Market foundation

A price is not a number, it is an agreement

A price shows on your screen. That single figure hides three pieces of information, and confusing them is the first source of misunderstanding. There is the last traded price, a fact of the past: two people agreed, the deal is done. There is the best bid, the highest price at which someone commits to buy right now, and the best offer, the lowest at which someone commits to sell. You never buy at the last traded price, you buy at the best offer.

The gap between those last two has a name, the spread, and it is not just another fee. It measures a disagreement: nobody wants to buy at the price someone is willing to sell at, so nothing happens. A trade is born the instant someone gives up waiting and crosses the gap. Whoever crosses pays, whoever waits gets paid.

A displayed price only holds for a given quantity, never beyond it. If your size exceeds what is offered at the best price, the rest of your order fills at the next level, a worse one, then at the one after that. Your real price is therefore an average of the levels you consumed, and it drifts away from the displayed price as your size grows against what is waiting in the book. A price with no quantity attached to it means nothing.

The machinery has a formal name, the continuous double auction. Vernon Smith, back in 1962, ran artificial markets where a handful of students, each knowing only their own reserve value, converged within a few rounds on the theoretical equilibrium price. The work earned him a Nobel prize forty years later, and it states something counter-intuitive: a market does not need informed participants to produce a coherent price, it needs a matching rule and people on both sides.

Where your order really goes

A futures contract trades on an organised market, with one single central book: one contract, one exchange, one queue. That queue is served by price priority then time priority: the better price goes first, and at equal price whoever has waited longest goes ahead. Every trade is printed, which gives a volume genuinely exchanged, verifiable by anyone.

A share works differently, and the simplified version everyone repeats is wrong. There is not one book but several: the historic exchange, and the competing venues quoting the same share. Your broker decides where to route your order, and it owes you best execution, not any particular venue. What rescues the reading is that all those venues report their trades to a consolidated tape: the book is split, the published volume is still what changed hands.

On the spot currency market there is no central book and no exchange at all. It is an over-the-counter network where your broker shows you THEIR price, aggregated from the ones their liquidity providers send. Two brokers therefore display two slightly different prices at the same instant, sometimes two different daily highs on the same pair. The “volume” on your chart is not volume traded: it is the number of price changes your broker happened to see.

⛔ The difference is not a specialist's quibble. Any volume-based reading on a currency pair measures the refresh rate of one intermediary among hundreds. The test is within your reach tonight: compare the day's high for the same pair on two data sources. If you want currency volume that genuinely changed hands, the matching futures contract is centralised and published.

Crypto is an in-between case: each venue has a real book and real volumes, but there are dozens of them and no official consolidation. “The price of bitcoin” is an average of prices all of which are real and all of which differ.

The clearing house, the invention that makes a market safe

Go back to the question that sorts every product: who pays me if I am right? On an organised market the answer is not “the person who sold to me”. The moment your order is matched, a clearing house steps in through an operation called novation: one trade becomes two obligations, the buyer facing the clearer, the clearer facing the seller. Neither side knows who was opposite, and neither needs to.

The clearer does not provide that service blindly. It demands initial margin from both camps, then settles gains and losses every day, by margin call. Your loss does not wait for the position to close before being debited, which is why a futures account can be shut on a Tuesday over a position that would have been right by Friday: the clearer's whole job is never to be exposed to more than one day of movement.

The arrangement took hold in Chicago in the nineteen twenties, after decades in which one operator's default travelled back up the whole chain. It does not abolish counterparty risk, it stacks it into a waterfall and puts a number on every step. Should a member default, the loss is absorbed first by the margin that member had posted, then by its share of the default fund, and only then by the shares of every other member. Each step of that waterfall is published and known in advance, which is precisely what is missing when your counterparty is a company you know nothing about.

⚠️ Where there is no clearer, the question comes back untouched. On a CFD or a spot currency position, your counterparty is the broker itself. Segregation of client money and a regulator's supervision stand in for the clearing house without playing quite the same role. That is the real reason the entity holding your account, and the regulator it answers to, deserve more attention than the spread quoted in the advert.

What actually moves a price

There are never “more buyers than sellers”. Every single trade has exactly one buyer and one seller, for the same quantity, at the same price, always. The sentence you hear everywhere is arithmetically false, and it blocks any understanding of what is happening.

What moves price is not a headcount, it is the exhaustion of what is offered. Price ticks up when the last seller present at the current price has been served and the buyer still wants more: he then has to reach for the next level, higher up. On the way down it is the exact mirror. Price is the mark left by the imbalance between what is demanded immediately and what is offered patiently.

Hence the most useful distinction in the book. There are orders that wait, limit orders, which manufacture liquidity, and those that take, market orders, which consume it. Only the second kind moves price. A wall of buy orders sitting below is not buying pressure: it is an intention that costs nothing until it is touched, and it can be pulled in a millisecond.

One consequence follows, and it carries through the rest of the lesson: a large move on a thin book is not necessarily a large order. The same quantity moves price far more at three in the morning than in the middle of the afternoon, and what you read as conviction is often just an empty book.

What the market has already bought

News does not move price, it moves people, who move price by placing orders. The distinction sounds pedantic, and on its own it explains the phenomenon that infuriates every beginner: a number better than expected, and the market falls.

The reason is mechanical. Before the release a consensus exists, it is published, and positions have been taken on the strength of it. If the number comes out in line, everything that had to be bought has already been bought: no new buyer is left, only holders looking to cash in. The market falls on good news because the good news had already been paid for.

What counts is therefore never the level of the number, it is the gap to what was expected, combined with the positioning already in place. Two identical numbers produce two opposite reactions depending on what the market was carrying beforehand. It also explains why the second reaction, ten minutes later, often runs against the first.

The idea has a lineage worth knowing. Louis Bachelier, in his 1900 thesis Théorie de la spéculation, already wrote that the speculator's mathematical expectation is zero, since price incorporates what is known. Seventy years later Eugene Fama gave it its modern formulation, the efficiency hypothesis, in three degrees of strength. You do not need to believe the strong version to keep the useful part: what is known is already in the price, only the unexpected moves it.

⛔ In the seconds following a major release the spread widens, depth vanishes, and an order does not fill at the price you were looking at. Neither does your stop.

The three sessions, and what their overlap does

The currency market runs continuously from Sunday evening to Friday evening, which makes people think it is the same at every hour. It is nothing of the sort. Activity follows the office hours of the big financial centres, and three of them carry weight: Asia around Tokyo, Europe around London, America around New York.

The hours, in universal time to avoid misunderstandings. Tokyo runs roughly 00:00 to 09:00 UTC, London 08:00 to 17:00, New York 13:00 to 22:00. What matters is not the list, it is the overlap: from 13:00 to 17:00 UTC, London and New York are working at the same time. That is the window with the most people on both sides of the book, and therefore the one where a large position opens and closes at the lowest cost.

Organised markets add two singular moments. The opening auction and the closing auction accumulate orders then cross them all at once at a single price. The close is anything but a formality: on the large European venues it regularly concentrates more than a fifth of the day's volume, because it is the reference price for index funds.

⚠️ Two traps to do with the clock are expensive. Countries do not switch to summer time on the same dates, so “three in the afternoon” is not the same market moment in March and in November. On currencies the trading day rolls over at 5 p.m. New York time, when liquidity briefly drops to almost nothing. At the weekend, finally, the market reopens on Sunday evening sometimes a long way from the last price: a stop does not protect you against a gap, it becomes a market order filled at the first price available.

Why the same pattern is not worth the same at three in the morning

A chart pattern is not a magic object, it is the trace of decisions taken by people. A break means participants agreed to pay above a level others were defending. If nobody was there to defend it, the break means nothing, however pretty it looks.

In the quiet hours three things change at once and they stack up. The spread widens, so your entry costs more. Depth thins out, so your exit slips. The participants still around are mostly automated flows and hedging orders indifferent to your levels. The same rectangle, the same triangle, the same double bottom, with ten times fewer people behind them.

In practice the pattern still triggers, it simply gets no follow-through. Price crosses the level, nothing arrives behind it, it comes back inside two candles later. That is the mechanical explanation for a large share of false breaks, and your own record can confirm it without you taking anyone's word for it.

⚠️ Beware of the opposite conclusion, which is wrong. Quiet hours are not forbidden and busy hours are not safe: a major release lands in the heart of the busiest window, and that is when spreads widen most violently. The rule is not “when there are people, it is fine”, it is “the number of participants changes the meaning of what you read, in both directions”.

Liquidity, the condition for everything else

The word turns up in every section, so it deserves its definition. Liquidity is the ability to exchange a given quantity, right now, without moving the price. It has three dimensions and never only one: the spread between bid and offer, the depth available at each level, and resilience, how fast the book rebuilds after being emptied.

What it costs you comes in two pieces, and the first one is almost always miscounted. A round trip does not cost you two spreads, it costs you one. You buy at the offer and sell at the bid: each fill costs you half the spread against the mid, and two halves make one spread, not two. Checking it takes ten seconds, open a position and close it straight away with the price unmoved: the loss shown is one spread.

Slippage sits on top, the difference between the price you asked for and the one you got, and that one really is paid twice, once per fill, as soon as your size exceeds what is offered at the best price. A round trip therefore costs one spread plus two slippages, and none of it shows up on your chart: the candle records the price that existed, never the price your order was given.

The arithmetic is worth writing down, using the illustrative figures in the chart. A round trip costing three points in the heart of the overlap and ten points two hours after the American close is not “seven points dearer”: it is a method whose average gain must be seven points higher to deliver the same result. If your average target sits around twenty points, you have just handed over more than a third of it by choosing the hour badly. The hour belongs to the strategy, not to convenience.

⛔ Liquidity also decides what your stop is really worth. A stop is a market order triggered at a level: it guarantees a trigger, never a price. On a thin book, or inside a gap, it fills wherever somebody was still standing, and that place can be a long way from what you had planned.

Practising: the hour survey

This lesson is checked against your own record, not against my word. The exercise fits in four columns added to your journal and requires no new knowledge. For your next thirty trades, note the entry hour in universal time, the matching session, the spread displayed when you sent the order, and a cross if a major release fell within the hour.

Four buckets are enough: Asia, Europe alone, the Europe and America overlap, then after the American close. Once you have thirty trades, work out for each bucket the number of trades and the average result per trade. Do not look at total profit per bucket: it is dominated by trade count and tells you nothing. The number that speaks is the average per trade.

⚠️ Be honest about what thirty trades prove, which is next to nothing. Split across four buckets, seven or eight are left in each, a sample where two unlucky trades are enough to reverse the ranking. What you get is not a conclusion, it is a hypothesis to watch over the next three hundred: one bucket looks dearer than the others. A journal that skips that caveat turns noise into method, and that is how a perfectly good hour of the day gets crossed off on the strength of three bad trades.

Add ten minutes for the second part, which has the merit of depending on no sample at all: it measures rather than tallying outcomes. Record the displayed spread on your instrument at four fixed moments of the same day: the Asian open, the European open, the middle of the overlap, two hours after the American close. One pass is enough, because the spread at a given hour is a fact and not a statistic. You end up with your own cost curve, and it shifts the hour at which you apply your method without touching the method.