You traded for a year without tracking anything, and now you want the history. The good news is that it is largely recoverable — brokers keep the transaction record, and most of what you need can be reconstructed from it.

The bad news is what cannot be recovered, and knowing that in advance saves you from inventing it.

What the broker actually has

Every fill, with date, time, price, quantity and fees. Every assignment and exercise. Every share transaction. That is the complete factual record and it is enough to rebuild realised P&L exactly.

What it does not have is structure: which leg replaced which, which trades belonged to one campaign, why you did anything. Those were never in the broker's data model, so they were never stored.

Inferring roll chains

This is the interesting part of the job. A roll leaves a recognisable fingerprint in the transaction record: a buy-to-close and a sell-to-open on the same underlying, same option type, on the same day — usually within seconds, because they were one ticket.

That pattern is reliable enough to reconstruct most chains automatically. The signals, in descending order of confidence:

• Same timestamp to the second, same underlying, opposite actions — almost certainly one roll ticket.
• Same day, same underlying, same type, matching quantity — very likely a roll.
• Same day but different quantities — possibly a partial roll, and worth flagging rather than assuming.
• Different days — probably two separate decisions, not a roll.

A workable order of operations

1. Export everything. The full transaction history, not just trades — you need assignments, share movements and fees. See exporting trade history.

2. Normalise the format. Every broker exports differently. Map to one shape: date, action, underlying, type, strike, expiry, quantity, price, fees.

3. Reconstruct positions. Match opens to closes by contract identity — underlying, type, strike and expiry together identify a contract uniquely.

4. Infer chains. Apply the roll patterns above, flagging anything ambiguous.

5. Rebuild share lots. For each assignment, create a lot and apply the premium as a basis adjustment.

6. Reconcile the total. Your rebuilt realised P&L should match the broker's annual figure. If it does not, the gap is almost always fees or a missed assignment.

Two limits worth knowing before you start

Retention. Most brokers keep online history for a limited window — often two years, sometimes less for exports. Older records may require a support request. If you are reading this and have not exported anything, export now regardless of whether you plan to rebuild.

Corporate actions. Splits, mergers and special dividends adjust option contracts in ways the raw transaction record represents awkwardly. If your year included one, expect that ticker to need manual attention.

Is it worth doing?

For tax purposes the broker's 1099-B already handles the reporting, so a rebuild is not strictly required.

It is worth doing if you want to answer questions the 1099-B cannot: which delta band got assigned, whether rolling helped, what your wheels actually returned, how much the spread cost you. Those need structure, and structure is what a rebuild recovers.

A reasonable compromise is to rebuild the current year properly and leave older history as a simple realised total. The recent year is where the lessons are.

What can go wrong

Guessing at ambiguous rolls. A wrong link corrupts a chain permanently and invisibly.

Importing trades only. Without assignments and share movements the wheels cannot be reconstructed at all.

Skipping reconciliation. If the rebuilt total does not match the broker's, something is missing — and it is usually fees or an assignment.

Assuming history is always available. Export before you need it.