The Strategy
US$800 a month into a 2× Nasdaq-100 fund and another US$800 into a 2× S&P 500 fund, both under the same rules. Each fund is measured against its own record high. Sitting close to that high I invest only a fifth of the cash I'm holding for it and bank the rest. The further it falls, the bigger the share I spend, and at −60% I spend all of it. Every rule is written out below.
The intention
Most people do the opposite without meaning to. Confidence rises with the market, so we buy hardest near the top and go quiet after a crash, which is when future returns are best. Counter-cyclical leverage flips that instinct and makes it mechanical, so it doesn't depend on how brave I feel that week.
On the way up my leverage falls. I buy a small slice and let cash pile up next to it. On the way down it rises. The cash converts into shares, faster the deeper it goes. The cash isn't sitting idle. It's waiting for a trigger I already agreed to.
The concept is Henrique Centieiro's counter-cyclical leverage framework, published on TradingView. The drawing is mine; any clumsiness in it is mine too.
Optimal leverage
Leverage does not just scale a return up. Tony Cooper's 2010 paper works out the long-run compound return of a leveraged fund and finds it traces a hill. Returns climb as leverage rises, reach a peak, then fall away. Go far enough past the peak and the long-run return turns negative while the index underneath it is still going up.
The peak sits at roughly the index's return divided by the square of its volatility. Volatility is squared in that sum and the return is not, so volatility carries far more weight. Everything below follows from that.
The shape of Cooper's result, drawn from his return and volatility figures for the two indexes. The vertical scale is left unlabelled because the height of the hill depends on which decades you measure, so the position of the peak is the only part worth reading off it.
| Index | Annual return | Volatility | Peak leverage |
|---|---|---|---|
| S&P 500 1950 to 2009 |
7.0% | 15.3% | about 3× |
| Nasdaq-100 1971 to 2009 |
7.7% | 20.2% | about 2× |
| US market 1885 to 2009 |
3.9% | 16.7% | about 2× |
The Nasdaq-100 returned more than the S&P 500 across these windows and it still has the lower ceiling, because it was about a third more volatile and volatility is the term that gets squared. The livelier index is the one that tolerates less leverage. Across the ten markets Cooper tested, the optimum usually landed near 2, reached 3 in a few cases, and never reached 4.
Each extra turn of leverage adds return along with the volatility, so you are being paid for the risk you take on. Being early on the curve also buys you room to be wrong. If the real peak turns out to sit lower than the estimate, you give up part of the gain rather than all of it.
Each extra turn still adds the full volatility and now subtracts return, so you carry more risk to earn less. The error runs the other way too. A stretch of lower returns or higher volatility pulls the peak to the left, and a position that was slightly past it becomes badly past it, with nothing to announce that it happened.
Figures and formula from Tony Cooper, Alpha Generation and Risk Smoothing using Managed Volatility, Double-Digit Numerics, 2010. The curves above are my drawing of his result, not a reproduction of his charts.
The tier ladder
One number decides everything: how far the fund has fallen from its record high. That sets the percentage. Each fund is measured against its own high and runs its own ladder, so a crash in one does not touch the other's cash. The percentage applies to all the cash I have available, which is last month's reserve plus this month's US$800. The price that decides it is the price I actually pay on the day I place the order, not a month-end figure or an average. So the tier is fixed by the same number that appears in my log, and anyone can check the two agree.
Deploy 20% of all available cash
Four fifths of everything I hold stays in cash. This is the boring state and it might last years. Most of the work is not spending money.
Deploy 33% of all available cash
A third of the pile. On a 3× fund a −20% print means the Nasdaq-100 itself has barely moved.
Deploy 67% of all available cash
Two thirds, in one order. Whatever is left carries into next month and gets counted again.
Deploy 100% of all available cash
The entire reserve, in the month the headlines are worst. After that I'm fully invested and all I have left is next month's US$800. This is the tier I'll find hardest to do, and the one most likely to hurt.
Drawdown is measured on the fund itself, not on the index behind it. The fund falls roughly three times as hard, so these tiers are reached far more readily than the index would suggest.
Worked example
Starting from nothing, with US$800 going in on the same day every month, which is what I actually contribute. The rules take their slice, whatever is left waits for next month, and the market does what it does. This is where the reserve comes from.
| Month | The fund vs its high | Tier | Cash available | Deployed | Carried forward |
|---|---|---|---|---|---|
| Month 1 | −4% | Baseline 20% | $800 | $160 | $640 |
| Month 2 | −9% | Baseline 20% | $1,440 | $288 | $1,152 |
| Month 3 | −14% | Baseline 20% | $1,952 | $390 | $1,562 |
| Month 4 | −43% | Deep dip 67% | $2,362 | $1,583 | $779 |
The first three months look like nothing is happening. Each one buys a small parcel and leaves four fifths of the cash alone, so by month four there is $1,562 sitting idle.
Then the market falls, and month four puts $1,583 to work in a single buy, nearly ten times what month one managed, at prices 43% below the high. That money only exists because the three months before it refused to spend.
Two things this table leaves out, both of which the real log on the Progress page includes: the reserve earns 4% a year while it waits, and every buy costs $3 in brokerage and gets rounded down to whole shares.
Why not just buy every month
Buying the same amount every month, with no tiers at all, is the simpler version of this. It is a real strategy and for most people it is the better one. Holding a reserve only pays off if prices fall after I start, and nobody knows in advance whether they will.
Collins ran both against TQQQ over twenty years, changing nothing but the month you begin.
| When you started | Plain monthly buying | Holding cash for the falls |
|---|---|---|
| At a low Jan 2003 | $1.91m | $1.91m |
| At a high Oct 2007 | $335k | $389k |
Start at the bottom of 2003 and the two finish neck and neck. Nothing is gained by holding cash, because the market only went up from there. Start at the top of 2007, right before the crash, and holding cash finishes about sixteen percent ahead. Same fund, same contributions, same twenty years. The only difference was the starting point, and nobody gets to know theirs in advance.
I am beginning in August 2026 with the Nasdaq off its record high and the S&P close to one, which is neither a clean low nor a fresh peak. Holding the reserve is what I am paying to not have to guess which one it turns out to be.
There is a second reason, and it is the one I think actually keeps a person in the game. A plain monthly buyer has to endure a crash. Someone holding a reserve gets to use one. On a twenty-year horizon you end up hoping for the pullbacks, because that is when the shares go on sale. That is a completely different experience to watching the number fall with nothing to do about it, and I suspect it is the difference between still being here in year four and quietly giving up.
The discipline this actually requires
None of the numbers on this page happen without one boring behaviour: contributing on schedule, and deploying the reserve when the rules say to, in exactly the months it feels most insane to do so. The 80/20 and tier methods only beat plain buying because they spend into the fear. Skip the contributions during a crash, or freeze and hold the cash "until things calm down," and what is left is whatever I happened to buy before it started. Collins has a figure for that version: $1,000 put into TQQQ once, in January 2000, was still worth about $1,000 twenty years later. A simulated 3× Nasdaq fell around 99% in the crash that followed, and a position that stops being added to has no way back from a fall like that.
A 2026 paper out of the University of Waterloo gets to the same place from another direction. Forsyth, van Staden and Li test leveraged funds against a century of market data, rebuilt back to 1926 because the funds themselves only date from 2006 and their real record covers an unusually kind stretch of history. They find these funds reward a strategy that decides how much to hold and keeps changing it, and punish one that buys and then sits. Their other warning is about which index sits underneath: a leveraged fund on a broad market behaves nothing like one on a narrow, volatile sector, and most of the horror stories come from the second kind.
So the hard part was never picking the fund. It is being the kind of person who still sends money in during the month the headlines are worst. That is the real experiment here.
Figures from B.D. Collins, $1,000 to $1,000,000. His 80/20 method starts at 20% near a high and deploys more as the market falls. My numbered tiers come from Henrique Centieiro; they put exact percentages on that same idea. The Waterloo paper is Forsyth, van Staden and Li, Making Leveraged Exchange-Traded Funds Work for your Portfolio, March 2026.
Lifecycle investing
Ian Ayres and Barry Nalebuff, both at Yale, set the idea out in Lifecycle Investing in 2010, and their argument is about diversifying across time rather than only across assets. In your first years of investing you have very little money in the market and decades of future contributions that are not in it yet. By the end you have a large balance and almost nothing left to add. Measured across a whole working life, you are badly underexposed early and heavily exposed late, which is the opposite of what most people assume they are doing.
Their fix is to take more exposure early, when the amount at stake is small, and less later, when it is not. Using stock data back to 1871 they found that borrowing when young and holding more conservative investments when older beat both standard lifecycle funds and being fully in equities. Instead of shifting from shares into bonds as you age, you start leveraged and reduce the leverage.
They cap it at two to one, and say so explicitly: past that, on their numbers, the risk of being wiped out early outweighs the extra exposure. Two is where I have ended up, which was not the plan when I started reading them.
They write about borrowing, because that is what leverage meant when they wrote. Margin loans and deep in-the-money call options are the tools in the book. Three-times funds barely existed at the time: UPRO launched in 2009 and TQQQ in 2010, the year the book came out, with no record to test the idea against.
A leveraged ETF gets to a similar place without a loan in my name. The borrowing happens inside the fund, so there is no lender, no margin call, and I cannot lose more than I put in. What I can still do is watch it fall ninety percent and hold it, which is a different problem and not the one they were warning about.
I started at 2×, but I have never actually carried 2×. Holding cash back to average in and to buy the dips means the leverage I am really running is under two, and it moves with the market rather than sitting still. That is the level I am comfortable holding for the next twenty years.
The cost is that this will not keep up with TQQQ or UPRO in a good run. What I get for it is a drawdown I can keep contributing through. Those two get their turn once a market has already fallen a long way.
Ian Ayres and Barry Nalebuff, Lifecycle Investing (2010), and their paper Life-Cycle Investing and Leverage: Buying Stock on Margin Can Reduce Retirement Risk.
When it really falls
TQQQ and UPRO are not on the monthly schedule. They are what I start buying into once they are 40% below their record high, measured on the fund itself rather than on the index behind it, with money set aside for that and nothing else.
Three times a daily move compounds, so the index does not have to fall anything like as far. An index down about 16% puts a 3× fund past 40%. TQQQ went through that level in the 2020 crash and again in 2022, when it finished 82% below its high.
The money is separate. It does not come out of the US$800 that goes into QLD and SSO each month, and it does not come out of either reserve. Those two schedules carry on unchanged through a crash, which is the point of them. If that money is not there when the fall comes, there is no buying.
I buy in over months rather than all at once, and then hold.
Other ways to do this
None of these is a bad idea. They are just not what the monthly schedule runs on, and the reasons are below.
TQQQ is 3× the Nasdaq-100 that QLD tracks. UPRO is 3× the S&P 500 that SSO tracks. A higher multiple compounds harder through a bull run, and for a while that is what I intended to buy every month.
What changed my mind is the other side of it. Three times the daily move falls a great deal further when the index turns, and takes far longer to come back from it. QLD lost about 60% in 2022 and TQQQ lost 82%, on the same index in the same year. That is the difference between a drawdown I can keep buying through and one I might not.
So they are not off the table, they are on a different schedule. A fall that deep is the one thing that makes the extra multiple worth carrying, which is why I start buying them once they are 40% below their record high rather than every month. That rule is set out under when it really falls.
Collins' second system uses a 14-period moving average to move in and out of the fund, sitting in cash through downturns instead of riding them. It sidesteps the drawdowns that make TQQQ frightening, and on his numbers it works. It also asks me to read a signal every month and to be right about when to get back in. For now I would rather have rules that never ask me to make a call. That is a preference, not a verdict on the method, which on Collins' numbers works well.
A separate system, from Jason Kelly rather than Collins. It pairs TQQQ with a bond or cash reserve, starts around 60/40, and every quarter rebalances the TQQQ position back to a 9% growth target, selling the surplus in good quarters and buying the shortfall in bad ones. It shares my instinct exactly: react to price with fixed rules, never forecast. Selling is half the method, and every sale in a good quarter is a taxable event. Under the rules coming in from July 2027 that costs more than it used to, because the gain no longer gets halved and a 30% floor applies whatever I earn that year. That is a cost to weigh rather than a reason to rule it out. It is on my list, and I may run a separate portfolio on it in future.
Please read this part twice
A soft website does not make a leveraged strategy safe. I'm publishing the failure modes with the same detail as the rules, because a strategy page that only lists the upside isn't a strategy page. It's a sales pitch.
At −60% the rules spend the entire reserve. If the fall keeps going to −70% or −80%, and in 2022 QLD fell about 60% while TQQQ fell 82%, all I have left is the next US$800 a month. There is nothing below that.
QLD and SSO target twice their index's daily move. Over months and years in choppy markets, compounding drags the result well below 2× the index return, sometimes below zero while the index is flat.
An Australian buying a US-domiciled fund is also taking an unhedged AUD/USD bet, in an index where a handful of companies carry most of the weight.
The ladder is easy on a spreadsheet. Emptying the entire reserve into a market that has more than halved is a different thing on the day. Nothing here has been through that, so treat the deepest tiers as written rather than proven.
Read the risk section twice. The tier table can wait.
The mechanics
There is only one price in play each month: whatever the fund costs when I place the order. That single number sets the drawdown, the tier, the deployment percentage and the share count, and it is the number published in the log. If I ever get filled at something different from the price I based the decision on, both figures go on the record.
That is what my broking app charges to place the trade, not anything the fund charges. The tier decides the share count on its own, and brokerage comes out of the cash afterwards, so the fee never shrinks the buy, it just leaves a little less in the reserve. Over twenty years that is 480 monthly buys across the two funds, $1,440, which the site tracks and reports rather than hiding.
Multiply the pile by the tier percentage, then round the purchase down to full shares. Take a baseline month and a price of $90: 20% of an $800 contribution is $160, which buys one share at $90. Brokerage of $3 comes out of the cash as well, and the other $707 waits in the reserve.
I contribute US$800 a month to each fund, so the AUD it costs me changes with the exchange rate. I'm choosing a steady position size in the currency the fund actually trades in, and accepting a variable bill at home.
Comments
What would you do differently?
If you run leveraged ETFs, or a set of rules of your own, I'd like to hear how you settled on them and whether they have held up. Flaws in mine are worth hearing too, and better now than in year four.