Where does the automated retail trader start to find edge?

[rosy2]

Quote from Algo_Design_Kid:

MOST are Liquidity Providers.

Don't think you can play their 20ms game.

That is a fallacy.

You basically "get in where you fit in" my friend.

No one is going to give you their style if it is truly profitable.

My advice to you would be to focus on MEAN REVERSION.

GL

agreed. also you need to look at transaction cost. if scratching a trade costs next to nothing then you're competitive.

the retail guy might have an edge if his broker offers some ridiculous deal that gives him money. olde discount brokers did this in the mid 90s for bonds until some guy milked it for a week.

 

[jjw]

At Rithmic, we are trying to close the gap between retail trading and institutional HFT. We have recently released features in our software that will enable a trader to get market data and release an order based upon that market data in under 250 microseconds (assuming that the trader's analysis that results in the order release does not take too long).

Until recently a trader would have to put up a lot of capital to be able to get into this game (or be part of a group with such capital). Now, if one has c++ programming capabilities, one can trade pretty much as fast as anyone, without a massive capital budget.

This is very new and will continue to evolve, but it is exciting to be able to offer technology that really levels the trading playing field.

 

[tripledtrader]

Quote from garchbrooks:

What are the guys who are succeeding doing, in a general sense -- what types of strategies? Momentum, multi-day reversion to the mean type stuff, and what kind of hold types are we looking at?

Many strategies that worked before, still work today. But one adaptation needs to be made to all strategies, and that is not placing passive orders. If you're placing passive limit orders, the odds are stacked against you. You'll be sub-pennied when you're right (they step ahead of your order), and filled when you're wrong. It's a lose-lose for liquidity providers. So it is important to realize that you will probably have to pay the bid-ask spread in the new HFT world. If you're trading liquid securities the one cent spread, may not be a big deal. Trading illiquid securities in a scalping fashion, is nearly impossible. So time frame has to increase if you're trading illiquid stocks.

 

[Algo_Design_Kid]

Quote from jjw:

At Rithmic, we are trying to close the gap between retail trading and institutional HFT. We have recently released features in our software that will enable a trader to get market data and release an order based upon that market data in under 250 microseconds (assuming that the trader's analysis that results in the order release does not take too long).

Until recently a trader would have to put up a lot of capital to be able to get into this game (or be part of a group with such capital). Now, if one has c++ programming capabilities, one can trade pretty much as fast as anyone, without a massive capital budget.

This is very new and will continue to evolve, but it is exciting to be able to offer technology that really levels the trading playing field.

I have never used your software, and it could be top notch, most likely is - But you cannot account for the latency in the network of the retail user.

This is where we lose.

Interesting if you could post some results that could back this up unless you are still in testing phase.

Thanks

 

[garchbrooks]

Quote from Algo_Design_Kid:

I have never used your software, and it could be top notch, most likely is - But you cannot account for the latency in the network of the retail user.

This is where we lose.

Interesting if you could post some results that could back this up unless you are still in testing phase. I would be personally interested.

Thanks

Also interested in their costs. It says you can deploy your code onto their colocated servers.

I'm sure it's expensive, but I'd be interested in them as an upgrade path, or perhaps doing like what some groups that use Genesis do and pool their resources and get the bots deployed to a colocated server or colocated virtual machine.

 

[Algo_Design_Kid]

Quote from garchbrooks:

Also interested in their costs. It says you can deploy your code onto their colocated servers.

I'm sure it's expensive, but I'd be interested in them as an upgrade path, or perhaps doing like what some groups that use Genesis do and pool their resources and get the bots deployed to a colocated server or colocated virtual machine.

If you had a truly viable strategy, no matter how honest the other party was, would you throw it on their servers?

No offense to sponsor, or whomever you are speaking about (didn't think you could do that with them), but I wouldn't trust my dog with a viable strategy - if I had a dog even.

 

[garchbrooks]

Quote from Algo_Design_Kid:

If you had a truly viable strategy, no matter how honest the other party was, would you throw it on their servers?

No offense to sponsor, or whomever you are speaking about (didn't think you could do that with them), but I wouldn't trust my dog with a viable strategy - if I had a dog even.

The code on the servers doesn't matter, because my code could be cryptic or even shitty. The part that matters is the fills that I get and where they store them. If that's the risk, then we have that risk from people working at the CME, the clearing firms, and so on and so forth. And from what I've seen, people who have any quantitative trading ability don't tend to work in the back offices, -and- the backoffice guys don't get paid well to begin with.

 

[Algo_Design_Kid]

Quote from garchbrooks:

The code on the servers doesn't matter, because my code could be cryptic or even shitty. The part that matters is the fills that I get and where they store them. If that's the risk, then we have that risk from people working at the CME, the clearing firms, and so on and so forth. And from what I've seen, people who have any quantitative trading ability don't tend to work in the back offices, -and- the backoffice guys don't get paid well to begin with.

There is a thread on here refuting your first comment, located somewhere on this site.

I would be weary, but speed shouldn't matter to you in this manner because IMO you will just end up getting ripped off/not understanding your true competition in the end.

Hopefully you prove me wrong, would like to hear the success story!

 

[garchbrooks]

Quote from Algo_Design_Kid:

There is a thread on here refuting your first comment, located somewhere on this site.

I would be weary, but speed shouldn't matter to you in this manner because IMO you will just end up getting ripped off/not understanding your true competition in the end.

Hopefully you prove me wrong, would like to hear the success story!

I don't doubt your comments at all, not one bit -- at least regarding speed.

But on the topic of security, the reality is that if you run, say, a firm in Chicago doing HFT, your employees end up interviewing everywhere else and people know stuff about other people's stuff everywhere else. The time you spend worrying about your binaries (not even source) on some other guy's server should be spent on internal employee security policies anyway. And, to top that off, some retail bozo making money independently would probably draw less attention than some firm that's doing serious volume and making serious profits.

I'd be less worried about someone hacking my server and taking my binaries because there's no way they can reverse engineer the binary and get all the changing inputs on a day to day basis without serious effort, and that serious effort would probably be better spent researching their own edge.

I believe the guy who says you have to remove, but at the same time, unless the edge prices in a slippage of 5-15 cents or so, I fail to see how retail traders can avoid getting victimized in the short run. HFT participants force the retail guy to withstand more risk, but I don't think the reward necessarily goes up. So, if you were doing like some kind of mean-variance study, the profile would look a lot worse for retail than HFT. Yet, it feels like ET has the same number of successful retail trader posters as it did in 2004, and it's a mystery to me how they are surviving.

In 2004, I remember seeing the markets price new information in such a manner that a retail trader could see the price adjust and move slowly. Now the MMs burst up and down, like before, but all the information gets compressed into much smaller "bars", to the point where some guy using "NinjaTrader" (or whatever) gets a faulty analysis without some really sobering assumptions worked into the model -- massive slippage, crazy commissions, and market orders filled nowhere near where a signal is generated.

 

[Algo_Design_Kid]

Quote from garchbrooks:

I don't doubt your comments at all, not one bit -- at least regarding speed.

But on the topic of security, the reality is that if you run, say, a firm in Chicago doing HFT, your employees end up interviewing everywhere else and people know stuff about other people's stuff everywhere else. The time you spend worrying about your binaries (not even source) on some other guy's server should be spent on internal employee security policies anyway. And, to top that off, some retail bozo making money independently would probably draw less attention than some firm that's doing serious volume and making serious profits.

I'd be less worried about someone hacking my server and taking my binaries because there's no way they can reverse engineer the binary and get all the changing inputs on a day to day basis without serious effort, and that serious effort would probably be better spent researching their own edge.

I believe the guy who says you have to remove, but at the same time, unless the edge prices in a slippage of 5-15 cents or so, I fail to see how retail traders can avoid getting victimized in the short run. HFT participants force the retail guy to withstand more risk, but I don't think the reward necessarily goes up. So, if you were doing like some kind of mean-variance study, the profile would look a lot worse for retail than HFT. Yet, it feels like ET has the same number of successful retail trader posters as it did in 2004, and it's a mystery to me how they are surviving.

In 2004, I remember seeing the markets price new information in such a manner that a retail trader could see the price adjust and move slowly. Now the MMs burst up and down, like before, but all the information gets compressed into much smaller "bars", to the point where some guy using "NinjaTrader" (or whatever) gets a faulty analysis without some really sobering assumptions worked into the model -- massive slippage, crazy commissions, and market orders filled nowhere near where a signal is generated.

Well it is hard to demonstrate the speed difference from 04-10, I believe liquidity providers make the market more efficient for us if your timeframe isn't less than 1 minute.

Without the added liquidity we are talking increased spreads. It would be hard to deduce IMO the real difference between the two unless you saw all participants and all cancelled and actual trades.

Too much of a brain-**** for me anyway, you should stick to what you know (or thought you knew after the market takes it )