Most Winners Arrive in the First Three Days

Yowster, over on the SteadyOptions forum, made an observation about the pre-earnings scatter charts that turned out to be worth a full afternoon of work:

Most cycles that hit the P&L target do so within the first 7 days, once you get beyond that point you wind up with significantly more losses than wins.

He is right — and the data is even more concentrated than his seven days suggest: the median winner arrives on day 3. But the reason he is right is not the obvious one, and the practical conclusion is the opposite of what you would expect.

What was measured

Every completed pre-earnings cycle in the backtest history: 46,717 of them, across 72 tickers, restricted to long straddles and long strangles. Entry at T-x business days before the report, exit rule as the scanner defines it — the first close at or above +10%, otherwise the last clean session before the announcement.

For each cycle I recorded the day on which the target was reached, counting from the day after entry.

The raw picture

Day 1 takes 22.8% of all winners, day 2 another 19.0%, day 3 another 14.7% — 56.5% of them in the first three sessions. Day 4 adds 11.3%, day 5 8.6%, day 6 6.3%, day 7 4.8%, bringing the cumulative total to 87.5%.

The median winner arrives on day 3. Nearly nine out of ten arrive within seven sessions.

Why that table proves nothing on its own

Here is the trap, and it is the reason this needed checking rather than believing.

A cycle that hits its target on day 3 is gone. It cannot hit again on day 8. So the pool of candidates shrinks every single day — 46,717 positions are alive on day 1, only 15,495 are still alive on day 7. Even if the daily chance of hitting were exactly constant, most hits would still land in the first few days, purely because that is where most of the positions are.

Any front-loaded histogram of this shape appears automatically. To know whether something real is happening, you have to ask a different question: among the positions still alive on day k, what fraction hits that day?

That is a hazard rate, and it is immune to the shrinking-pool effect.

The hazard rate says it is real

On day 1, 46,717 positions are alive and 4,745 of them hit — a rate of 10.2%. Day 2: 38,997 alive, 3,940 hits, 10.1%. Then the rate starts falling: 9.4% on day 3, 8.7% on day 4, 7.9% on day 5, 7.0% on day 6, 6.4% on day 7. From day 8 it flattens out — 6.1%, then 6.0% on day 10, still around 5% on day 12, with only 3,987 positions left by then.

If the effect were purely mechanical, this column would be flat. It is not. The daily chance of reaching the target falls by roughly 40% over the first eight sessions, then settles onto a plateau around 6% and stays there.

So a position that has not worked in its first week is not simply unlucky yet. It is in a genuinely worse state than it was on day one, and it stays there.

So should you cut it at day seven?

This is where the obvious conclusion breaks.

I compared two decisions for every position still alive and still un-hit at day k: close it at that day's price, or hold it under the normal rules until the target or the forced exit.

At day 1, closing returns −3.2% against −5.3% for holding: a 2.0 point advantage. At day 3 it is −7.4% against −8.8%, worth 1.4 points. At day 5, −11.2% against −12.2%, one point. At day 7, −14.9% against −15.3% — four tenths of a point. From day 10 the two are identical at −19.6%.

Exiting is always very slightly better than holding. And the advantage shrinks to nothing exactly where you would want to use it.

Cutting at day seven saves four tenths of a point. That is not a rule, it is a rounding error — and it certainly does not justify the extra decision, the extra commission, and the extra spread.

Look at the level rather than the difference: a position that has not hit by day seven is already at −14.9%. The damage is not caused by holding it too long. The damage is already done by the time you notice it isn't working.

What this actually tells you

The decision that matters is the entry.

Once a pre-earnings long-volatility position is open, there is very little left to manage. It either reaches its target in the first few sessions or it slowly stops being likely to. No exit discipline recovers the difference, because the difference was set when the position was bought — at what price, on which name, how many days out.

There is a related figure that makes the same point from another angle. The win rate climbs steadily with the length of the window: 14% for a position entered two sessions before the report, 44% at eight, 61% at fifteen. More sessions, more chances to catch the move. That, too, is a decision made at entry and nowhere else.

What I have not tested

Why the hazard rate decays is a separate question, and I have not answered it. Two candidates seem plausible.

The first is the shape of the volatility ramp. The position is not buying the earnings gap — it exits before the announcement — it is buying the rise in implied volatility that anticipates it. If a large part of that rise happens early in the window, then the easy gains are early by construction.

The second is gamma: as expiration approaches, the position's sensitivity to the underlying's actual movement changes, and so does what it takes to produce a 10% gain in value.

Both are testable. Neither is tested here, and I would rather say so than offer a mechanism I have not measured.

What I can say is that the pattern Yowster spotted survives the check that would have exposed it as an artifact — and that it points at the entry, not the exit.

OptionBench is a research and analysis tool, not an investment advisor. Nothing here is a recommendation to buy or sell any financial instrument. Backtested results are hypothetical and do not guarantee future performance.