
{/* banner prompt: Cinematic realistic photograph, 16:9 wide banner composition: a server room aisle with rows of blinking racks on one side and, in the foreground, a wooden chessboard with a single knight piece in sharp focus, cool blue server light meeting warm lamp light. Muted teal and warm amber colour grade, soft natural light, shallow depth of field, high detail, photorealistic, no readable text, no numbers, no logos, no visible faces. */}

This continues the final part of the [ASX order book series](/blog/2026/10/12-reading-the-asx-order-book-how-intraday-traders-actually-find-edge/), after [the liquidity harvester and the catalyst sniper](/blog/2026/11/08-the-liquidity-harvester-and-the-catalyst-sniper/).

When I talked through these trader profiles with a colleague, he summed up the harvester types in one phrase: human high-frequency trading machines. It's a good line, and it's mostly right. But the places where it's wrong turned out to be the most interesting part.

<KeyPoints>

- The best scalpers resemble HFT systems in holding period, trade count and reliance on the book.
- They cannot compete on speed. They compete on judgement at auctions and catalysts.
- That judgement is codifiable, which makes it interesting for systematic traders.

</KeyPoints>

## Where the comparison holds

The best discretionary scalpers whose journals I read share a surprising number of traits with high-frequency trading (HFT) systems:

<Table data={
    {
        headers: ['Trait', 'HFT system', 'Human scalper'],
        rows: [
            ['Holding period', 'Microseconds to seconds', 'Seconds to minutes'],
            ['Trade count', 'Enormous', 'Thousands per year'],
            ['Edge per trade', 'Tiny', 'Small, a few ticks'],
            ['Main input', 'Order book and trade feed', 'Order book and tape'],
            ['Main enemy', 'Costs and latency', 'Costs and slow execution'],
            ['Exit discipline', 'Instant, rule-based', 'Time stops, liquidity exits'],
            ['Favourite moments', 'Auctions, imbalances, mechanical flows', 'Auctions, imbalances, mechanical flows'],
        ]
    }
} />

Both are harvesting small, short-lived mispricings in the microstructure of the market, at volume. Both treat the order book as the primary source of information. Both obsess over costs. Both exit the moment the expected move fails to appear.

## Where it breaks down

### Speed

An HFT system reacts in microseconds. A human reacts in hundreds of milliseconds at best, more realistically seconds. Any edge that's about being first, such as racing to hit a stale quote, belongs to the machines. Humans simply can't play that game.

### Taking versus providing liquidity

Many HFT strategies make money by **providing** liquidity: quoting both sides, earning the spread, managing inventory. The human scalpers I read about mostly **took** liquidity: paying the offer or hitting the bid to get in when a setup appeared, and leaning on somebody else's resting order as protection. They pay the spread rather than earn it. That's a fundamentally different business.

### Context

This is the big one. An HFT market-making system typically cares about very local information: the book, recent trades, maybe related instruments. A discretionary scalper's decision at the open folds in a wide range of context:

- What did the commodity, the futures and the stock's foreign listing do overnight?
- Is this gap caused by news, or by one large seller?
- Is today a rebalance, expiry or settlement day?
- What happened the last time this stock opened like this?
- Is the sector in favour, or rotating out?
- Is the market in a regime where fades work, or where breakouts work?

None of that is about speed. All of it is about **judgement at a specific moment**.

### Selectivity

An HFT system is always on. It quotes all day, every day, in every instrument it covers. The human scalpers' best months were the ones with the fewest trades. They made their money by declining most opportunities and pressing a few.

## So where does the human edge actually live?

Putting those differences together, the remaining human edge seems to sit in a narrow but real space:

<Steps>

1. **Event moments** where the price is set by an auction or a burst of forced flow, not continuous competition: the open, the close, halts, rebalances, settlements.
2. **Interpretation**: judging whether news is material, whether a gap is mechanical, whether a move is proportionate.
3. **Cross-domain context** that's awkward to encode: commodity cycles, corporate actions, the history of a particular stock's register.
4. **Selectivity and size**: knowing when *not* to trade, and when one read deserves much more size than usual.

</Steps>

In other words, discretionary scalpers aren't competing with HFT. They're operating in the gaps HFT leaves: moments that are too infrequent, too contextual, or too lumpy for a speed-optimised system to bother with.

## The edge is shrinking at the margins

The journals also showed signs of the machines encroaching. Mispriced moves that once lasted minutes being corrected in seconds as market-making algorithms stepped in. Illiquid opens snapping back to fair value faster than before. Changes to auction mechanics altering setups that had worked for years.

The human edge isn't disappearing, but it's being pushed toward the moments that are hardest to automate. The traders who did well adapted by leaning further into context and selectivity, and by adding slower styles where speed matters even less.

## What this means for system builders

From a systematic trading point of view, the interesting conclusion is that the human scalper's edge is largely **codifiable**, just not as a speed game. The heuristics from this series can be expressed as features:

- auction mispricing versus overnight leads and sector peers;
- protection relative to expected volume;
- resistance ahead;
- persistence of large orders;
- gap cause classification from announcements and order composition;
- time stops calibrated from historical MFE.

I sketched how to compute several of these in [measuring order book imbalance in code](/blog/2026/10/20-measuring-order-book-imbalance-in-code/). The point isn't to out-race anyone. It's to encode the *judgement* and apply it consistently at the moments that matter, without the tilt, fatigue and boredom that the journals show are the human scalper's real enemies.

## What this means for human traders

If you're a discretionary trader, the lesson is to stop trying to be fast and start trying to be **right at the right moments**:

- Don't compete in continuous, liquid trading where you're slower than everyone.
- Focus on auctions, catalysts and mechanical events where context matters more than milliseconds.
- Build your context advantage: sector knowledge, the corporate calendar, the history of the stocks you trade.
- Let the machines provide the liquidity you lean on, and pick your spots.

The phrase "human HFT" captures the intensity and the precision of the best scalpers. But the reason they survive alongside the real HFTs is that, at the moments they choose, they're doing something the machines don't.

<SeriesNav series="asx-order-book" />

<Disclaimer />


63eb12b7c46dd8325556bccd52a7769102782fdf753b9889a0ff9673b3a301db

How discretionary ASX scalpers compare with high frequency trading algorithms, where human judgement still beats speed, and what systematic traders can learn from them.

The best discretionary scalpers on the ASX look a lot like high-frequency trading machines: thousands of trades, holding periods in seconds, edge read straight from the order book. But they can't compete on speed. What they actually compete on is judgement at specific moments, and that tells both human traders and system builders where the remaining edge lives.

Tag: Market Microstructure

Tag: Market Microstructure

The Human HFT: What Discretionary Scalpers Do That Algorithms Can't