Cheap Means Two Things — VIX and Nat Gas Need Opposite Rulers

1. The market is always right.

2. Every price is already determined.

3. Every view stays open to all probabilities, the way quantum mechanics does.

4. All evolution happens through repetition.

"Buy it when it gets cheap."

Almost everyone says this. I said it myself for years. Then one day I looked at the sentence again and realised it settles nothing at all.

Cheap compared to what?

Until you answer that, "buy it when it gets cheap" is an instruction with no content. So I took the same sentence and held it against two markets — VIX and natural gas. The answers came out opposite.

That is what this piece is about, and at the end I leave you a three-step procedure you can apply to whatever you happen to be watching.

A word before we begin, because it decides the order of everything else.

An indicator someone drew on a chart is, in the end, someone's opinion dressed up as a line. What matters more is the thing the chart is already made of: the live map of how each market moves with, and against, every other one. Correlation and inverse correlation are not decoration on top of price — they are price, updating in real time, with every participant's money already melted into it. The markets themselves, read against one another, are the true indicator.

Human-made indicators come second. Not never — second. That order is the whole point.

I say this with some apology to the people who make a living selling indicators. I am not saying your work is worthless — I build indicators myself, and I know the care that goes into them. I am saying something narrower and, I think, fair: an indicator cannot certify the question it is being asked. Sold on its own, without the market relationship that makes it applicable, it is a good tool handed over without the one instruction that decides whether it helps or misleads. The honest version of that sale includes the order.

Here is why the order matters, using this piece as the evidence.

VIX is itself an indicator, and a well-built one. Yet against the Nikkei it measures −0.09. Put VIX on a Japanese chart and every signal it gives you is noise, no matter how carefully the indicator was constructed. Nothing inside the indicator can warn you about that. Only the correlation map can, and it tells you before you place a single trade.

Natural gas is called an energy market, and it sits at 0.16 against oil as a whole. The label says one thing and the relationship says another. Any indicator you hang on gas while thinking of it as oil will be right about its own arithmetic and wrong about the market.

Nasdaq and the Dow sit at 0.72 — the only meaningful gap inside the index group. That single number tells you which two charts are worth keeping side by side and which pairs are the same information counted twice. No oscillator will ever tell you that, because an oscillator only ever sees one chart.

And roll cost, the trap in the middle of this piece, is not visible on any indicator at all. It lives in the relationship between one contract month and the next. The market states it plainly; you only have to look where the market keeps it.

So the order is: first ask what this market is and what it is tied to, then choose the ruler, and only then reach for an indicator to read it faster. Reverse that order and the indicator inherits every mistake you made before you switched it on. It will still draw a confident line. That is precisely the danger.

Used in the right order, human-made indicators are genuinely valuable. They compress work you would otherwise do by eye, they keep you consistent on days when your judgement is not, and they let you check the same condition across a hundred charts in a minute. What they cannot do is tell you whether the question you are asking belongs to that market at all. That answer only ever comes from the market's own relationships.

That is the spirit of everything below. Nothing here is a line I invented. It is all counted from what the markets actually did.

1First, let us agree on the words

Four terms will make everything that follows easier. If you already know them, please feel free to skip ahead.

What VIX actually is

VIX is calculated from S&P 500 option prices. What it says is: how much movement does the market expect over the next 30 days.

Think of options as insurance. When people get nervous they will pay up for insurance. Option prices rise, and the VIX calculated from them rises too. That is why it is called the fear index.

The important part is that VIX is not a forecast that prices will fall. It is a reading of how nervous people are right now. A thermometer, not a prophecy.

I measured whether it really moves opposite to stocks

Words are cheap, so here are the numbers. Daily bars, and the three figures are the 30-day, 90-day and 1-year windows.

These are correlation coefficients. Reading them is simple.

The US indices sit between −0.7 and −0.85. That is a genuine inverse relationship. Days when VIX rose were days when stocks fell.

This is where people go wrong most often

Look at the bottom line. The Nikkei is −0.09. That is close to zero, which means essentially no relationship at all.

VIX is an American fear index. It is built from American options and it tracks American stocks. "When VIX rises, stocks fall everywhere" is a common saying, and this table shows it is not true.

So before you lean on VIX, check whether the market you are watching is American. Watching Japanese stocks with VIX on your screen is like adjusting your heating by reading the thermometer next door.

Futures, margin, tick

This piece talks about futures, so let me settle three words.

And if your deposit runs short, the broker closes the position for you. That is forced liquidation. There is only one way to prevent it — put in enough money.

Two kinds of number appear in this piece

This distinction changes the conclusion later. To say it in advance: measured by the average, something looks like it works; measured by the median, there is a stretch where it loses.

2VIX — cheap compared to itself

First we fix the ruler. For VIX, I used this one:

Not an absolute number, but the instrument's own recent range. Whether VIX is 12 or 15, if it sits at the low end of the past year, we call it cheap.

Measuring from 1990, there were 1,687 such days. I counted what happened afterwards.

Twenty trading days is roughly a month. Seven times out of ten it was higher.

Why it comes back

If you stop at the number, you will be caught off guard the first time it fails. You need the reason.

VIX is not a price. It is the price of anxiety, and anxiety does not last. Neither people nor markets can stay frightened indefinitely. Bad news arrives and it spikes; the news is digested and it settles back.

The other direction works the same way. Deep calm does not last either. When quiet days stretch on, people stop buying insurance, options get cheap, and VIX drifts low. Then something happens and it jumps again.

So VIX cannot travel far in either direction. It gets pulled back toward the middle. This property is called mean reversion.

Stocks are different. If a company keeps growing, there is nothing strange about its price multiplying several times over. There is no middle to return to. That is why the same ruler must not be used on them.

How to see the lower band with your own eyes

I counted where the monthly lows clustered. That is 440 months since 1990.

In other words, when VIX travels down it has generally stopped around here. The all-time low did go further (8.56), but that was rare.

To check this on your own chart:

Draw one horizontal line and from then on you will see it at a glance.

3But futures carry one trap

Up to here it sounds like "buy VIX when it is low and hold." With futures the story changes, and this is the most important part of the piece.

What roll cost is

Futures expire. The September contract ends in September. To keep holding, you must move into October before expiry.

The problem is that the next month is usually more expensive.

Why? Insurance makes it easy to see. Two months of cover costs more than one month, because more can happen in a longer window. VIX futures behave the same way. The further out the month, the more "something will probably happen by then" is priced in.

So every roll means selling the cheap one and buying the dear one. Even if the market does nothing at all, your position quietly shrinks.

I measured how much it shrinks

Over 15.7 years:

The index went nowhere while the rolled position was cut roughly in half, year after year. That entire gap is roll cost.

Stocks have no such cost. Buy a share, leave it alone, and you still own the same number of shares. Futures are different. Doing nothing is itself an expense.

If you do not know this, you can be right about direction and still lose. It genuinely happens: someone decides "VIX is at the floor, I will hold it", waits a few months, and finds VIX higher while the account is smaller.

4So the wick is twenty days long

I subtracted that cost from the gains above. All medians.

Here is how to read it.

The middle column, roll cost, keeps growing with time. The left column, the gain, barely grows past twenty days, because VIX usually does its returning within a few weeks.

So the two lines cross somewhere near thirty days. Through twenty trading days something is left; from thirty trading days the cost wins.

By the average it still looks positive out to ninety days. But that is a handful of large spikes lifting the number. The ordinary single attempt loses once it passes thirty days. This is exactly why the median matters.

And there is one more awkward twist

Roll cost is at its most expensive when markets are calm. When things are quiet the near month gets cheap and the gap to the far month widens.

And a calm market is precisely when VIX is cheap.

The place where you buy well and the place where the cost bites hardest are the same place. That is the real reason a short window is necessary. Hold a good entry too long and the cost that comes with that entry eats the advantage the entry gave you.

Three rules

The third one is hard. The moment of "surely if I hold just a little longer" will arrive. But the table is unambiguous. What waiting buys you is not the spike — it is a certain, measurable cost.

Posting generous margin is a good habit and it prevents forced liquidation. But what it prevents is liquidation, not cost. The thing that limits cost is the calendar.

5Natural gas — cheap compared to the world

I held the same ruler against natural gas. It failed.

Buying in the "bottom 10% of the past year" gave a 50.5% chance of being higher twenty trading days later. The base rate is 49.5%, so that is a coin toss.

Why it failed is the heart of this piece

Natural gas fell from about $15 in 2005 to $1.4 in 2020. Shale supply came flooding in and the price came down structurally.

What happens when you hold a "bottom of the past year" ruler against a market like that?

It keeps triggering all the way down. At $5 it is at the low end of the past year; at $3 it is too; at $2 it is again. Each time, the ruler says "cheap" and you buy.

It was a ruler for catching a falling knife. The ruler was wrong, not the market.

So I changed the ruler

I switched to an absolute level: regardless of its own range, only when the close is below 2.0.

I also fixed how I counted. If you compute "return 120 days later" every single day, today's result and tomorrow's overlap by 119 days. You are counting the same event twice. So I collapsed each consecutive stretch into a single event. By calendar days it is 271 days; the real number of events is 12.

At 180 days all twelve were higher, and even the worst single case was +25.8%.

Natural gas has roll cost too — about a quarter per year over 19.4 years. Even after taking that out, the 180-day median leaves +36%.

Why 2.0 in particular

There is nothing sacred about the number itself. It matters because the area around 2.0 is where production costs sit.

When price falls below cost, producers lose money. Nobody can absorb that indefinitely, so rigs are shut and production stops. Supply falls and price recovers. That is the force which builds a floor.

VIX has nothing like this, because nobody "produces" VIX. What VIX has instead is a different force — human emotion cannot sustain itself for long.

Different forces require different rulers.

And that explanation doubles as the condition for discarding this idea. If shale costs fall further, the 2.0 line falls with them. Do not memorise the line; follow the cost.

6The same ruler across 21 markets

So far this has been about two markets. The question that remains is: which kind is the market I am watching?

So I held one identical ruler against 21 markets — "what if you had bought in the bottom 10% of the past year?" Changing the rule per market would be fitting rather than measuring, so the ruler stayed fixed.

The figures show how many percentage points better than usual the outcome was. Zero means it was no different from buying on any random day.

Returns quickly — 20 days · 120 days

Needs a long horizon

Not clear either way

Buying cheap actually hurt

The most surprising line is the bottom one

In stock indices, buying at your own one-year low did worse than buying on a random day.

Nasdaq is −9.4%p. After buying at the low, the share of cases higher twenty days later was 53.1%; buying on any random day it was 62.5%.

Why does this happen? Two things overlap.

First, an index is an asset that mostly rises. So even a random purchase is already 62% likely to be higher twenty days later. The baseline itself is high.

Second, a one-year low usually appears while the decline is still under way. The low is not printed after the fall ends; it is printed and re-printed on the way down.

So the market where "buy the dip" is used most often is the market where it worked least. What worked in indices was not waiting for lows — it was simply holding.

How to use the table

Now look at natural gas again. In this table it sits under "not clear either way" (+1.0%p).

Yet when the ruler was switched to the absolute level of 2.0, it won 12 out of 12 at 180 days. Same market, different ruler, opposite answer.

So being in the "not clear" group does not mean there is no opportunity there. It means the right ruler has not been found yet.

7Three steps to apply to your own market

If you take one thing from this piece, let it be this procedure.

Step 1 — does this market have a middle to return to?

Ask yourself this:

Step 2 — does this market have a floor?

Step 3 — choose ruler and horizon from those answers

You could have a hundred markets and still only need these two questions. Is there a middle to return to, and is there a floor?

8On size — cheap and light are not the same thing

Many people go looking for markets with low margin. But low margin usually means the contract is small, not that the risk is low.

What you should look at is this:

This is not set by your broker. It is the market's own character, and it is the same wherever you trade. Measured as the median daily range over the past year:

Natural gas moves roughly ten times as much in a day as the euro does.

So if you actually want to do the "hold natural gas for 180 days" described earlier, it has to be a small contract rather than the full one. Reduce the size and each day's swing shrinks with it, which is exactly what lets you stay in.

You are trading size for time.

What you are choosing is not a cheap market. It is a daily swing you can live with.

The arithmetic goes like this:

If step 3 exceeds 30%, three days against you wipes out the margin. You cannot sit in a position for months at that size.

9Five places people go wrong

10How to buy with the worst case already calculated

Everything so far has been about where, by which ruler, and for how long. One thing remains.

How much money do you have to put in?

Some markets have a calculable maximum loss

You often hear that futures losses are unlimited. For a short position that is true, because there is no telling how far a price can rise.

But buying is different. If a market cannot go below zero, then everything you can lose is already fixed.

Put that whole amount in and the phrase "forced liquidation" simply stops applying to you, because your margin can never fall short.

Where to find margin numbers

Margins are set by the exchange (CME, CBOT, NYMEX, COMEX) and your broker adds on top. So there are two numbers to know.

Below are the maintenance margins published by CME Group. I took these directly from exchange data on 3 September 2026, front-month contracts. Initial margin sits somewhat above these because your broker adds to them.

The column worth studying is the micro one. Grain micros run $100 to $225. The point is not that this is pocket change — it is that this method can be carried out without a large account.

For VIX and natural gas the exchange data does not publish the front month, so I use the numbers I read directly from my own broker screen.

There is one fortunate thing about this method.

If you decide to fund the full contract value, you do not need to know the margin at all. Margin is the number that decides when you get thrown out, and if the whole value is sitting there, there is nothing to be thrown out of. What you need to know is not the margin — it is the contract value.

Let us work through VIX

VIX cannot go below zero. So the maximum loss on one contract bought at 16.65 is

Initial margin is $8,525, so funding roughly twice that (1.95×) means that in theory nothing can liquidate you, whatever happens.

And here is the weight of a single step.

So about a sixth of your margin swings in a single day. Fund only the initial margin and you can be gone within days. Funding the full value is not greed; it is what lets you sit through that swing.

The same arithmetic for natural gas

Here the gap opens wide. The margin is cheaper than VIX ($3,321 against $8,525), yet the full amount is nearly twice as expensive ($30,170 against $16,650).

This is where you can see, in numbers, that a cheap margin and an affordable position are two different things.

Reducing the contract brings it into range

The same natural gas comes in three sizes.

The micro is one tenth of the full contract. At around $3,000 you can fund the whole thing.

Reducing size is not timidity. It is what turns the worst case into a number you can actually hold.

This is also the real reason micro and mini contracts exist. People tend to think "I use micros because I have little money", but the order is the other way around.

Reduce the size and you create time. Same judgement, same level, and yet one of you can sit through it and the other cannot.

The mini sits in between. For natural gas the three multipliers are 10,000, 2,500 and 1,000, so you can pick the one that matches what you are able to fund.

Grains have their own minis. If a micro is one tenth, a mini is roughly one fifth — the size to reach for when the micro feels too small and the full contract feels too heavy. Exchange maintenance margins:

Three sizes means three speeds. At the same level with the same judgement, the person in a micro can wait months while the person in a full contract cannot last weeks. What you are choosing is not the market — it is how much time you can afford.

Put the other way: a market with no micro is hard to use this way.

Fortunately, more markets have micros than you might expect. Going through the exchange listings one by one, corn, wheat, soybeans, soybean meal and soybean oil all have them, and so do gold, silver, copper and platinum. Grains are one tenth of the full contract; silver and platinum are one fifth.

The only three without a micro were oats, live cattle and lean hogs. Those three are only candidates when you can carry the full contract size as it is.

There is one thing I must flag, though. That list only means the exchange lists those contracts — not that you can buy them. Which contracts are actually offered differs by country and by broker. It is common for something to exist at the exchange and be missing from your own screen.

So I cannot tell you which micro and mini contracts are supported in your country. Please check with your own broker directly. Three things to confirm: whether they carry it, what the margin is, and what the commission is. Because a micro is small, commission weighs relatively more on it.

Why the yearly chart is the right place to choose from

Whether a price is low should not be judged on daily bars. Daily data is crowded into the last few years, which distorts any claim of "historically cheap".

You want yearly bars. Count how many years closed below today's price and you will know immediately whether this really is the low end. Monthly bars are for looking more closely inside that answer.

I ran this ruler across markets that have floors, and I ruled out two things first.

A watchlist — not "buy now" but "if it comes down to here"

Let me be straightforward: for most of these, now is not the time. Grains and metals alike are sitting in the upper part of their yearly ranges.

So instead of "buy now", I write down the level to wait for. The trigger is the median of the last ten years of yearly lows — not one year's extreme, but the floor that has actually been touched repeatedly.

How to read it.

Grain micros cap the worst case somewhere between $1,800 and $4,700. Metal micros run $8,000 to $19,000, which is larger. Fund that amount and there is no liquidation, and you can wait months at a yearly-chart floor.

The closest right now are soybean meal (−18%) and natural gas (−19%). Gold, silver and copper are near all-time highs, so they are not candidates for this method today.

I have put the yearly charts of the watchlist markets here as well. Saying "it is high" or "it is low" in words gives you no way to check. Please look for yourself at where each one stands.

Agricultural and livestock markets work the same way

Corn, wheat, soybeans, oats, soybean meal and soybean oil, and livestock such as lean hogs and live cattle, all have floors.

For crops the floor is the cost of planting. When price falls below cost, less of that crop goes into the ground the following year. Less acreage means less supply, and price recovers.

Livestock is similar. When meat is worth less than the feed it takes, herds are reduced. It takes time, but supply falls in the end and price comes back.

Because that force exists, an absolute ruler (a fixed price) works here. It is the same logic as using 2.0 for natural gas.

But it needs time. Farming turns over once a year. Livestock is slower still. So you choose on yearly bars and you give it months, sometimes a full year. This is not a twenty-day method.

Things already measured in earlier pieces

This piece does not stand alone. Numbers measured in earlier pieces are used here directly. Rather than say the same thing twice, let me simply note where the joins are.

In an earlier piece I measured 42 markets against two axes. Natural gas came out at 0.16 against oil as a whole — energy in name only, not the same group. Against volatility it was −0.083 and against the dollar −0.062, both effectively coin tosses.

Which means natural gas is not a market that moves because something else moves it. With no larger axis to lean on, the phrase "cheap relative to something else" does not even parse. That is why this piece uses a fixed number, 2.0, for gas. The relative ruler does not work there, so an absolute one was the only option.

In that same measurement, the cohesion of the agriculture and livestock group was 0.226. Cohesion is the average of how alike the members of a group are, and compared with 0.82 for the index group that is very low. Even grains trade apart from one another.

So looking at a neighbouring grain will not answer your question. Once the yearly chart has given you a level, the next thing to look at is outside information — USDA reports, weather in the growing regions. The watchlist in this piece fixes the level; it does not tell you why price arrived there.

In an earlier piece I ran seven years of data with five different leverage settings. From 2× everything was liquidated. The only survivor was 1×.

When this piece says "fund the full contract value" and "approach it with micros", it is saying the same thing. Leverage of 1× is full funding. A different market, a different method, measured separately — and the same conclusion.

An earlier piece said to keep five metals on screen together: gold, silver, copper, platinum and palladium. Gold and silver sit at 0.73 while gold and copper are at 0.25, so you need all five to tell whether the move is a currency story or an economic one.

Four of the metals on this watchlist are four of those five. If you built that screen after the earlier piece, you are already set up. Just write the trigger prices beside them.

The same goes for Nasdaq. Earlier in this piece I used a low VIX as a caution signal for Nasdaq, and in the previous piece Nasdaq and the Dow were at 0.72 — the only meaningful difference inside the index group. So when VIX reaches its lower band, check whether the gap between Nasdaq and the Dow is widening too. That way you get a second piece of evidence rather than counting one signal twice.

There is a three-layer risk frame I set out earlier. The head is futures, the body is cash equities, the tail is options. And outside all of it sits zero risk — organising material, building lessons, building indicators and sharing them. Nothing to lose, and both you and the other person grow.

This piece is about the head, futures. But it is about handling the head as if it were the body.

What "futures are dangerous" really refers to is size. Buy a market that cannot go below zero and fund the whole contract value, and at that moment the position becomes the same as cash. No liquidation, a calculable maximum loss, and months of patience available to you. That is exactly what a micro does — it shrinks the head down to the size of the body.

And this piece itself is zero risk. I have written down what I measured, along with what would show me to be wrong. Whatever you decide to do with the method is yours to judge, but the way of measuring is yours to take and use.

The three earlier pieces were about one rule for every chart you open, where the seesaw tips when news breaks, and the difference between watching together and buying together. All three were about what to look at together.

This one comes next. Once you have decided what to look at together, the question becomes where to buy. That is why there are almost no correlation figures here, and a great deal about level and size instead.

To summarise

That is how to buy with the worst case already calculated. It is not a method for being right. It is a method for choosing a size that survives being wrong.

And this is a method for people who take the long view

Looking at the watchlist now, most entries are still −20% to −70% away. You may well think, "so there is nothing to do."

But look at the list again. Every one of those triggers is a level actually touched within the last ten years. They are not invented prices. Natural gas went below 2.0 in 2016, in 2020 and in 2024. Corn, wheat and soybeans each had their own such years.

Markets come back. They simply come when they come, not when you would like them to.

What do you need when that happens? Three things.

The value of this method is not in what you buy today. It is in whether you are ready when the day comes.

That is why you choose on yearly bars. A yearly chart prints one bar per year. You cannot become impatient with it. And when a level does arrive, it tends to stay for months, sometimes a year.

Only for the person who knew in advance and waited does that level become an opportunity. To everyone else it is just a frightening day when prices fell a long way.

11Can you trust these numbers — how they were measured

Plenty of pieces put probabilities in front of you; few say how they were counted. Please do not take these on trust — here is enough to check them yourself.

1) Overlapping days inflate the sample

If you compute "return 120 days later" every day, today and tomorrow overlap by 119 days. You are counting the same event over and over. The sample looks like a thousand cases when it may really be a handful.

So for natural gas I collapsed consecutive stretches and counted 12 events only. By calendar days it is 271; the real events number 12.

To be more honest still: those 12 are themselves clustered in 2016, 2020 and 2024. It is fairer to think of them as three or four episodes rather than twelve independent ones. Please read the results with that limit in mind.

2) Median, not average

VIX occasionally multiplies several times over. A single case like that lifts the whole average.

The question that matters in practice is "will this one attempt win?" So every conclusion here is stated as a median. As you saw earlier, by the average even 90 days is positive, while by the median it is negative from 30 days.

3) What would show this piece to be wrong

All three are things you can find out by measuring again. This is a matter of checking, not of believing.

Measurement conditions

12What this piece does not cover

Entry detail (which bar, which signal), scaling in, stop placement, taxes and commissions are all left out.

What is here is only three things.

The rest is yours to work out.

This is a record of the past, not a promise about the future. It is not a recommendation to trade. The judgement and its consequences belong to you.

Thank you for reading. If you take one sentence away from this piece, I hope it is this one.

Before you say something is cheap, decide what it is cheap compared to.