HFT Myths

Quote from RedDuke:

Can you please describe the required steps for direct.

I did all my trading via FCMs, through their TT, PATS, zen fire or IB APIs and platforms.

If you connect to CME servers directly, who and how clears the trades? How does CME know who you are, and who gives you logon/feed credentials?

Thanks a lot
More...
When you apply for a connection at CME, you'll need to be pre-approved by a CME approved clearing firm before you're allowed to connect. You then get the credentials from the exchange directly.

Might want to try Advantage Futures as a step up from home retail into the DMA world. They advertise helping you every step of the way to connecting. I haven't worked with them directly but have heard good things about them.
 
Quote from hft:

Yes, there are some non-banks that I just don't have experience with that would be competitive as well, I know NewEdge in particular is just as competitive as the banks, particularly overseas.

And yes, Barclays and CS are just as competitive as the rest.

The differences between the pricing among all the firms mentioned are pretty small - once you're scaled up it's more of a matter of who suits your style of trading better and what kind of relationship you've established with them and their quality of service in areas other than pricing (i.e. how fast do they answer your inquiries, how much headache do they create/eliminate for your back office operations).
More...

Thank you for your generosity in sharing. I wish you much success!
 
HFT: do you have last look privileges on hotspot or currenex ? That sounds like it would be rather convenient for a hft-firm if you can sit on an order for n milliseconds, and then decide to either fill or reject it. Basically a license to print money, if n is large enough. OTOH, it sounds rather inconvenient if you dont have it, but other participants do. How many ms is the typical last look?

What do you think of the recent EBS plans to eliminate fifo order execution, and introducing randomized execution? That a problem for the typical hft strategies?

Have you heard about the Lucid Markets incident on Reuters, where they subscribed to a boatload of marketdata connections and got more snapshots that the others? Do you think its abusive? Apparently EBS and Reuters make good money on selling multiple marketdata connections to hft firms, but it seems like ebs at least gave in to their sell-side overlords for now (with half-pip pricing and random execution etc).
 
Quote from Pippi436:

HFT: do you have last look privileges on hotspot or currenex ? That sounds like it would be rather convenient for a hft-firm if you can sit on an order for n milliseconds, and then decide to either fill or reject it. Basically a license to print money, if n is large enough. OTOH, it sounds rather inconvenient if you dont have it, but other participants do. How many ms is the typical last look?

What do you think of the recent EBS plans to eliminate fifo order execution, and introducing randomized execution? That a problem for the typical hft strategies?

Have you heard about the Lucid Markets incident on Reuters, where they subscribed to a boatload of marketdata connections and got more snapshots that the others? Do you think its abusive? Apparently EBS and Reuters make good money on selling multiple marketdata connections to hft firms, but it seems like ebs at least gave in to their sell-side overlords for now (with half-pip pricing and random execution etc).
More...

Last-look is vital to market-making in FX. It's not all free money though. Similar to how you can last-look deals, you can receive reports on how often participants last-look you and turn off your flow to them. So if you last-look too often, you lose order flow.

The last look period is in the tens of millis range. They've changed it a few times (lower) over the years. Not that the actual number really matters, anything in the millis and above changes the game completely, for us at least.

The EBS plan is crazy, and I think/hope they don't go through with it. It'd really throw HFT's in a loop. Market-makers would widen out quotes and market-takers would pay up more crossing bigger spreads. I'd give it a 50 delta of passing though, since EBS has always catered to non-HFT's. I'd also bet that the plan would be rescinded in short order after their volumes drop though. They'd also lose possibly irrecoverable market share to reuters, hotspot, and currenex in the meantime though. Will be interesting to see how it pans out.

I'd heard from a high-level about Lucid but don't know much about it. I don't blame it completely on Lucid though, Reuters needs to manage their systems better. We've taken advantage of some Reuters quirks as well. All exchanges could use some oversight to ensure that their rules are enforced properly.

Those are all interesting points that you bring up. Are you in the industry - have any thoughts on them yourself?
 
What are your thoughts on exchanges giving structural advantages to certain groups - DMMs, floor brokers, super high rebate tiers, naked short selling exceptions, etc? How much to do you think these things help for HFT compared to a startup HFT shop without any of those advantages?

Do you know what level you need to go entirely self-clearing for US equities and avoid all the per share fees?
 
After doing extensive statistical analysis the NMS system is a seriously flawed.

Pre and post sessions are exempt from RegNMS quote protections and there are 105 trade exception codes that exempt a substantial amount of intraday trades that trade through the NBBO. The most prevalent are ISOs.

The most telling sign the game is some what rigged to allow market insiders to screw the retail traders are the 1 second flicker quotation exemption given to all trading centers.

The SEC allows trades to be filled at the worst bid or offer over the preceding one second. Exchange time stamps are in one second increments therefore they have up to 2 seconds to shake and bake you.

They even set special rules up to define what trades can set last, increment volume, set highs, lows etc. Those with retail feeds see one set of data while the institutional feeds get a completely different set of data.

The electronic blue sheets FINRA and the SEC use to enforce trade violations don't even have timestamps prior to November 2012. Plus they rely on the foxes to inventory the hens and count the eggs.

Consider in 1 second the exchange in memory 300ns matching engines have the capacity to process 3,000,000+ trades.

The trading center game:

1 - 2 second to legally stick it to you.
Matching engines that run at 300ns without audit trail
A enforcement system that didn't have any time stamps.
Subpenny advantage
Order Flow priority
Visibility to all resting orders...


Given these facts no retail trader should ever engage in any sub second strategy. You can't win in these time frames when they can legally stick it to you.

For subsecond HFT... Either JBO and play as an insider or get gamed.

If you do trade retail HFT:
1. build and test your strategies using second bars based on the lowest bid and highest offer.
2. Set a minimum of 1 second between trade actions
3. Carefully monitor and manage all exchange resting orders

This is your reality... play smart
 
Quote from hft:

Glad to see a post from a fellow HFT'er :). Good questions to ask, that's why I can't answer many of them :p.

1. There's quite a few that come to the top of my head. In no particular order: Jump, Teza, Spot, Peak 6, Optiver, IMC, Jump, Allston, Volant, XR, Ronin, EWT, Eagle 7, Wolverine, and the list goes on for a while longer.

2/2/3/4/5/6/7. Can't tell you.

8. No. We refuse to even pay for Windows or Office. It's open-source or in-house here.
More...

I cannot confirm nor deny :D. Maybe you can shed some light on what your typical day is like.
 
Thanks to start the thread. I have the following questions, hoping not too naive.

1. Is this a fastest win all game? or is it a profit sharing game?

2. Given your knowledge, if you want to start your own, can you compete? What is the cost to setup, hardware cost + labor cost?

3. If given you a limited resource, say 10% of your current company cost, can you still get a profitable solution?

4. In such a short time, do statistics really works? If there exists an optimal solution, and you are implement an sub-optimal, e.g. a little bit delay than competitor, actual probability distribution is not the exact assumed distribution in the model, will the profit drop like a waterfall, or not that sensitive?

5. How similar between strategies developed by competitors given most of us get the similar education and intelligence? Are you compete totally innovated ideas or just try to get some variants of a main theory? or only adjust some parameters empirically?

6. What is your company's edge on earth? Enough R&D fund? Smarter employee? Fastest?

7. Given you are profitable now, what is the failure possibility in the near future due to competition?
 
1. Assume total profit for hft in a market is fixed. With more competitors, are some hft profiting from other hfts?

2. in hft, are 20-80 rules working, that is 80% hft lose and only 20% profit?

3. Any hft explore the weakness of other hft? Like virus, you are trying to find other people's loophole
 
You must log in or register to reply here.
Back
Top