
{/* banner prompt: Cinematic realistic photograph, 16:9 wide banner composition: a sailboat on open water adjusting its sail as the wind changes direction, dark clouds clearing on one side and sunlight breaking through on the other, choppy water, cinematic seascape. 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 is the fifth special topic in the [ASX order book series](/blog/2026/10/12-reading-the-asx-order-book-how-intraday-traders-actually-find-edge/), after [halts, placements and SPPs](/blog/2026/11/03-halts-placements-and-spps-corporate-action-mechanics-for-traders/). It's about a moment every trader in the journals I read went through, usually more than once: the setup that paid for months suddenly doesn't.

The question at that moment is always the same, and getting it wrong is expensive in both directions. **Is this variance, or has the market changed?** Abandon a good strategy during a normal drawdown and you lock in the loss. Keep trading a strategy whose edge has gone and you bleed indefinitely.

<KeyPoints>

- First test whether a losing stretch is inside normal variance for your statistics.
- Regime change usually has a cause: liquidity, volatility, catalysts, structure or crowding.
- Size down rather than stop, follow the volatility, and be ready to press when it returns.

</KeyPoints>

## Step one: is it within normal variance?

Most traders massively underestimate how long normal losing stretches last. The [win rate vs payoff](/blog/2026/10/21-win-rate-vs-payoff-two-shapes-of-a-profitable-trader/) post showed how often long losing streaks occur by pure chance. The same logic applies to drawdowns.

The practical test is to compare your recent results against what your historical statistics would produce by chance.

```python
import numpy as np

def drawdown_percentile(hist_r: np.ndarray, recent_r: np.ndarray, runs: int = 20_000, seed: int = 0):
    """How unusual is the recent stretch, given historical per-trade R results?
    Resamples historical trades to build the distribution of results over the same number of trades.
    """
    rng = np.random.default_rng(seed)
    n = len(recent_r)
    sims = rng.choice(hist_r, size=(runs, n), replace=True).sum(axis=1)
    return (sims <= recent_r.sum()).mean()

# a result around 0.2 is unremarkable; below 0.02 is genuinely unusual
```

If your recent stretch sits comfortably inside the historical distribution, it's probably variance. If it sits in the extreme tail, something may have changed. This doesn't prove anything on its own, but it stops you from reacting to noise.

## Step two: look for a cause

Variance has no cause. Regime change usually does. These were the causes that showed up in the journals:

<Table data={
    {
        headers: ['Cause', 'What you notice'],
        rows: [
            ['Lower market-wide liquidity', 'Thinner books, smaller auctions, setups that need size no longer have it'],
            ['Fewer catalysts', 'Quiet news flow, fewer stocks in play, more forced trades'],
            ['Lower volatility', 'Moves too small to clear costs, more scratches'],
            ['Choppy, directionless market', 'Breakouts fail, fades work, then neither works'],
            ['Market structure change', 'Auction timings or rules change, a platform or broker changes'],
            ['Crowding', 'Your setup becomes popular and the edge is competed away'],
            ['Sector rotation', 'The sector where your setups lived goes quiet while another heats up'],
        ]
    }
} />

Market-wide activity on the ASX varies a lot from year to year. In a year when listings shrink and daily value traded falls meaningfully, a setup that depends on liquid, active opens simply has less raw material to work with.

## Step three: what to do while you find out

The responses that worked were remarkably consistent across the journals.

<Steps>

1. **Size down, don't stop.** Keep trading the setup at reduced size so you keep collecting data and stay sharp, without paying full price for the uncertainty.
2. **Trade fewer, cleaner setups.** In a poor regime, only A+ setups deserve capital. Several traders went to deliberately small, quick scalps in choppy markets to avoid getting chopped up.
3. **Protect the account.** A defensive stretch with flat P&L is a success if the market isn't offering your setups.
4. **Review by sub-setup.** Often a strategy hasn't died, one variant of it has. Break your statistics down: by time of day, stock price band, sector, gap size, long versus short.
5. **Be patient.** Many regimes come back. Some of the best months in the journals came right after long quiet stretches, when volatility returned to a sector that suited the trader's style.

</Steps>

## Step four: adapting when it really has changed

When the evidence says the regime has genuinely shifted, the traders adapted in a few ways.

**Follow the volatility.** Intraday edges need movement. When one sector went quiet and another heated up (commodity cycles were a frequent driver), the traders who did best moved their attention to where the action was, while keeping the same core skills.

**Add a complementary style.** Several traders who started as pure scalpers added a slower style alongside: multi-day swings, catalyst trades held overnight, or a value-oriented portfolio. The scalping skills still mattered for entries, but the extra style kept the account growing during quiet intraday regimes. Some ended up primarily as swing traders.

**Adjust to structural change.** When market structure changes (auction arrangements, broker platforms, product rules), the core idea usually survives but the details need re-learning. A period of small size while you recalibrate is cheap insurance.

**Retire it.** Sometimes an edge really is gone. Accepting that is easier if you track statistics per setup and have other setups to fall back on.

## Press when it comes back

The flip side of regime awareness is pressing when conditions suit you. When a regime turns in favour of active traders, it tends to happen for many of them at once: record months cluster, because the market is finally serving up the volatility their styles need. The traders who'd survived the quiet stretch with their capital and discipline intact were the ones who could take full advantage.

That's the real reason to manage a bad regime carefully. Not just to lose less, but to still be there, sized correctly and thinking clearly, when the good regime returns.

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

<Disclaimer />


db509cf37e128f5df17983091e651ed3e4ac14986298c5471e523699286eb026

How to tell if your trading strategy has stopped working or is just in a drawdown, how to detect market regime change, and how day traders adapt.

Every trader in the journals I read went through stretches where a reliable setup suddenly stopped paying. Sometimes it was variance, sometimes the market had changed. How to tell the difference with your own statistics, what to do while you find out, and how traders adapted when the regime really had shifted.

When Your Setup Stops Working


{/* banner prompt: Cinematic realistic photograph, 16:9 wide banner composition: an antique brass balance scale on a dark wooden table, one pan holding many small silver coins and the other pan holding a few large gold coins, the scale perfectly level, soft directional window 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 post opens the second track of the [ASX order book series](/blog/2026/10/12-reading-the-asx-order-book-how-intraday-traders-actually-find-edge/): the business side of trading. The first track ended with [measuring order book imbalance in code](/blog/2026/10/20-measuring-order-book-imbalance-in-code/). Now we look at the numbers that decide whether any of that craft actually pays.

Every serious journal I read posted the same dashboard metrics: number of trades, win rate, average win, average loss, the ratio between them, commissions, net profit. And the most interesting thing in those dashboards wasn't any single number. It was the **shape** the numbers made together, and how differently the traders were shaped.

<KeyPoints>

- Expectancy needs both win rate and payoff. A high win rate alone means nothing.
- Profitable traders cluster into high win rate, low payoff harvesters and low win rate, high payoff snipers.
- Each shape has different losing streaks and different psychological traps.

</KeyPoints>

## The only formula that matters

Expectancy is the average amount you make per trade:

```text
expectancy = (win rate × average win) - (loss rate × average loss)
```

Two related numbers come up constantly:

```text
payoff ratio  = average win / average loss
profit factor = gross profit / gross loss = (win rate × average win) / (loss rate × average loss)
```

A profit factor above 1 means you make money before costs. And the breakeven win rate for any payoff is:

```text
breakeven win rate = 1 / (1 + payoff ratio)
```

<Table data={
    {
        headers: ['Payoff ratio', 'Breakeven win rate', 'Feels like'],
        rows: [
            ['0.5', '67%', 'You must win two out of three just to stand still'],
            ['1.0', '50%', 'A coin flip'],
            ['1.5', '40%', 'You can be wrong more often than right'],
            ['2.0', '33%', 'Lose two, win one, still break even'],
            ['3.0', '25%', 'Lose three, win one, still break even'],
        ]
    }
} />

That table explains most of the confusion beginners have. A 70% win rate sounds fantastic. With a payoff of 0.4 it loses money. A 35% win rate sounds terrible. With a payoff of 3 it's an excellent business.

Every trader whose journal I read learned some version of that lesson, usually by staring at a dashboard with a lovely win rate and a disappointing bottom line.

## The two shapes

Across the journals, profitable traders clustered into two very different shapes. Here they are with synthetic numbers.

<Table data={
    {
        headers: ['', 'The harvester', 'The sniper'],
        rows: [
            ['Typical style', 'Fast scalps, auction trades, many per day', 'Catalyst trades, few per week, often held longer'],
            ['Win rate', '75%', '42%'],
            ['Average win', '1.1 R', '3.2 R'],
            ['Average loss', '1.0 R', '1.0 R'],
            ['Payoff ratio', '1.1', '3.2'],
            ['Expectancy per trade', '+0.58 R', '+0.76 R'],
            ['Profit factor', '3.3', '2.3'],
            ['Trades per month', '150', '25'],
            ['Monthly expectancy', '+86 R', '+19 R'],
        ]
    }
} />

Both are profitable. Per trade, the sniper makes more. Per month, the harvester makes far more, because volume multiplies a small edge. But that volume comes with a catch that we'll see in the next post: every one of those 150 trades pays brokerage and spread.

## They feel completely different

The numbers above hide the experience of trading them.

**The harvester** is green most days. Losing streaks are short. The psychological challenge is boredom, precision and the occasional big loss that wipes out a week of small wins. Their biggest danger is one oversized, averaged-down loser.

**The sniper** is wrong more often than right. Losing streaks of six, eight, ten trades are normal. The psychological challenge is staying patient and sized correctly through long dry spells, and actually holding winners long enough to collect the big payoff. Their biggest danger is cutting winners early out of relief, which destroys the payoff that makes the whole thing work.

Here's a rough sense of how long losing streaks get, assuming independent trades.

<Table data={
    {
        headers: ['Win rate', 'Chance of at least one run of 6 losses in 100 trades', 'Chance of at least one run of 10 losses in 100 trades'],
        rows: [
            ['75%', 'about 2%', 'well under 1%'],
            ['50%', 'about 55%', 'about 4%'],
            ['42%', 'about 82%', 'about 16%'],
        ]
    }
} />

If you're built like a sniper and expect to feel like a harvester, you'll abandon a perfectly good strategy in the middle of a normal losing streak.

## You can simulate it

The fastest way to build intuition is to simulate your own numbers.

```python
import numpy as np

def simulate(win_rate, avg_win, avg_loss, trades=250, runs=10_000, seed=1):
    rng = np.random.default_rng(seed)
    wins = rng.random((runs, trades)) < win_rate
    pnl = np.where(wins, avg_win, -avg_loss)
    equity = pnl.cumsum(axis=1)
    drawdown = (np.maximum.accumulate(equity, axis=1) - equity).max(axis=1)
    return {
        "median_final_R": float(np.median(equity[:, -1])),
        "p5_final_R": float(np.percentile(equity[:, -1], 5)),
        "median_max_drawdown_R": float(np.median(drawdown)),
    }

print(simulate(0.75, 1.1, 1.0))
print(simulate(0.42, 3.2, 1.0))
```

Run that with your real win rate and payoff and you'll get a realistic picture of the drawdowns you should expect. Most people are surprised by how deep they are.

## Improving payoff without wrecking win rate

A common arc for harvesters goes like this: a very high win rate for a long time, but a payoff stuck around 1, so the edge per trade is tiny and costs eat a big share of it. The usual diagnosis, once they look honestly at the data, is that they aren't taking enough risk on their best setups.

What changed their numbers wasn't a new setup. It was management:

<Steps>

1. **Keep fast time stops on failures.** Losers stayed small and quick.
2. **Stop scaling out so eagerly on the best setups.** Hold more of the position when the book still supports the trade.
3. **Size up the A-grade setups** and size down or skip the marginal ones.

</Steps>

The win rate barely moved. The average win grew meaningfully. And because expectancy is win rate times average win, the whole business changed shape.

The sniper's equivalent problem runs the other way: improve win rate by being more selective, without cutting winners. Several journals found that a hard cap on the number of trades per day did exactly that. (That's a whole post later in the series.)

## What to track

If you keep a journal, track these at a minimum, overall and per setup:

<Table data={
    {
        headers: ['Metric', 'Why'],
        rows: [
            ['Win rate (scratches separated)', 'Scratches are neither wins nor losses. Lumping them in distorts both numbers.'],
            ['Average win and average loss', 'The two halves of payoff'],
            ['Payoff ratio', 'Tells you your breakeven win rate'],
            ['Profit factor', 'A single health number'],
            ['Expectancy per trade, after costs', 'The real edge'],
            ['Longest losing streak', 'Calibrates your psychology'],
            ['Largest single loss as a multiple of average loss', 'Flags the blow-up risk'],
        ]
    }
} />

That last metric matters more than it looks. In nearly every journal, the months that went badly weren't months where the averages were bad. They were months with one loss many times larger than the average.

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

<Disclaimer />


3ca4cfcda98caf772d7b7592d7d25d1b5b64dd8a7c780fc2115af9ef57dc1503

Win rate vs risk reward explained for day traders: expectancy, profit factor, breakeven win rate, losing streaks and the two shapes of profitable traders.

A 75% win rate can lose money and a 40% win rate can make a fortune. Expectancy, profit factor and breakeven win rates explained with synthetic examples, plus why the two profitable trader shapes feel completely different to trade and how to improve payoff without wrecking your win rate.

Tag: Trading Statistics

Tag: Trading Statistics

When Your Setup Stops Working

Win Rate vs Payoff: Two Shapes of a Profitable Trader