Your statements and your records

Two documents, one of them a record 

Two different things get called the account statement, and the difference matters more than it sounds. The first is the account history inside the trading terminal: a live view, generated on demand out of the account's present state, listing open positions, closed positions and cash movements. The platforms YAL runs, MetaTrader 5, both produce one. The second is the periodic statement issued by the firm, covering a stated period, closed off at a stated date, and issued as the firm's own account of that window. Both are drawn from the same underlying ledger and will usually agree. Only the second has an issuer, a period and a date on the face of it, and that is what makes it the thing meant when a record is in question.

A statement covers a period, not a position. It opens with a balance carried in, itemises every transaction and every adjustment that touched the account inside the window, and closes with a balance carried out. The arithmetic of the whole document is therefore a closed loop: the opening balance, plus and minus everything listed, equals the closing balance. When that loop does not close, either something is missing from the list or one of the figures is wrong, and confirming that it closes is the cheapest check available on the entire document.

The lines one closed position occupies 

Inside the window, a closed position generally takes a row of its own or a pair of rows, and its fields fall into four families.

• Identification. A ticket or deal identifier, the instrument, the direction and the size. The identifier is the handle the whole record hangs on, because a query about a single fill is asked in terms of it and answered in terms of it.
• Execution. The opening price with its timestamp and the closing price with its timestamp. These record what was done, which is a separate matter from what was requested.
• Charges. Commission, applied per side or on notional depending on the pricing arrangement in force, and financing, applied once for each daily rollover the position was held through. On a shares CFD held across an ex dividend date, a dividend adjustment lands here too.
• Result. The gross figure produced by the price difference multiplied by the size, and the net figure once the charges above have been applied to it.

A published typical spread is not a term of any individual fill either. It is a statistic describing many fills over a period, while a row on a statement records exactly one. At YAL the typical EUR/USD spread on the Raw Spread account is 0.1 pips, with a commission of $3.50 per lot, per side. Reconciling a single closed position is arithmetic on that position's own prices and charges, and it is not a test of whether a published average held on the day.

Reconciling one closed position 

Reconciliation is the act of recomputing a figure from its own inputs, independently of the document that states it, and then accounting for every unit of difference between the two. It is ordinary bookkeeping and it is not an accusation. Applied to a closed position, it means taking the instrument's contract convention, the size, the two prices and the charging basis in force, working out what the round turn should have cost, and reading that against the charges the statement actually applied.

Two features of that block are the point of it. The first is that the computed charges and the printed charges agree, which is the ordinary case, and it is exactly that ordinariness which makes a disagreement legible on the day one occurs. The second is that the largest single line in the total appears nowhere on the document, because it was collected inside the opening price. A reconciliation confined to the printed charges reconciles the smaller half of what the round turn cost.

The second calculation is the one that explains what the balance moved by. Gross and net differ by exactly the charges, and the charges are applied in the same direction whichever way the price went.

The asymmetry in the last two rows is arithmetic and not a claim about anything. A charge is a debit in both cases, so it reduces a credit and enlarges a debit by an identical amount. That is the whole of the relationship between the two figures a statement prints, and it is why gross and net are never interchangeable when a row is being checked.

Where a difference usually comes from 

When a computed figure and a printed figure disagree, the ordinary explanations are a long way from exhausted before an error is reached. The common ones, in the order they tend to turn up:

1. The fill was not at the price requested. An order executes at the price available when it reaches the market, which in a fast or thin market is not the price that was on screen when it was sent.
2. The two prices were read from the same side of the quote. A position opened on one side of a two sided quote is closed against the other, so comparing two prices taken from one side produces a difference of roughly the spread.
3. Financing was applied for a rollover that was not counted. The charge attaches to each daily cut off a position is held through, and the number of cut offs a holding period spans is not always the number of nights that intuitively passed.
4. A conversion rate applied. Where an instrument settles in a currency other than the account's, the figure was converted at a rate struck at the time, not at a rate visible now.
5. An adjustment landed against the instrument rather than the position. A dividend adjustment on a shares CFD, or a corporate action, moves a figure without any transaction of the holder's having taken place.

Only once those are ruled out does a difference become a question for the firm. Framed that way, reconciliation is not mainly a search for wrongdoing, and treating it as one produces a great deal of alarm about ordinary mechanics. Its real product is a baseline. An account whose rows have been checked has an established normal, and an abnormal row on it is conspicuous. An account that has never been checked has nothing to be conspicuous against.

What the record is for 

The practical value of a retained statement surfaces at the moment a question is asked about one execution. A query that names a ticket identifier, an instrument, a timestamp and a printed price is a question about a specific event in the firm's own logs, and it can be answered out of them. A query that describes an impression of what happened is not answerable at all, however strongly it is held. The difference between the two is entirely a matter of what was written down at the time, by whom, and whether it survived.

The chain of documents that lets one figure be traced back to the event that produced it is conventionally called an audit trail. On a trading account it runs from the order as it was sent, through the confirmation of the fill, to the statement line reporting it, to the balance that line moved. A missing link is a step in the account of events that has to be taken on trust by whoever reads it, and trust is precisely what a record exists to make unnecessary.

How a query becomes a formal complaint, and what a regulated firm is obliged to do once it becomes one, is the subject of the next lesson and is not covered here. What belongs to this one is the plainer point underneath it: a complaint is made out of records, and the records have to exist before the thing complained about happens.

What is worth keeping 

The material worth retaining is broader than the statements, and most of it is issued once and never issued again.

• Every periodic statement in the form it was issued in, rather than a report regenerated later out of the account's current state.
• Confirmations of deposits and withdrawals. These sit outside the trading ledger and are the record of money entering and leaving, which is a separate question from what the trading did.
• The version of the firm's terms and its costs schedule that was in force at the time. Terms are revised; the version governing a transaction is the one that applied on the day, and superseded versions are rarely retrievable afterwards.
• Correspondence with the firm, including the reference numbers it quoted, since a reference in an email is often the only link between a conversation and a row.
• Any record of the platform's own state at a moment in question, such as a capture showing a price, an order and a clock in the same frame.

How long any of it is kept is not a question with a single answer, and this page does not supply one. Retention periods follow from the tax rules and the limitation periods of the jurisdiction a person is resident in, they differ substantially between jurisdictions, and why that question belongs locally rather than to a broker is the subject of a later lesson. The one observation that travels everywhere is that a retention decision can only be made while the records still exist, and an account closed and forgotten tends to take its history with it.

Where practitioners disagree 

Two arguments run through this subject and neither resolves. The first is about frequency. One school reconciles every closed position, on the grounds that a systematic error repeats, that the earliest instance is the cheapest to find, and that a baseline is only built by repetition. Another treats that as unsustainable for anyone trading often, and reconciles a period at a time instead, confirming the loop from opening balance to closing balance and opening individual rows only when the loop refuses to close. The second position is stronger than it first sounds, because a practice abandoned after a fortnight detects nothing at all, while a lighter one that is actually kept detects everything that reaches a balance. The limits are symmetrical too: per position work is thorough and frequently abandoned, per period work is sustainable and silent about an error that nets to zero across the window. Neither is put forward here as the one to adopt.

The second argument is about whether a terminal report counts as a record at all. One view holds that it plainly does, since it is drawn from the same ledger as the issued statement and reports the same transactions. The other points out that it is generated on demand out of the account's present state, that a correction posted later changes what it prints for a date already past, and that it carries no issuer and no period on its face. Both descriptions are accurate, and they answer different questions: one about where the figures come from, the other about what the document is evidence of.