[benreesman]

No, algorithmic trading is not about leeching from anyone in any intrinsic way. Like all trading there are good and bad actors, and bad actors get a lot of press.

The two biggest roles that “algorithmic” or “high-frequency” or “electronic” traders play are:

- market making: someone has to warehouse a bunch of thing A a and thing B so that buyers or sellers can find a counter party. this used to be an Italian guy from Jersey called a “specialist” and he quoted in eighths so that the minimum he could make per unit was $0.125. he was not giving you a better deal than optiver.

- arbitrage: thing X costs a lot in Chicago and is cheap in New York, or vice versa. buy in the cheap place, raising the price, sell in the expensive place, lowering the price. wash/rinse/repeat: shit costs the same in both places now. people have been doing this since before the wheel, on foot.

[ncmncm]

Neither of those describes sub-millisecond, programmed trading.

It is, instead, simply gambling. Like poker, it is not really a game of chance, so those with better algorithms and faster hardware win, but it provides society absolutely nothing of value. Generally it amounts to leeching off of other traders who might be providing something of value.

It has got so many have custom electronics connected directly to the fiber that delivers market event reports, and trigger sending an order even before such a packet finishes arriving, the response calculated in an FPGA in under 100 nanoseconds. Nowadays those traders have equipment in cages in the same building as the exchange, and the fibers to all the cages are exactly the same length.

They may have a collection of order packets already formatted, and just pick one of them to send according to the update.

For a while some would trigger sending an order early, and then add a bad checksum if calculation indicated it was not right. The exchanges banned that, not because it was unfair, but because it added load not paid for.

A way to preserve value for the non-gamblers would be to place a (proportionally) small tax on trades, so whoever does it most frequently pays most.

[TedDoesntTalk]

How are market makers using algo trading to warehouse a bunch of stock?

[benreesman]

The use of models and algorithms is in an effort to not get caught on the wrong side of a big move, the technical term for this is adverse selection, more colloquially “getting lifted”. If someone is coming through with a massive order and you fail to anticipate this and cancel in time, you’ve effectively subsidized a block trader. Market makers want to participate on both sides picking up a small fee (typically one tick) for providing liquidity in a market that is seeing price action in both directions. They don’t want to get run over by agency execution people trying to lay off 21 billion in Tesla stock.

The common misconception is that in order to sell you a share of AAPL, the market maker first buys it cheaper and then sells it to you in some unfair way. There isn’t time for that: they already had inventory.

[kirse]

they already had inventory.

Lots of wiggle room in that "had inventory" definition as well.