Price sources and how a quote is built

A share has an exchange. A currency pair does not. There is no single venue on which the price of one currency against another is established, and no official closing price for it either. What exists is a network of banks and electronic venues, each quoting to its own counterparties, and a price for a currency pair is a summary of what that network is quoting at a given instant.

This is why two brokers can display different prices for the same pair at the same moment without either being wrong. Each is summarising a different set of sources, updating on different intervals, and applying its own handling to the result. The differences are usually a fraction of the spread and occasionally larger, and they are a property of a decentralised market rather than a discrepancy to be resolved.

Aggregating competing feeds 

A firm receives a stream of two way prices from each of its liquidity providers, each with a quantity attached. Aggregation is the process of merging those streams into one book: every provider's bid is sorted with every other provider's bids, every ask with every other ask, and the top of the resulting book is the best bid from any provider paired with the best ask from any provider.

The pairing has an important consequence. The best bid and the best ask usually come from different providers, so the aggregated top of book is narrower than any individual provider's own quote. Adding providers to an aggregation therefore narrows the aggregated quote mechanically, up to the point where the best prices stop improving, which is the practical reason firms maintain relationships with several rather than one.

The final row is the mechanism in one line. No provider was quoting the aggregated price, and the aggregated price is nonetheless real: an order to buy would go to provider C and an order to sell to provider B, and each would be filled at the price that provider was quoting.

What sits between the feed and the screen 

A raw aggregated book is not displayed as it arrives. Several filters sit in between, and each of them is a deliberate design decision with a cost as well as a benefit.

• Bad tick filtering. A single provider publishing an erroneous price far from the others is excluded rather than allowed to set the top of book, since an erroneous price would otherwise trigger every resting order near it.
• Rate limiting. A feed updating thousands of times a second is throttled to a manageable rate before display, which means the displayed price is a sample of the underlying stream rather than every update in it.
• Staleness checks. A provider that stops updating is dropped from the aggregation rather than left contributing a price that is no longer live.
• Minimum quantity thresholds. A provider quoting an excellent price in a quantity too small to be useful can be excluded from the top of book so the displayed price is one that can actually be transacted in.

Rate limiting is the filter most visible to an account holder, because it explains why a chart drawn from a broker's feed and a chart drawn from another source do not agree tick for tick. Both are sampling the same continuous process at different intervals, so their highs and lows differ slightly and their candles are not identical. Neither is a record of every trade, because in a decentralised market no such record exists.

Instruments that do have a venue 

A share contract references a listed security, so the price comes from the exchange the share is listed on, delivered through a market data vendor. There is a single authoritative price, and the questions that remain are about the data licence, the delay if any, and which of a share's multiple listings is being referenced where a company is listed in more than one place.

A cash index contract references a published calculation. The index provider computes the level from constituent prices and publishes it on an interval, and outside the constituents' own trading hours no live calculation exists. Firms that quote an index contract outside those hours are quoting a derived price, most often from the corresponding futures contract, and the instrument's specification states this rather than leaving it to be inferred.

Commodity contracts reference futures prices from the exchange the futures are listed on, so the same authoritative source applies. What differs is that the contract references a particular contract month, so the price being read changes at every roll date rather than being one continuous series.

What the firm adds to the aggregated price 

Between the aggregated book and the displayed quote a firm applies its own commercial handling, and there are only two ways to do it. Either the aggregated price is passed through unchanged and the firm charges a separate commission, or a margin is added to each side of the aggregated price and no separate charge is made. Both are disclosed, both cover the same underlying costs, and the choice between them is a packaging decision rather than a difference in economics.

What is not a packaging decision is whether the handling is symmetrical. A margin applied equally to both sides moves the mid price nowhere and widens the quote. A margin applied to one side only moves the mid price, and the direction it moves in is information about the firm rather than about the market. Execution policies state which is applied, and it is a disclosure worth reading.