3 ms·

It shifts the problem so that every bot is waiting for the end of each second to gather as much information as possible from external sources before placing its

It shifts the problem so that every bot is waiting for the end of each second to gather as much information as possible from external sources before placing its orders. So when a human trader places a market order, they will have a disadvantage equal to the volatility over 1 second. By contrast, in a real-time market, the HFT bots make your market orders very close to the current price everywhere else. It's not clear to me that batching actually solves all our problems. It might be better, but it doesn't seem 100% clear cut.

mkehrt 11y ago
Maybe randomly pick an auction length from like 0.5 to 1.5 seconds each time, so you don't know what batch you're in?
- unabridged 11y ago
  There is currently a dark pool using a similar method: "The London Stock Exchange’s Turquoise Midpoint Dark Pool has a service known as Uncross, which matches orders at the midpoint of the bid and offer with auctions held randomly every 5 to 10 seconds." https://blogs.cfainstitute.org/marketintegrity/2014/11/10/are-frequent-batch-auctions-a-solution-to-hft-latency-arbitrage/ https://blogs.cfainstitute.org/marketintegrity/2014/11/10/ar...
  - gd1 11y ago
    Which is kind of the point. If people think that rolling auctions are better than a CLOB, they are free to set up exchanges that use auctions and see if anybody trades on them. Spoiler: No one will. Or are they arguing that the government should step in and tell people how to trade?
unabridged 11y ago
Yes, in the case where external pricing information is released a short time before the window closes, collocated HFT will be able to win. But I'm not sure it would be any worse than the current situation. Also with batch auctions there can be multiple collocation points in different cities. But when the external information is released earlier in the window (which should be just as likely if random, and maybe even mandated for press releases to occur at the beginning of the window) the playing field between collocated HFT and smaller traders is much more level than the current situation. But most information used in trading isn't from external sources, its from the order book itself. Collocated HFT with 1ms latency can read the book, calculate next move (let's say 20ms), and submit new order in 1 + 20 + 1 = 22ms. A trader with 100ms latency and the same system would take 220ms to respond. This is where most of the HFT advantage comes from, they can see the order book and respond before it even arrives at your computer. With a batch auction everyone sees the previous results at the beginning of the window and has 1s to respond. Low latency is still desirable to allow for more processing, but its no longer overpowered. edit: 1 sec was chosen completely arbitrarily, the best choice could be more or less, or even a random sized window.
Gustomaximus 11y ago
Easy solution, randomise batch release between 0.9 and 1.5 seconds rather than an absolute time.
cdetrio 11y ago
> when a human trader places a market order, they will have a disadvantage equal to the volatility over 1 second. No, there's no disadvantage for the human's 1-second-old stale order because batch auctions use uniform-price clearing (aka single-price clearing, all orders in a batch interval transact at the exact same clearing price).

mkehrt 11y ago

Maybe randomly pick an auction length from like 0.5 to 1.5 seconds each time, so you don't know what batch you're in?

unabridged 11y ago
There is currently a dark pool using a similar method: "The London Stock Exchange’s Turquoise Midpoint Dark Pool has a service known as Uncross, which matches orders at the midpoint of the bid and offer with auctions held randomly every 5 to 10 seconds." https://blogs.cfainstitute.org/marketintegrity/2014/11/10/are-frequent-batch-auctions-a-solution-to-hft-latency-arbitrage/ https://blogs.cfainstitute.org/marketintegrity/2014/11/10/ar...
- gd1 11y ago
  Which is kind of the point. If people think that rolling auctions are better than a CLOB, they are free to set up exchanges that use auctions and see if anybody trades on them. Spoiler: No one will. Or are they arguing that the government should step in and tell people how to trade?

unabridged 11y ago

Yes, in the case where external pricing information is released a short time before the window closes, collocated HFT will be able to win. But I'm not sure it would be any worse than the current situation. Also with batch auctions there can be multiple collocation points in different cities. But when the external information is released earlier in the window (which should be just as likely if random, and maybe even mandated for press releases to occur at the beginning of the window) the playing field between collocated HFT and smaller traders is much more level than the current situation. But most information used in trading isn't from external sources, its from the order book itself. Collocated HFT with 1ms latency can read the book, calculate next move (let's say 20ms), and submit new order in 1 + 20 + 1 = 22ms. A trader with 100ms latency and the same system would take 220ms to respond. This is where most of the HFT advantage comes from, they can see the order book and respond before it even arrives at your computer. With a batch auction everyone sees the previous results at the beginning of the window and has 1s to respond. Low latency is still desirable to allow for more processing, but its no longer overpowered. edit: 1 sec was chosen completely arbitrarily, the best choice could be more or less, or even a random sized window.

Gustomaximus 11y ago

Easy solution, randomise batch release between 0.9 and 1.5 seconds rather than an absolute time.

cdetrio 11y ago

> when a human trader places a market order, they will have a disadvantage equal to the volatility over 1 second. No, there's no disadvantage for the human's 1-second-old stale order because batch auctions use uniform-price clearing (aka single-price clearing, all orders in a batch interval transact at the exact same clearing price).