Rolling is where most options logs quietly start lying. Not through carelessness — through a completely reasonable data model that happens to be wrong.

A roll is one decision that generates two transactions. Record them as two independent trades and the arithmetic breaks in a specific, consistent, flattering direction.

The specific mistake

When you roll, the broker's combined ticket shows a net credit — say $90. But two things happened inside it: you paid $810 to close the old leg, and you collected $900 opening the new one.

A log that records “new position opened, credit $900” and treats the closure as a separate line has now counted $900 of income that was 90% funded by a payment it may or may not have recorded against the right position. Add several rolls and the total premium collected climbs impressively while the account does not.

Model it as a chain

The fix is structural rather than arithmetic. A replacement leg is not an independent trade; it is the child of the leg it replaced.

Once that relationship exists, three things become computable that were not before:

• The buyback is charged once, to the leg that was closed. No double count in either direction.
• The chain has a single cumulative P&L. You can ask what the whole position earned, not what six unrelated tickets did.
• Duration is the chain's duration. Annualising the last leg alone hides the weeks the earlier legs tied up capital.

What to record on a roll

Judging the roll, not just recording it

Once the chain exists, you can ask the question that actually matters: was this roll worth doing?

The test from the rolling lesson is to judge the new leg as though it were a fresh trade. A $90 credit for thirty more days on $9,500 of capital is about 1.2% annualised — well below anything you would open deliberately. That is a delaying tactic wearing the appearance of a credit.

You cannot ask that question at all without recording the buyback against the position it belongs to.

Doing it in a spreadsheet

It is possible: add a “parent trade ID” column and a formula that sums credits less buybacks across every row sharing a root. In practice it is where most manual systems break down, because it turns a flat table into a recursive one.

A workable compromise for a manual log is to keep one row per chain rather than per ticket, updating the same row on each roll — cumulative credit, cumulative buybacks, current strike, current expiry, days held so far. You lose per-ticket detail and keep the number that matters.

What can go wrong

Recording only credits. The buybacks are real cash and the largest source of overstated performance in options logs.

Charging the buyback twice. The opposite error, from the same missing link.

Annualising the final leg alone. Ignores every week the earlier legs held capital.

Losing the thread after four rolls. Almost nobody can state a chain's true cumulative P&L from memory, and that is exactly when a wrong number is most likely to be believed.