HOW TO SHORT STOCKS LIKE MARTIN LUK.

Shorting a stock means borrowing shares, selling them, and buying them back later. If the price falls, you buy them back for less and keep the difference. If it rises, you pay more than you sold for.

For a swing trader, it's the mirror of buying a breakout. Except it isn't a clean mirror. Stocks fall faster than they rise, they bounce harder, and they can gap against you with no limit on how far.

The most common mistake is to short a stock because it's already fallen a long way. The approach in this post does the opposite. It waits for a stock in a downtrend to bounce back up into resistance that's still falling, and it only shorts once the bounce visibly fails on a lower timeframe. The process is Martin Luk's, from a study of his own short trades, and it's the core of everything below.

Why shorts behave differently

Three things make the short side its own game, rather than just the long side upside down.

Falls are fast, and so are bounces

Selling tends to come in bursts. Holders sitting on losses all sell at once when support breaks, margin calls force more selling, and a stock can give up weeks of gains in a few sessions. That speed is the attraction.

The catch is that a falling stock doesn't fall in a straight line. It drops, then rips back up as early shorts take profits and bargain hunters step in. Those bounces are often sharper than the pullbacks in an uptrend, and they're where most shorts get stopped out.

The upside is capped

A long can make many times its risk if a stock runs for months. A short can make at most 100%, and in practice much less, because falling stocks keep bouncing.

Martin makes the point directly: shorts are mathematically less favourable than longs, so the entry has to be more precise to make up for it. That's why everything in his process is about getting in close to a level that tells you quickly when you're wrong.

Crowding and the squeeze

Short sellers have to buy back eventually. When a lot of them pile into the same falling stock, any good news forces them to buy at once. Their buying drives the price higher, which forces more of them to buy.

That's a short squeeze. It's why a downtrend can produce a double-digit up day out of nowhere, and why the most crowded shorts are often the most dangerous ones to join late.

The setup: a bounce into falling resistance

Martin's main short is simple to describe.

A stock is in a downtrend, below its moving averages. It rallies back up into those falling averages (mostly the 21- and 50-day exponential moving averages), gets rejected there, and drops straight away, often without building any base first. He shows it on charts ranging from LRCX in 1990 and ADBE in 1992 to QS in 2021 and TSLA in 2025, to make the point that the pattern repeats across decades.

Two refinements came out of his study.

The tighter the EMAs, the better. When the 21 and 50 are bunched close together, the rejection works more often. A bounce into a tight cluster meets all the overhead supply in one narrow band, while a bounce between two widely spaced averages has room to keep going.

Anchored VWAP is a second form of resistance. An anchored VWAP (AVWAP) is the volume-weighted average price of every share traded since a chosen day, which here is a swing high. When it's falling, most buyers since that high are underwater, so a rally back to it gives them a chance to get out at break-even. Martin looks for a falling AVWAP sitting near tightening EMAs, and his examples include TEM, GEV, MSTR and TSLA.

He also restricts the universe to fast-moving technology names he already trades, not utilities or consumer staples. And he keeps two signals apart. The higher-timeframe picture (a weak stock bouncing into resistance) only puts a name on the watchlist. It isn't an entry.

This is a different trade from the parabolic short. A parabolic short fades a stock that has gone nearly vertical, up 50–100% or more in days, and bets on a snap back to the 10- and 20-day averages. The downtrend short sells a stock that's already weak, on a rally into averages that are falling.

So one is a mean-reversion trade against a blow-off, and the other is a trend trade, with the trend. The entries and stops look alike, but the charts you're hunting for are opposites.

The triggers: let the lower timeframe say no

The first thing Martin tried was flipping his long entries, by shorting a break of the opening-range low or the prior day's low. It didn't hold up, even in hindsight. It only worked on days when a stock gapped up and reversed at once.

What fixed it was a well-known principle: every reversal on a higher timeframe starts as a breakdown on a lower one. He combined that with Qullamaggie's intraday VWAP entries for parabolic shorts, and settled on two triggers. He only takes either one while the stock is at higher-timeframe resistance:

• Trigger A, the hourly break. Short when price breaks below the low of the previous hourly bar. The bounce has stalled at the daily resistance, and the first lower low on the 60-minute chart is the sign it's turning.
• Trigger B, the VWAP failure. Let the stock bounce intraday into the day's VWAP, or into an AVWAP anchored to the high of the day, and short when it fails there on a 5-minute chart.

He's blunt about the limits. Shorting every bounce into falling EMAs doesn't make money on its own. The details of execution decide the result, and the examples further down show how much.

The stop: above the high of day

The stop goes above the high of the day, or above the high of the candle that broke down. On a VWAP-failure entry, a move back above VWAP can serve as the stop.

Martin's typical short stop sits 1.5–5% away. Because the trigger fires right next to the high, the stop is naturally close. If the high of day is far away by the time the trigger comes, your entry is late.

But a stop is an order, not a guarantee. A short held overnight has no protection from a gap up on news, and with a tight stop, a gap through it can cost several times the planned risk. (Read how stop distance sets position size for why a tight stop also means a larger share count, which makes a gap through it more expensive.)

Sizing: 0.5% and a position cap

Martin risks 0.5% of the account per short, and caps any single position at 35% of the account. He goes smaller still in higher-ADR names, when he's testing the waters, or when trades aren't working. Here's the arithmetic on a $100,000 account:

The cap is what stops a very tight stop from turning into a huge position. So there are two separate limits, risk per trade and position size, and whichever one binds first wins. (For the method in general, see position sizing for swing trading.)

Covering: quick swings, taken in pieces

Martin treats shorts as quick swings, not trends to trail for months. Here are his covering rules:

• Cover about 15% once the trade is more than 3R in profit (three times the initial risk).
• Cover about another 15% into likely support, like prior swing lows, moving averages and anchored VWAPs.
• Cover the rest if the move looks stretched to the downside, or if an hourly bar closes back above the 50-period EMA.

He describes these as his long sell rules turned upside down. The hourly 50 EMA is the key one. It's close enough to get you out before a bounce takes back most of the gain, and loose enough to let a real breakdown run for a week or two.

When not to short

Most of the time, the best short is no short.

In a strong market. Martin turned to shorts in late December 2024, because breakouts were failing and leaders were losing key averages after a long, euphoric run. That's the environment the setup is for. When the indexes are above rising averages, weak stocks get bought too.

Whether reversal shorts work better when QQQ and IWM are below their 9, 21 or 50 EMA is the question Martin named as his next study. (For index filters on the long side, see market filters with moving averages.)

When you're new. Qullamaggie's advice, per community notes of his streams, was not to short until you're already very successful on the long side. A beginner's worst case on a long is known. On a short, it isn't.

To fix a slow patch. Martin made only about 8% in the 2022 bear market, because nearly every short he tried failed, and he names overtrading in corrections as one of his own weaknesses. Shorting because you're bored, with no longs to buy, is how that happens.

The risks a long never carries

• Unlimited theoretical loss. A long can lose at most what you paid. A short has no ceiling: if the stock doubles, you lose twice what you sold. The stop is the only thing between you and that tail, and it doesn't work through a gap.
• Short squeezes. Heavily shorted stocks can jump hard on any good news as shorts rush to cover. And the squeeze often comes after the stock has already fallen, just as you start to feel safe.
• Overnight gaps. Earnings, upgrades, takeover rumours and index news all land outside market hours. A short held overnight can open far above its stop, and you get filled at the open.
• Borrow availability and fees. You can only short shares your broker can borrow. Popular shorts become hard to borrow, carry a daily borrow fee that can be large, and can be recalled at short notice. So check the borrow status and rate before you plan the trade, not after.
• Dividends. If you're short over the ex-dividend date, you owe the dividend to the lender.

Some traders avoid the mechanics altogether. They don't short stocks directly, and express bearish views through inverse ETFs instead, so every trade is a long with a known worst case.

A real short: COIN in May 2026

Coinbase (COIN) shows the whole process on one trade. When the 21- and 50-day EMAs start converging and both slope down, Martin doesn't short the low. He waits for the stock to bounce back up into them, because that's where trapped buyers sell and the downtrend tends to pick back up.

COIN rallied right into both averages and stalled there. The entry came when the hourly chart made its first lower low (trigger A), and the stop went just above the high of the day.

From there the covering rules did the work: a piece at 3R, a piece at the prior low, and the rest on the first hourly close back above the 50-hour EMA. The trade came to +7.5R.

It didn't catch the whole move, and that's by design. You get paid on the way down and keep one firm line for the rest.

A tight cluster again: NET in January 2026

Cloudflare (NET) is the cleanest version of Martin's main setup. Through December 2025 its 21- and 50-day EMAs slid lower side by side, only about 2% apart, so any rally back up would meet both of them at once.

In early January it bounced right up into that band and stalled just under the 50-day. The first hourly lower low got it in, the stop went just above the high of the day, and both partial covers came on the next day's drop. The rest rode the slide until an hourly close back above the 50-hour EMA, and the trade came to +9.7R.

A bounce into the 21-day EMA: ARM in December 2025

The cluster doesn't have to be tight for the first test to hold. Arm Holdings (ARM) fell hard through November 2025, and by early December both EMAs were pointing down, with the 50-day still a few percent above the 21-day.

The bounce stopped at the 21-day EMA, well short of the 50. The hourly trigger got it in, the stop went above the high of the day, and it took a few days to get going. Once it did, it slid for most of the month: a piece at 3R, a piece at the November low, and the rest on the first hourly close above the 50-hour EMA, for +7.4R.

An anchored-VWAP short: MSTR in January 2026

Strategy (MSTR) in January 2026 shows what to do when the EMAs aren't tight. If the 50-day is too far above the price to matter, Martin looks at the 21-day EMA and an AVWAP anchored to the last swing high instead.

MSTR bounced into the band between those two lines and failed there. The same hourly trigger got it in, with the stop above the high of the day, and the trade came to +8.5R.

Into both averages at once: TWST in August 2024

Twist Bioscience (TWST) broke down hard in early August 2024 and spent the rest of the month drifting under its averages. By late August both EMAs were falling and only about 3% apart, so a rally back up had a narrow band of resistance waiting for it.

At the end of the month it popped up through the 21-day and stalled right under the 50. The hourly trigger got it in, the stop went above the high of the day, and the first partial cover came the next morning. The second came at the early-August low a week later, and the rest came off on the first hourly close above the 50-hour EMA, for +8.3R.

A stall at the 21-day EMA: SNPS in March 2025

Synopsys (SNPS) slid through late February and March 2025, and both of its EMAs turned down with it. Like ARM, this is the version where only the 21-day comes into play: the 50-day was too far above to reach, so the 21-day did the work.

In late March it bounced into the 21-day and went no further. The first hourly lower low got it in, with the stop above the high of the day. A piece came off at 3R two days later, another at the mid-March low, and the rest on the first hourly close above the 50-hour EMA, for +6.6R.

A tight pair rolling over: CAI in January 2026

Caris Life Sciences (CAI) spent December 2025 chopping around its averages, then rolled over in mid-January. By late January the 21- and 50-day EMAs were both sloping down and only a little over 2% apart, the kind of tight cluster Martin likes.

The bounce reached the 21-day and stalled just under the 50. The hourly trigger got it in, the stop went above the high of the day, and two days later it fell through both partial cover levels at once. The rest rode the slide until the first hourly close above the 50-hour EMA, for +6.8R.

Building your own short playbook

The most useful part of Martin's write-up isn't a rule. It's how he found them.

He didn't start from a scan. He started from his own journal, with the list of every ticker he'd traded since 2020. For each stock, he went back to its IPO, looked at every long downtrend, and asked where the best risk-reward short would have been.

• About 5 charts were enough to see the main idea: bounces into falling EMAs that get rejected.
• More than 200 charts were needed to refine the execution: which trigger, where the stop goes, when to cover.

The point is that the process was low-friction. Martin stopped archiving charts in a place he never went back to, and kept the study tied to names he already knew. And he kept asking specific questions (which EMA, which timeframe), because vague rules slow you down when the trade is live.

You can run the same loop on your own trades. Pull every short you've taken. For each one, mark where the bounce stalled, which average or AVWAP it was under, how far the high of day was from your entry, and whether you held it overnight.

Then compare your shorts with your longs on win rate, average R and profit factor. If your shorts lose money, check two places first: wide EMA gaps (Martin's tight-cluster point), and overnight gaps through tight stops. Our guide to reviewing your trades covers the routine, and MFE and MAE show how far each short went against you before it worked.