Hi Remiraz,
No expected future problems with profit sharing under the heading of this thread:
"Let's collaborate - ..."
Let's collaborate - components of a good ATS
- Thread starter 65Matt
- Start date
No expected future problems with profit sharing under the heading of this thread:
"Let's collaborate - ..."
Collaborating? no problem as the venture will likely never lead to anything profitable.Quote from Remiraz:
not sure if i'm getting the joke here...
1/10 of a ms! what a bunch of BS.
Quote from dersu:
That's what my company has. It is making money constistently on high-frequency arbitrage using custom developed trading system with the most critical part of it a high-efficiency market depth processor that provides marked depth info with latency less than a millisecond (actually 1/10 of ms). It conforms to posted ATS "spec" and the funniest thing - it is running on abunch of cheap PCs under Windows
missing from the discussion is a classification of automated systems. I am aware of only two significant ideas in this area.
1) Mean reversion. This one was first put to use in the early-1980s at Morgan Stanley i believe. In original form it looked something like this: you short the 200 most appreciated stocks over last 5 days, and buy the 200 fallen most over the same period. You hold one day. This strategy consistenly made money over some 20 yrs for them and several others. it was making something like 0.5% per day, and was diminishing only gradually over the years. not profitable any more in original form, though modifications are the bread and butter of statistical arbitrage.
most current automated systems fall into this category, one way or another.
For example market-making, index convergence trades, and the like all are based on mean reversion.
2) Trend following. The earliest reference i find is to the group calling themselves Turtles, though they didnt trade with computers, but they did seek to make the rules completely objective. This stuff generally works worse than mean-reversion, is prone to inconsistency, and not really suitable for high-frequency trading. Unlike 1, it is possible to game the practitioners of this art.
these two are in some sense opposite of each other, i mean for every indicator it is possible to construct a strategy which goes along with it, or goes countertrend when the indicator peaks. Because of that, i wonder if anything fundamentally different is even possible. Except no 3 below...
3) Real arbitrage: component stocks versus index futures, and no doubt many other kinds of real arbitrage i am not aware of. Like convertible arb, triangular arbitrage in currency, buying oil or metal in one part of the world and selling in another etc... ok these do not really fall under category of automated trading but the first example i mentioned is: the best guys at this is Cooper Neff in Phily.
That i think is the big picture. Did i miss something?
1) Mean reversion. This one was first put to use in the early-1980s at Morgan Stanley i believe. In original form it looked something like this: you short the 200 most appreciated stocks over last 5 days, and buy the 200 fallen most over the same period. You hold one day. This strategy consistenly made money over some 20 yrs for them and several others. it was making something like 0.5% per day, and was diminishing only gradually over the years. not profitable any more in original form, though modifications are the bread and butter of statistical arbitrage.
most current automated systems fall into this category, one way or another.
For example market-making, index convergence trades, and the like all are based on mean reversion.
2) Trend following. The earliest reference i find is to the group calling themselves Turtles, though they didnt trade with computers, but they did seek to make the rules completely objective. This stuff generally works worse than mean-reversion, is prone to inconsistency, and not really suitable for high-frequency trading. Unlike 1, it is possible to game the practitioners of this art.
these two are in some sense opposite of each other, i mean for every indicator it is possible to construct a strategy which goes along with it, or goes countertrend when the indicator peaks. Because of that, i wonder if anything fundamentally different is even possible. Except no 3 below...
3) Real arbitrage: component stocks versus index futures, and no doubt many other kinds of real arbitrage i am not aware of. Like convertible arb, triangular arbitrage in currency, buying oil or metal in one part of the world and selling in another etc... ok these do not really fall under category of automated trading but the first example i mentioned is: the best guys at this is Cooper Neff in Phily.
That i think is the big picture. Did i miss something?