The Arms Race in High Frequency Trading

[DeeDeeTwo]

Quote from nitro:

It is creative and expert programming.

Yes.

But the logic is hardly rocket science...
They are all doing the same thing...
Running variations on Market Making or Relative Value algorithms...
But with low latency.

A lot of this is a commodity circa 2009.

One can simplify all this...
And view Algo Trading as a separate Zero Sum Game...
Where only the Black Boxes are playing...
And competing against other Black Boxes.

So what do you get?

The usual Zero Sum Game distribution...
Where 2% make a lot of money...
Another 4-8% make a small profit...
And 90% are running Black Boxes at a loss.

But the people servicing this sexy, sexy niche...
They always get paid.

 

[nitro]

I agree completely.

Quote from Euler:

I agree, nitro. Hardware and latency certainly deserve due consideration and expenditure, but obsessing over these alone (as the original blog author seems to imply that many are doing) will not keep one in the game long-term, I think. It's the entire system that counts, and as you say, other aspects are more important, in general.

High freq trading is a very competitive business and only those that provide liquidity at the most competitive prices, yet still profit, will survive. Anyone who's in it and profitable today should assume that their edge could be gone in a matter of weeks, lost to even better systems. In the end, as you say, society benefits with more efficient markets.

 

[nitro]

Well, sure. Overshoots are part of markets, and the effect is that on an overshoot, the people that are responsible for bringing back prices to "FV" will step in and take risk that prices are out of whack.

As far as a mutual fund that is insensitive to price, they will not be in business for very long if that is the way they continue to do business, imo. Every place that I have ever been in uses VWAP or TWAP or some other sophisticated method of acquiring an asset to minimize their own effects on liquidity. They are not perfect, but they are a far cry from "insensitive to price."

I don't think there is less liquidity at all. There is only possibly less liquidity to the initial buyer, not to the traders that bring things back inline on overshoots. Further, who do you think is on the other side of the trades of the mutual funds? MMs more often than not. Who do you think notices that they are getting lifted? So you want them to take note of the fact that someone is trying to run them over as cheaply as possible, and stand there and take it? That does happen to some extent in super liquid assets like ES, ED and equities like MSFT that have huge size deep into in the book, but on anything but A tier stocks? Come on. I would love it if anytime I wanted a fill on 5000 strangles the other side would sit still for me too.

All markets participants anticipate and react, and mutual funds are the sardines for the sharks, dolphins, whales, birds, etc. Maybe mutual funds should hire traders? Novel.

Quote from black diamond:

I would agree in most situations, but don't you think aggressive front running could mess this up? The scenario I am thinking of is a HFT takes liquidity in advance of say a price-insensitive mutual fund order, scares some more out of the book, then the mutual fund trade goes through at a worse price than it would have, and the price reverses later when the book replenishes. So in this case you don't have a step function, you get an over-shoot then a correction.

I don't know how much mutual fund trading actually goes into the order book nowdays, but I think the effect would be similar if it gets crossed somewhere based on a price from the exchange. Besides, I am sure there is somebody worth some sympathy whose trading would get affected even if it is not a mutual fund.

This would be bad for society - less liquidity and worse price discovery.

 

[nitro]

Quote from rosy2:

I agree. IMO the hardware and latency hype you hear in the news is advertisements so firms buy the latest stuff. Latency and hardware are important but just because you have the fastest infrastructure around means nothing unless you have some kind of edge.

It seems some firms are trying to create this ultra-fast system but have no strategies to put on it. And some strategies dont even need the extra speed.

Half agree. I have never been to a firm that buys the hardware or fast pipes, and then hopes it can make use of it. There is almost always a business case first. It doesn't mean it isn't risky, it just means that it is a logical risk like amything else in this business.

 

[nitro]

Ok. I sort of agree that most of the algos running on these things are not rocket science. I guess I miss the point.

Quote from DeeDeeTwo:

Yes.

But the logic is hardly rocket science...
They are all doing the same thing...
Running variations on Market Making or Relative Value algorithms...
But with low latency.

A lot of this is a commodity circa 2009.

One can simplify all this...
And view Algo Trading as a separate Zero Sum Game...
Where only the Black Boxes are playing...
And competing against other Black Boxes.

So what do you get?

The usual Zero Sum Game distribution...
Where 2% make a lot of money...
Another 4-8% make a small profit...
And 90% are running Black Boxes at a loss.

But the people servicing this sexy, sexy niche...
They always get paid.

 

[stock777]

Better to call it what it is.

High Tech Frontrunning.

Micro Arbitrage.

Rebate humping.


Whatever you call it , makes for a very twitchy tape most of the time. I don't like it.

Too much like battling the bots (set on high skill ) in a video game.

Was better in the old days.

Tell you what, I rarely if ever see a description of what these hft guys are actually doing. It can't all be top secret proprietary , if I tell you I'll have to kill you stuff.


How about a simple flowchart style example of what a basic hft does to make their billions a week.

 

[jack hershey]

Quote from stock777:

Better to call it what it is.

High Tech Frontrunning.

Micro Arbitrage.

Rebate humping.


Whatever you call it , makes for a very twitchy tape most of the time. I don't like it.

Too much like battling the bots (set on high skill ) in a video game.

Was better in the old days.

Tell you what, I rarely if ever see a description of what these hft guys are actually doing. It can't all be top secret proprietary , if I tell you I'll have to kill you stuff.


How about a simple flowchart style example of what a basic hft does to make their billions a week.

Would you punch up the symbol set you want used?

 

[buylo]

Quote from propseeker:

yup... funny thing is, it's also the highest cost with inefficient and error-prone programming... either way, you pay.

Indeed.......nothing like good old fashioned price discovery and fundamentals. My profits have been supporting programming models that haven't worked worth a shit the past 5 years. I've stopped introducing myself to programmers, because they're gone soon enough and replaced by others.

 

[lolatency]

My background is in tweaking and accelerating hardware and software for this stuff. I've been in this game for 5 years. It is -the- best game in town and it is -the- best job to have. My only mistake was moving into options MM space for a while, thinking it was as exciting as stock space. Most of the options guys are slow beasts, and many of the managers don't see the forest from the trees.

My background is in hardware design, systems software, and a (recent[1]) graduate degree in statistics. I was born for low latency trading.

A lot of firms will fail simply because you can't get the programmers on Wall St. to do this correctly. You have to run out to Silicon Valley, and even then, you have to find the guys who love trading a lot.

[1] I still have some soft spots, but enough understanding to make it happen.

 

[jack hershey]

Quote from rosy2:

I agree. IMO the hardware and latency hype you hear in the news is advertisements so firms buy the latest stuff. Latency and hardware are important but just because you have the fastest infrastructure around means nothing unless you have some kind of edge.

It seems some firms are trying to create this ultra-fast system but have no strategies to put on it. And some strategies dont even need the extra speed.

Here is an example.

HF would be necessary here to deal with running 5 to 10,000 contracts. Each entry(1), reversal(8) and last cover (1) would have to be partial fills interlaced into the T&S and the book that is showing. This really isn't front running too much as it is just staying on the right side of the market.

The two most obvious places where improvement is needed are the multitrade sections.

By taking the trading frequency up to the next two fractal levels of reversals two times it becomes apparent that in those 50 trades you no longer have to give up the volatility of the bars.

As shown going up the volatility isn't much and if doing 10 to 20 partial fills per turn a lot of it is captured by the interval required by the partial fills. spreading the partial fills by an interval of time (100millisec) or by limiting the % of cummulative is still kinda crude but workable.