TRIED & TRADED
← Field guide · what a strategy is

A default is not an instruction

  • Every number you do not choose gets chosen for you.
  • It then sits in your results looking exactly like a decision you made.

What happens

  • A strategy has more settings than anyone states.
  • Software fills the rest in — a width, a rounding rule, a fill assumption, an expiry choice.
  • The result comes out with those values baked in.
  • Nothing in the output marks which numbers were chosen and which were inherited.

Ours did exactly this

  • Our capture had a default spread width of $5.
  • Nobody chose it. It was simply what the code did when the width was not specified.
  • It ended up in more than a hundred thousand captured trades, every one of which looked like a deliberate $5-wide spread.
  • It was found and fixed on 6 August 2026, and it will be written up in full under What Went Wrong.

That count is from our own capture data, not from a strategy result. No test on this site has published a result yet.

Why it is worse than an ordinary mistake

A wrong number you chose is a mistake. A wrong number you inherited is invisible.

  • It is consistent, so nothing looks odd. Every trade has the same flaw and the results look tidy.
  • It survives review, because reviewing the strategy means checking the numbers that were written down.
  • It can be the thing producing your edge, and you would never know to test it.
  • And it does not travel. A $5 width behaves completely differently on a $30 stock and a $600 one, so a default that suited one name quietly ruined another.

What to do about it

  • List every setting the run touched, not just the ones in the strategy description.
  • For each one, say whether you chose it. The honest answer is often no.
  • Change the defaults and re-run. If the result moves, the default was part of the strategy.
  • Publish the defaults alongside the specification, so a reader can see what you did not decide.

The general form

  • This is not a options-software problem. It is what happens whenever a tool has more knobs than a description has words.
  • The fix is the same everywhere: make the implicit explicit, then test whether it mattered.

Common questions

How do I find the defaults in my own testing?

Change one and see whether the result moves. Anything that moves the result and is not in your written specification is a setting you inherited.

Is a default always wrong?

No — often it is perfectly sensible. The problem is not the value, it is that nobody weighed it.

Why does the width matter so much?

Because it sets the maximum loss. A $5-wide spread risks roughly the same money whatever the share price, which means it is a large position on a cheap stock and a trivial one on an expensive stock.

Did the $5 default invalidate your earlier work?

It changed what those trades were. They are being recaptured with the width stated rather than inherited, and the correction will be published with both numbers side by side.

How this shows up in our tests

No result is affected, because no result has been published.

The correction is recorded and the recapture is under way. When the first test publishes, its width will be a stated number in the specification rather than whatever the software would have picked.

The centre

What is a
specification?

Where width bites

The bull put
spread

The method

What is a
backtest?