised of 20 short lines of code, reducible to 16, that lives both simultaneously in our charting software and in Excel. We could also, in a pinch, run the algorithm in our heads. That’s how simple it is. Our algorithm is easy to take for granted now. Too easy. And I have to remind myself that it took 20 years of trial and error to develop it. In fact, this is the algorithm that convinced me of the relative unimportance of sophistication.

So don’t be discouraged if you are forced back to the drawing board time and again in search of an algorithm that will enable you to join the house in its bets against the squares. It’s worth the blood, sweat and tears.

I actually want you to succeed!

I’m betting against you, but I want you to win, too. I sincerely do. Notice I’m not charging you for anything. I don’t run a fund in which I want you to become an investor so I can charge you a fee. I don’t need you to come to a seminar. I don’t need you to buy a subscription to a special Insiders Only newsletter. And we don’t need your traffic to generate ad revenue.

We won’t necessarily share our secrets with you, but we do want to share the process whereby which we arrived at those secrets, hence this betting series. We want to inspire you to develop your own secrets. We want you to become self-reliant. We want you to shed your predilection for seeking out confirmations of your biases. We want you to kill your gurus. We want you to be radicals bent on reintroducing as many little inefficiencies into the market as you are able. We want you to form your own opinions and place your own courageous bets.

That’s all we want.

More in Part IV.

How We Bet — Part II: Coeur Mining Case Study

When betting, we look for an intersection between value and good odds. Value, you will recall from Part I, is determined by identifying a clear discrepancy between benchmark implied probability and case study implied probability, where the odds against a trade are lower in the case study than in the benchmark.

Coeur Mining is an interesting case study, as we have a lot of historical data with which to work. This enables the Odds Engine to report statistics with a high level of confidence.

Let’s begin with some foundational statistics:

The period over which data was analyzed was 29 years. Over that period of time, Coeur experienced 11 cycles, 6 of which were down cycles and 5 of which were up cycles. The average down cycle lasted 3.7 years while the average up cycle lasted 1.4 years. In other words, Coeur shares trade in a down cycle 75.86% of the time.

Counterintuitively, value bets are not to be found on the long side in up cycles, which undercuts the old saying, ‘the trend is your friend.’ Rather, the long trades with the best odds and highest implied probability of success occur at the tail end of down cycles of the typical length. Interestingly, the reverse is not true for playing the short side. Short bets with the best odds require that one is willing to aim for a modest target while trading good size in a down cycle. Value diminishes quickly beyond the S4 price point.

A quick aside: The algorithm with which we calculate price points R6-R3 and S3-S6 is proprietary. Without that algorithm, we would not have a functioning Odds Engine, and for obvious reasons, we can’t share the math. We publish the Engine results in an effort to give readers a sense of our process — a process that enables us to see what the fundamental data available to the public doesn’t say.

There is a means by which one may trade when the odds are in one’s favor a high percentage of the time. This means that one is trading with the house against the general public as well as against contrarians — both segments of the betting population with a penchant for betting on their favorites at almost always the wrong time. They bet their favorite at what they thought was the bottom, only to find it was the middle of a down cycle. They bet their favorite when the cycle finally turns up, only to find that the bet was made at or near the top. Both groups have a poor understanding of the implied probability of virtually every trade they enter and the house eats them for lunch.

Betting isn’t easy: Humans weren’t designed to play the odds effectively, as the odds rarely conform to one’s expectations of what is logical. Wise guys know this and keep an eye out for the bets the public deems sure things — because the public thinks of their favorites as logical winners and the logical winners rarely defy the odds. Hence, we look for those price points where probability and value intersect.

Discovering those price points that represent value at good odds is a laborious, pain-staking process. And if one isn’t willing to put in the time to develop a method and process for betting well a little over 50% of the time (hopefully a little higher), one’s probability of success will be low.

You aren’t betting horses…

Just like at the track, when betting stocks by buying shares, you are betting against everyone else. But that’s where the similarity with horse betting ends. The payouts for favorites at the track are typically pretty dismal. Only squares bet on the favorite all day long. You can win all day at the track betting favorites and not ever be able to pay your mortgage.

With stocks, however, one can, and I dare say should, bet on the price point with the best odds, where value is glaring. But in order to ensure that the payout is satisfactory, particular attention must be paid to position size. To succeed long-term, one’s goal cannot be the price points with a low implied probability of success, especially if dismal odds are required to compensate for a small position.

More in Part III.

How We Bet — Part I

We do a lot of handicapping, but we acknowledge that won’t make us better traders. The value of handicapping ends with stock picking. Once stocks are picked and opinions are formed, one must gauge whether or not one should place an actual bet.

We are primarily value bettors. This is not to suggest we only place value bets. Rather, the better part of investment capital is allocated to those issues where a clear discrepancy between benchmark and implied probability is evident, while the remainder is allocated to long shots. In our portfolio, the ratio of value bets to long-shots is roughly 70/30.

For every issue we study, odds are calculated for 12 discrete price points; from the odds, an implied probability for each price point is generated. The implied probabilities are then weighed against benchmark implied probabilities for the sector to which the issue belongs.

Whether a bet is placed on the long or short side is determined by whether or not its cycle — as we calculate it — is up or down.

We maintain benchmark probabilities for the following sectors: gold, silver, platinum, uranium, oil services, and agriculture.

Over the course of the next week, we will be publishing herein a visual representation of the math with which our betting forms (totes) are produced. As odds are read differently depending upon the part of the world in which you live, output will be in decimal, fractional and American forms. We are Americans, but we think in European fractional terms, so when discussing odds, it will be primarily in the terms of fractions (e.g. 13/1 against…).

In addition, it’s important to note that we also compute conditional probability sets, but for simplicity’s sake, these won’t be included on our published totes.

Is this really how we place our bets?

It is. It’s also how we quantify our exits. However, the totes that we will be publishing will be those wherein value is present a high percentage of the time relative to a benchmark, though this isn’t necessarily the price point at which we will choose to exit. Our own tolerances allow us to exit at price points that have a fair confidence interval with a preset margin of error.

Stay tuned!

URANIUM: Now the Real Carnage Can Begin!

Now that the majority of our initial Odds Engine downside targets have been reached (see here and here and here and here), the real carnage can begin:

And from our post of 12 December 2019 — Cycle Death and the Uranium Junior — were you curious about our estimated time-frame for a bottom:

P.S. It’s important to remember that the 900% return you expected on the stock you bought at $0.50/sh. and which has declined by 90% to $0.05/sh., now is represented by a rise to a mere $0.50/sh. There are still good trades to be had down the pike, but this requires that position size scales proportionately as shares cheapen.

$POTRF: SOPerior Fertilizer Corp. — 2020 Tote Board

“It is wise not to treat something that is very very unlikely as if it were impossible.” -Nigel Turner

This is a live example of the current tote for POTRF. Null values do not imply that additional price-points are improbable. Rather, we do not have historical precedents on which to run our tote algorithms.