The Averaging-Down Death Spiraleric

This is part twelve of the ASX order book series. The previous post was about sizing by setup grade. This one is about the most dangerous sizing decision of all: adding to a position that's going against you.

I read several trading journals end to end, written by traders with very different styles and very different results. Their worst losses had almost nothing in common in terms of setup, stock or market. But they had nearly everything in common in terms of behaviour. Each catastrophic loss followed the same sequence.

Anatomy of the spiral

The final loss is a large multiple of the original planned risk. In the journals I read, single losses from this sequence wiped out weeks or months of earlier gains in a single day or two.

Why it's so tempting: it usually works

The cruel part is that averaging down usually works. Most adverse moves do retrace at least partially. So a trader who averages down gets rewarded most of the time: the position comes back, they exit at breakeven or a small profit, and they learn that not cutting is fine.

Here's a synthetic simulation of a simple rule: on a losing trade, instead of stopping out at -1R, double the position and wait for breakeven, with a hard limit of -10R.

Copy import numpy as np

def averaging_policy(trials=100_000, seed=7):
    rng = np.random.default_rng(seed)
    results = []
    for _ in range(trials):
        # after the first -1R move, does price retrace to the new breakeven
        # before reaching the disaster level?
        if rng.random() < 0.85:
            results.append(0.0)    # bailed out at breakeven
        else:
            results.append(-10.0)  # forced out at the disaster level
    r = np.array(results)
    return r.mean(), (r == 0).mean(), (r == -10).mean()

print(averaging_policy())   # roughly (-1.5, 0.85, 0.15)

Under these made-up assumptions, the averaging trader "saves" the trade 85% of the time, and feels great about it. But the average outcome is about -1.5R per losing trade, versus -1R for the trader who just takes the stop. And the distribution is horrible: most of the time nothing happens, and occasionally a single trade costs ten times the planned risk.

This is the variable reward trap we'll look at in the psychology track: the behaviour is reinforced by frequent small rewards while the rare large punishment is dismissed as bad luck. Several journals explicitly noted they kept getting rewarded for bad discipline, and that this made it very hard to change.

Fading strength: the most common version

Across the journals, the single most common setting for the spiral was shorting a stock that was going up hard. The reasoning is always similar: the move is irrational, the valuation is absurd, it's up too much in a day. All of that might even be true. But a stock moving parabolically with heavy volume can go much further than seems reasonable, and short sellers being forced to cover add fuel.

The traders who survived this learned a simple rule, usually after paying for it more than once: trade with the trend and the catalyst, not against it, unless the book shows a clear, specific reason the move is ending. "It's up too much" isn't a reason.

Planned scaling is not averaging down

There's an important distinction. Scaling into a position can be perfectly sound if it's planned:

Adding to a winner as it confirms your thesis is a different behaviour again, and one several experienced traders used deliberately: start smaller, add as the trade works, keep the stop tight relative to the newest entry.

Rules that stop the spiral

These are the rules the traders in the journals ended up writing for themselves. Nearly every one of them reads like a scar.

The recovery speed lesson

One thing I found encouraging across the journals: the traders who survived didn't stop having big losses entirely. What changed was how early they cut, and how fast they recovered. Setbacks that once took a very long time to recover from became much shorter, because they were cut earlier and the trader was trading well enough to make the money back.

The aim isn't to never be wrong. It's to make sure that being wrong is always survivable.