A Junior Quant's Guide to Ultra-Leveraged Trading
Bringing civility to the land of 100x leveraged products.
“There are only three ways a smart person can go broke: liquor, ladies and leverage.”
If you’re a quantitative trader or even just a general market participant, we all know the goal is to make money. However, some of us like markets so much that we earnestly try to make a business of it. At that point, however, it’s not just about making a profit — it’s about making enough profit — enough for employees, enough for better data, enough for better tools.
Now, a well-known joke in markets is that the best way to make $1m is to start with $2m — otherwise, your other option is to just chase low-probability, high-payoff scenarios.
Although we aren’t seen-it-all grizzled veterans, we’re not exactly fresh, naive tourists to pursue such a foolish approach.
So, this presents a problem that every prospective fund manager is eventually faced with:
“If we have a solid way of getting reasonable returns with low volatility, how can we source the highest amount of capital at the lowest cost?”
In traditional finance, we can get there a few ways:
Stocks: If you’re low-capital retail, you’re maxed out at 2x leverage for stocks under Regulation-T margin. Once you can put up a bit more capital, you’re then offered Portfolio Margin, which stretches things a bit further to 6-7x.
If you’re an institution (proprietary firm, hedge fund), you would go through what’s known as a prime broker (PB). The leveraged offered by a PB is about the same as portfolio margin but there are more bespoke services offered (e.g., short share locates).
Bonds: Even at a basic retail level, you can lever up 30x when purchasing treasury bonds. So, within a single day, a $30,000 Charles Schwab account can accumulate a $900,000 treasury bond position:
While attractive, no one actually uses this for trading since the margin rate offered by a broker (e.g., 12%) is generally a lot higher than the yield on any treasury (e.g., 4%).
Additionally, if you wanted to do something like buying a big bond position before a payment and only accruing 1-day of margin interest, you’d still be paying the “dirty price” of the bond which already factors in accrued interest.
Futures: Sure, futures definitely serve a real purpose for hedgers, insurers and whatnot, but really though — these are products basically designed for leveraged gambling.
If a trader decides to go with a discount futures broker, they may only need to post $500 of collateral for an S&P 500 futures contract (/ES). For reference, the notional value of an ES contract is 50 x index_price, so as of writing, that represents a notional value of ~$285k — effectively a 570x leverage rate.
Now, a hard lesson we learned from our 0-DTE days is that a product itself is not inherently an edge or a strategy — it may have very enticing features, but at the end of the day — it’s just an instrument. Whether or not it’s an effective instrument is entirely dependent on the specific underlying view you’re trying to capture.
So, now that we have a possible idea of how we can lever our capital up, we need to identify appropriate ways to use it.
We know better than to chase volatile, big payoffs, so our desired risk/return profile is bond-like volatility, with equity-like upside. The absolute percent returns might be small, but if we lever it up enough, the absolute dollar values should make it a worthwhile endeavor.
With that goal in mind, we set out to search for which kinds of strategies fit the bill and eventually, we landed on a pod-shop favorite:
Long-Short Relative Value
One of the first hedge fund strategies and ostensibly the most simple — you long one asset and short another — the offsetting PnLs roughly balance each other out for market neutrality.
You go from predicting the outright performance of any 1 asset to instead predicting the relative performance compared to another asset.
The predictions don’t even need to be something complex and proprietary, it can be common-sense stuff like estimating that large caps will fare better than small caps or that Bitcoin will hold its value better than a meme-coin.
Now, like most things — the nuance is where it gets tricky.
To see how, let’s take a look at a reasonable long-short trade:
Starting Theory: The S&P 500 Index will perform better than the Russell 2000 Index.
We’re stepping a bit out of our depth into the equity fundamental world, but it’s a concept that makes sense from a big-picture view — all theoretical academic factors aside; who will fare the best in the future U.S. economy? — the FAANG’s and the Berkshire Hathaway’s — or — the deep south egg-packaging firms and small, single-drug pharmaceutical companies?
Now, again, we’re not making an explicit directional view on “the market” itself, we’re just betting on the relative performance
If the market is positive, the S&P may go up 10% and R2K might only go up 3%, that 7% spread is our theoretical profit.
If the market is negative, the S&P may go down 10%, but R2K might be down 17%, that 7% spread is our theoretical profit.
Of course, this assumes that both assets remain correlated. If some rogue leader signs a new law that all S&P 500 members will be taxed at a 75% rate, but R2K members get a cut to 5% — the trade is over.
So, it’s not totally riskless, but realistically the correlation is likely to stick since both are driven by the same fundamental macro drivers. It’s a bit less quantitative, a bit less mathematically rigorous, but that can be all it takes to form the basis of a view.
With this being our starting point, we then run a very quick, raw test to see how this portfolio performed historically.
Just with a cursory test of the relationship, we see that even in times of distress (e.g., COVID, Russo-Ukrainian War, 2025 Trade War), the offsetting exposures balance out and the relationship generally holds.
There was a period in late 2020 to early 2021 where small caps outperformed — the consensus on why is that the small cap factor historically tends to outperform large caps immediately after recessions — Small Caps Have Been a Big Story After Recessions.
This gives us pause as, at the time of writing, there are some recessionary fears, so it’s a real risk factor to consider on a forward-looking basis — but we’re just in the experimental stage, so for now, let’s see what else we find.
Now that we have the view, we just need the right instruments. Thankfully, both the S&P 500 and the Russell 2000 have deeply liquid futures products, so all we need to do is use those instruments to systematically execute our view.
Here’s one way we might approach it:
At market open each trading day, we buy 1 /ES future and simultaneously short an equivalent-notional amount of /RTY futures.
We profit if the S&P 500 performs better than the Russell 2000 index over the course of that trading day.
At market close, we close the position.
Repeat.
Now, when doing a deeper simulation of a strategy, it is very important to be as realistic as possible. With that being a motto, here’s some of the few testing constraints:
Using NBBO (bid/ask) quote values instead of OHLCV values
For liquid products, OHLCV is honestly roughly fine enough, but as you get into more niche offerings, having to cross the bid/ask spread can be a night-and-day difference than just assuming you’d have been able to trade at the last price.
Using a real fee model
Assuming costless transactions is another major mistake, so it’s best to take the actual fees your broker charges (it’s always on their website) and simulate using the exact, real costs you’d incur.
So, with those constraints in mind, let’s see how the approach fared:
As demonstrated, the curve of the intraday version is about similar to that of just holding the long/short portfolio indefinitely.
Roughly similar, but it’s evident that the continuous rebalancing and daily accrual of transaction costs act as a draw on returns.
Further, going with an equal-dollar weighted amount introduces a new problem — beta mismatching.
Assuming a rolling 90-day beta of 1.35 for the Russell 2000, we assume that for every 1% the S&P 500 moves, The R2K will move by ~1.35%. Now, over time this is fine as the larger losses add up resulting in eventual R2K underperformance, but on an intraday basis, this means that we often take losses on days where the overall market goes up (e.g., S&P up 0.7% today, R2K up 0.95%), simply because R2K is correlated, but more volatile.
There are ways to beta-weight the portfolio so that the beta is theoretically net 0, but futures are only available in whole units, so we’re always going to be a little under or over exposed to beta.
Now, this can be handled by just expanding the duration of the trades, from intraday to weekly/monthly for instance. However, holding a futures position overnight, even when hedged, requires you to post the full maintenance margin of both products. Yes, even at full maintenance margin, you’re still levered up, but it’s significantly less and represents a legitimate capital constraint.
So, this was definitely an interesting starter experiment and got us into the weeds, but now that we’ve seen the problems, we can start to innovate for solutions.
To do that, we need to look somewhere we haven’t yet.
Crypto Perpetual Futures
We’re not crypto natives, but the inspiration for this idea came from one of the relative value scenarios we pitched earlier:
“The predictions don’t even need to be something complex and proprietary, it can be common-sense stuff like estimating that large caps will fare better than small caps or that Bitcoin will hold its value better than a meme-coin.”
If we’re using a similar logic and considering Bitcoin to be the S&P 500 of crypto — the most widely recognized, widely adopted — what would be the Russell 2000?
Well, it would be Ethereum, a currency that’s just as known, but only as the little brother to Bitcoin.
So, following the process set in the first example, we’ll start out with a cursory test to see what the relationship looked like historically. To execute this view, we would simply just long BTC and short ETH.
It turns out, this has been a pretty interesting trade for the past 5 years:
2021-2022 was a period known as the “DeFi Summer” — you remember, the NFT-COVID days — yeah, those days.
During that period, Ethereum was the darling of the crypto world and outperformed most other currencies, but since then, the relative performance has been rather linear, through negative and positive market regimes.
You may have even seen this meme floating around, which sort of crystallizes the sentiment regarding the relative performances:
So, with that baseline structure in tow, here’s where things get interesting.
Remember, we aren’t just looking for an optimal pair to long/short, we’re looking for an operational edge that will give us high enough amounts of leverage to make “worth it” amounts of money, but still gives us the flexibility to really hedge out our risks.
In comes the perpetual future.
As the name implies, perpetual futures are essentially just futures that never expire.
What makes these a suitable instrument are that they’re pretty linear — you basically say “I want $10,000 worth of BTC”, decide how much collateral you want to put up and you will get exactly that:
That simple, it’s just a raw way to trade an asset down to the exact penny amount and it’s available 24/7.
This is personally shocking to us because although it’s relatively new, it demonstrates that there’s a much deeper layer of sophistication than one would typically think of when just vaguely hearing the word “crypto”.
So, now that we have a view and an appropriate instrument, let’s build out a baseline strategy:
Each day, we initiate a short ETH, long BTC position to be held for the duration of U.S. Equity Market hours (9:30 EST to 4:00 EST)
Because Ethereum is highly correlated to, but more volatile than Bitcoin, we want to beta-weight the portfolio as opposed to a typical equal-dollar weighting
Example: Suppose the 90-day rolling beta of ETH is 1.36:
1 / beta → 1/ 1.36 → 0.75 → We want to short $75 of ETH for every $100 worth of BTC, this would make us beta-neutral.
When being beta-neutral, we can be more confident that we’re exclusively betting on the relative relationship of the assets and not having any exposure to whatever the general market does on that day.
At market close, we exit the position.
Repeat.
Let’s run a quick simulation to see how we would’ve fared in recent times:
As expected, the strategy performed pretty similarly to just holding the portfolio over the same horizon.
However, a crucial nuance is how the returns change based on the manner of execution. The exchanges prefer that you provide liquidity instead of taking it — this basically just means setting limit orders away from the current price and waiting for execution.
We compared the liquidity taker vs maker fees from DyDx, one of the largest exchanges in the space (no affiliation):
Now, a 0.05% fee seems relatively insignificant, but this is charged to your notional exposure not your collateral.
So, if you’re starting off with $1,000 bucks and deploy a gross notional of $20k across both pairs, you would need to pay ~$20 each trade if you wanted fast execution (instant -2%). Conversely, by executing in the way the exchange likes, you’d only pay $4 per trade.
Once we got a glimpse at some of the deeper nuances of the trade and still didn’t see any glaring red-flags, we set out to quickly go into prod just to learn more.
Trading POV + Risk Management
With these products, we are given more than enough leverage to bankrupt ourselves, so like any tool, it must be used wisely.
To get an idea of how much we wanted to risk, we need a way to model how much we might expect to make or lose on any given day. To do this, we run a very simple, raw metric:
For the PnL of each day, we calculate what % of notional value the move represented
A $50 absolute profit/loss on $20,000 of gross exposure represents a +/- 0.25% return.
Get the 95% confidence interval of the absolute daily returns:
(abs_avg_return) + (std_of_abs_returns*2)
This allows you to say; “in 95% of cases, we expect to make or lose no more than x% of our notional exposure on any given day.”
So, if your max expected risk was 0.5% of notional and you were starting with $1,000 but wanted to keep your daily PnL constrained to +/- $100 each day, you would choose to put up $20,000 for each trade (0.5% of 20k = 100).
A very basic measure, it does assume a normal distribution, but it is good enough for a rough system — the worst is always ahead so any risk model inherently underestimates risk, so realistically, this is fine.
Remember: Your PnL is still tied to your base collateral, so although you can lever up to make large amounts daily, you should size enough to be able to take your max expected loss at least 10x in a row. Boring, but at the end of the day, the goal is to never blow up.
So, with us having an idea of our daily expected variance, we then just set out to trading.
Pretty much everything went as expected PnL wise, but we encountered a major challenge:
Problem: The order execution constraints can be insufferably cruel.
We knew from the start that just to make this viable, we’d need to follow the exchange’s rules in order to be eligible for the maker fees of 0.01%.
Here’s a deeper look at what providing liquidity vs taking it looks like:
The top-of-book quote means the best bid/ask, so let’s say there’s a buyer at $100.00 and a seller at $101.00
If you submit a buy order at $101.00, you will be removing liquidity since there is now a lower quantity of the asset at that price to be traded with.
If you submit a buy bid of $100.50, you will be adding liquidity since there is now a higher quantity of the asset to be traded with.
Simple enough — just be more patient and deliberate with the orders you send.
Yeah, well, here’s what that’s like in reality:
Crypto prices move fast, so a competitive bid 10 seconds ago is now essentially just a stale order since the asset now costs 0.25% more.
You often need to cancel the order and send out a new one that’s closer to the current price, but not too close to the top-of-book.
In any other scenario, you can just wait longer for prices to revert, but you are making 2 simultaneous trades on different assets, so if one side takes 5 mins longer to execute than the other side, that introduces another problem.
You live with the mortal fear that as soon as your long BTC leg gets filled, there is a deep sell-off while you’re ultra-levered on one-way risk — a one way ticket to liquidation.
Much like with typical exchanges, if your collateral balance dips too far below the required amount, the crypto exchange will close the position for you, locking-in a loss.
You have to do this dancing gamut both ways — opening and closing — every day.
Now, there are indeed some solutions if one was truly committed to this specific trade:
More sophisticated, potentially automated order management strategies
Examples:
Only submit the long BTC order after the short ETH order has been filled
Simulate paying the taker fee on just one leg — take slower execution on the first order, then when it’s filled, immediately get filled on the other.
0-fee platforms (lighter xyz)
Hold the positions longer
While BTC and ETH are the most liquid pairs, there are financing interest costs that come with sustained exposure.
So, a key insight at this point is that even the most simple strategies require tons of specific nuance just for viability.
The profits here come not just from the pure asset returns, but from how strong your operational edge is (i.e., how much more can you save on fees?, how can you speed up the simultaneous execution?).
Final Thoughts
We didn’t want to give you an information overload, especially regarding some of the finer details on how perpetual futures work (e.g., funding rates as cost of capital), but this was just 1 example of the quantitative trading cycle:
Start with a strategy that works at a high-level
Relative Value
Define your ideal outcomes / mandate
Frequent realized PnL
Low volatility
High leverage
Find the optimal instrument(s)
Perpetual futures
Clearly define risk
Confidence-interval based approach
Learn the nuances in prod
Testing real-world execution
Another key takeaway from this is to see how the strategy was built around a clearly-defined market view — “Asset A should outperform Asset B for reason X and Y is the optimal instrument to execute that view”.
Many newer quants (all of us at one point) make the mistake of reversing that order. For instance, if you:
Created a dataframe of the returns of BTC compared to multiple other assets
Ranked by those which consistently underperformed
Went into prod; long BTC and short the ones in the bottom decile
You technically might even end up fine for awhile, but it always puts you in the situation where there’s really no reason for your selected portfolio to continue to do well, it just did in the past and will hopefully continue to do well.
However, when there’s a drawdown, you don’t know why it stopped working so that you can better react, you just lose money and go back to square one.
Don’t do that.
Start with a broad strategy class, form a clearly-explainable market view, find the optimal expression, lever up, and go home.
What you just read is part of a much bigger revival happening at The Quant’s Playbook.
We’ve gone quiet for a while — testing, refining, surviving — but now we’re back with brand new energy, new systems, and a renewed obsession with building the kind of content we always wished existed.
We’ll keep innovating in our traditional finance home turf, but we’re also diving deep into the more serious — and surprisingly profitable — corners of crypto you probably haven’t explored yet.
Thank you for sticking with us — truly. Your readership and support mean everything to us. Whether you’ve been here since day one or just recently stumbled across our work, the fact that you choose to spend your time reading what we create is something we hold in the highest regard. It’s your readership that pushes us to go all-in — working around the clock to bring you the kind of high-quality insights you won’t find publicly anywhere else.
Happy trading! 😄
Still needing to scratch that mental itch? If you enjoyed this post, you’ll probably love some of our others just like it: