Order Books, Spreads and Slippage: Why the Screen Price Is Not Your Price

Clients show me a screenshot of a price, then a screenshot of their trade, and ask why the two do not match. The trade is usually correct. What they were looking at was a quote, and a quote is not an offer to fill you at any size.

Execution is the part of trading nobody teaches, and it is the part that quietly takes a percentage of everything you do. Crypto venues make it worse than equities do, for reasons that are structural rather than sinister.

What the number on the screen actually is

Most price displays show the last trade, which is history, or the mid, which is the average of the best bid and the best ask. The mid is a price at which nothing traded and nothing is available. It is the midpoint of a gap.

The book underneath looks like this:

A market buy takes the best offer first. If your order is larger than that level, it takes the next one, then the next. Your average fill is worse than the best offer, and worse still than the mid you were quoted. That difference is slippage, and it is entirely a function of your size against the book’s depth.

Two orders of the same dollar value can have completely different slippage on the same venue five minutes apart. Depth is not a property of the asset. It is a property of the moment.

The three costs, separated

People collapse all of this into “fees” and then argue about the wrong one.

The commission is the disclosed number. It is the easiest to compare and usually the smallest.

The spread is what you pay for immediacy. Cross it to buy, cross it again to sell, and a round trip costs you the full spread before the asset has moved at all. On a thin pair this dwarfs the commission.

The slippage is what you pay for size. It scales with how much of the book you consume.

There is a fourth cost that only exists on-chain, which is that your pending transaction is public before it settles. Anyone can see it, and a searcher can place orders around it. The research literature calls the general phenomenon maximal extractable value, and the paper that named the problem, Flash Boys 2.0, is still the clearest description of how transaction ordering becomes a revenue stream. What it costs an ordinary swap, and what bounds the damage, is in MEV: the execution cost buried in your swap. On a centralised venue the order book is the exchange’s, and your order is not broadcast to the world before it fills. That is a real advantage of centralised execution and it rarely gets stated.

Order types, and the tradeoff each one makes

Do you need the fill more than you need the price?
  |
  +-- Yes, exit now, size is small vs the book
  |     -> market order, accept the spread
  |
  +-- Yes, but size is large vs the book
  |     -> split it, or use a limit ladder
  |
  +-- No, the price matters more than the timing
        -> limit order, accept that it may never fill

A limit order does not eliminate cost. It converts a certain cost into an uncertain one. You stop paying the spread and you start bearing the risk that the market leaves without you, which on a trending day is the more expensive outcome. Nobody ever gets a statement showing what the unfilled order cost them.

Splitting a large order across time is the standard answer to depth. It works because you are letting the book replenish between clips instead of eating through it in one pass. It fails when everyone else is doing the same thing in the same direction, which is precisely when you most want it to work.

Why crypto books are structurally thinner

Liquidity fragments across venues that do not share an order book. In Canadian equities, marketplaces operate under NI 21-101 and the trading rules in NI 23-101, with an order protection framework overseen through market regulation. That architecture exists to stop your order trading at an inferior price when a better one is displayed elsewhere.

Crypto has no equivalent. Fifty venues, fifty separate books, no obligation on any of them to route you to a better price on a competitor. Arbitrageurs stitch the prices together, roughly, most of the time, and the stitching frays exactly when volatility spikes. It is the same book a crypto ETF’s designated broker has to buy in when it assembles a creation basket, which is why a listed fund’s tracking error is usually an execution story rather than a management one. It also means the depth you are looking at may not be real, which is the subject of reading exchange volume.

Then there is the market maker’s position. A designated market maker in a listed equity has quoting obligations. A crypto market maker has none, and the rational response to a violent move is to widen the quote or pull it entirely. That is also why a liquidation cascade on a perpetuals venue prints through levels that looked perfectly liquid a minute earlier. Depth disappears at the moment you need it, which is a pattern the Bank of Canada returns to in its Financial Stability Report discussions of dealer behaviour under stress.

Each fill is its own transaction at its own price

Every fill is its own transaction at its own price. A single intended purchase that executed across six price levels is six data points, and the average is a derived number, not a reported one.

Your cost base is built from actual proceeds and actual costs, including the commission. It is not built from the price you saw when you clicked. If your records come from a screenshot rather than a trade file, they are already wrong, and the record-keeping approach that survives review starts from exchange exports. Most tax software for crypto reads those exports natively, which is the actual reason to use it.

The frequency and deliberateness of your execution also feeds the business versus capital account question. Someone running a limit ladder across three venues is describing a trading operation, whatever they call it.

Cutting execution cost beats chasing a better entry

Execution cost is the most reliably ignored number in retail crypto, and it is one of the few costs you can genuinely reduce. Not by picking better trades. By placing the same trades differently.

I would rather a client cut their round-trip cost by using limit orders and splitting size than have them chase a better entry. One of those is a repeatable edge and the other is a coin flip.

If you are reconstructing a year of trades and the numbers are not tying out to your statements, send me the exports and I will tell you where the gap is coming from.