Four rule changes, thirteen years apart. You are the one placing the orders, and every screen ends with a number you paid.
1997. Quotes move in eighths of a dollar
A specialist posts a price to buy from you and a price to sell to you. By rule the smallest gap
between them is one eighth, 12.5 cents. He is showing size, because that spread pays him to.
Bid, he buys at
$20.00
5,000 shares
Spread
12.5¢
Ask, he sells at
$20.125
5,000 shares
0
Round trips
$0
What it cost you
$0
What he earned
You did nothing and you are down money. You bought at his ask
and sold at his bid, and the stock never moved. That gap is not a fee anyone announced. It is the price of
demanding immediacy, and it is the oldest revenue line in the securities business.
2001. Decimalization. The tick drops to a penny
Same stock, same specialist, one rule change. He can now quote in cents, and competition pushes
the spread to the minimum. Look at what happened to the size he is willing to show.
Bid
$20.00
300 shares
Spread
1¢
Ask
$20.01
300 shares
Same trade in 1997
$12.50
In 2001
$1.00
Twelve times cheaper, and that is real. For anyone trading a hundred shares, the modern
market is not slightly better, it is dramatically better. Most of the argument about market structure ends here,
and for most investors it should.
Now try it with size
He only shows 300 shares at a penny, because a penny does not pay him to show more. Drag the
order up and watch what you have to climb through to get filled.
Order size100 shares
Round trip in eighths, 1997
$12.50
Round trip in pennies, 2001
$1.00
Find the crossover yourself. Somewhere around 3,750 shares the thin penny book becomes
the more expensive place to trade, and the old wide-spread market with real size behind it wins. Decimalization
did not make trading cheaper. It made small trading much cheaper and large trading harder, which is why
institutions started slicing orders into hundreds of small pieces, and why there is now an industry built on
doing that for them.
2010. Everything is fast now
Quotes update many times a second. You click on a price you can see. Between your click and your
order arriving, that price has had time to leave.
Bid
$20.00
300 shares
Spread
1¢
Ask
$20.01
300 shares
0
Orders sent
0.0¢
Average slippage
0
Filled worse than shown
Run ten without co-location, then ten with it. The spread on the screen says one cent
either way. What you actually pay does not, because in the time your order spends travelling, the people who can
see it coming have already moved. Co-location is sold by the exchange, by the rack, by the metre of cable.
This is the part that is genuinely contested. One reading is that you are being front run.
The other is that a market maker who cannot pull a stale quote has to widen it for everyone, and you would pay
that instead. Both are defensible and the evidence is mixed. Decide which you believe and be ready to say why.
May 6, 2010. What a thin book does under stress
Nothing here is new. Same penny spread, same 300 shares a level, same market makers who are free
to stop quoting whenever they like. A fund now has to sell 20,000 shares and does not care what it gets.
$20.00
0
Shares filled
$20.00
Worst price
20,000
Order left unfilled
You predicted this two screens ago. Three hundred shares a level and no obligation to
quote is a market that works beautifully for a hundred shares and has nothing underneath it when somebody needs
to move twenty thousand. The 2010 crash was not caused by a new kind of trader. It was the ordinary structure,
asked a question it was never built to answer.
Now tick the band and sell again. The band does not add a single buyer and it does not
calm anybody down. It stops a price from printing more than a set distance from where the stock just was, and
pauses trading instead. That is the whole fix, and it is why a bigger volatility spike in August 2024 produced
nothing like 2010.