1. The spread as a cost: what you lose the instant you enter

Costs 2: What you pay and cannot see

Some costs arrive on an invoice and some arrive hidden inside the price. The previous level dealt with the first kind: commissions, custody, currency, management fees. They show up on a statement and you can compare them between brokers. This level is about the others, the ones nobody bills you for and which, precisely because of that, take more money out of active portfolios every year than all the commissions put together.

We begin with the most immediate of them all, the one you pay in the very second you execute: the spread.

The spread is not a quote, it is a toll

The order book holds two prices side by side: the best at which somebody will buy from you (the bid) and the best at which somebody will sell to you (the ask). The distance between the two is the spread; the point halfway between them, the mid, is the best estimate available of the fair price at that instant.

An example with numbers in it. A share is quoted at 19.98 bid and 20.02 ask. The mid is 20.00 euros and the spread, 4 cents. It pays to express it in basis points: a basis point is a hundredth of a percentage point, that is, 0.01%. Four cents on twenty euros is 0.20%, or 20 basis points.

Send a market order to buy and you are filled at 20.02. You have paid two cents above the fair price: 10 basis points, half the spread. Your position is born with a latent loss of exactly that size, and not because you got anything wrong, but because you wanted in now. On the way out you will sell against the bid and pay the other half: the complete trade costs the whole spread, 20 basis points.

One detail separates those who measure from those who do not: the correct reference is the mid at the instant your order reaches the market, not the last traded price. Out of that comes the effective spread, twice the distance between your execution price and that mid. Sometimes it turns out better than the quoted one, because somebody improves on the price inside it; sometimes far worse, because your order ate through several levels of the book. The quoted spread is the promise; the effective spread is what you paid.

The multiplier almost nobody applies: turnover

Ten basis points sound like nothing, and they are nothing if you trade twice a year. They stop being nothing the moment turnover enters the picture: how many times a year the whole portfolio is renewed. Each turnover means two crossings of the spread, one to buy and one to sell, so the annual cost is two times turnover times half the spread.

• Turnover of 2 (you buy and sell twice a year): 4 crossings, 40 basis points. 0.40% a year.
• Turnover of 20: 40 crossings, 400 basis points. 4% a year.
• Turnover of 100: 200 crossings, 2,000 basis points. 20% a year.

Put it in euros and it hurts more. A portfolio of 100,000 euros that turns over a hundred times a year trades ten million in purchases and another ten in sales: twenty million of volume. 0.10% of twenty million is 20,000 euros. That strategy has to produce 20% gross merely to break even, before commissions, before tax and before it has made a single euro. And there is nobody to haggle it down with, because nobody charges it: the price takes it.

One basis point against fifty

The second variable is liquidity, and its range is brutal.

In a heavily traded large-cap security the usual spread is around one or two basis points. In a small, thinly followed one, fifty is perfectly normal, and at the most illiquid end you see spreads of a hundred or more.

Take the same strategy, with a gross edge of 30 basis points per complete trade, and run it in both universes:

• Liquid universe, spread of 4 basis points: the trade leaves 26 basis points net. The strategy works.
• Illiquid universe, spread of 50 basis points: the trade leaves minus 20 basis points. The same idea, the same signals, the same hit rate, and it loses money systematically.

The cruelty of it is that the juiciest inefficiencies tend to live where the spreads are widest: were they easy to exploit, they would not still be there.

And a third variable: the spread is not constant. It widens at the open and at the close, on news and, above all, in episodes of stress, when whoever is quoting prices cuts the size on offer and backs away from the mid to protect himself. In March 2020 the spreads on an enormous variety of instruments widened very markedly for several days. The cost of trading goes up precisely when you most want to trade.

Implementation shortfall: the only number that matters

All this arithmetic flows into the central concept of the level, set out by André Perold in the late eighties under a title that says it all: the portfolio on paper versus reality.

The implementation shortfall is the difference between what your strategy would have returned had every trade been executed instantly and without friction at the price ruling when the decision was taken, and what your real portfolio actually returned. Everything lost along the way fits inside it: the spread, slippage, the impact of your own orders and the trades you decided on and never got round to executing.

The practical consequence is easy to state and hard to accept: the return of a backtest is a hypothesis; the return minus the implementation shortfall is the result. A strategy is judged not by what it promises on paper but by the size of the hole between the paper and the statement. The next three modules take that hole apart piece by piece.