Corey, Nice posts. (or I should say that I agree with you regarding your assess towards Ash) lolatency and asiaprop, I don't know the details of what you guys do but considering that all 3 of us has exposure to institutional trading, I think we can all agree that the retail side does not have an edge over the institutional side in terms of market access like account setups ( commish and leverage), IT like latency, and other exchange side perks. ("Legally", order flow access is not allowed....) The key seems be a matter of doing what the institutional side won't bother doing. Few things would be: 1. Eliminating quant. desk operation by automating them by using generalized set of rules. 2. Reverse-engineering / anticipating a typical model. If you're talking about Sell-side there's a list of typical models like VWAP, TWAP, Arrival Price, Price React, Target Percentage of Volume, Volume Dispense, or onClose. You can anticipate when these models may trigger a major into the market. These days, the models are executed into the DarkPool so it becomes a bit tricky quantifying it to a point where it comes a clear cut quantified and confirmed "edge" but there are ways around it (which can't be done in a retail environment) Same can be done with Buy-side Market Making. The base concept of the models are the same all around, it's a matter of dealing with variations of the concept (underlying tendency). Like a double cross-Martingale system, you take "obvious" situations when a typical Mean Reversion models won't trade filtered by volatility and actually trading them. 3. Use of minor exchanges or trading platforms. Though, the liquidity is a major problem you can find arbitrage opportunities by trading against CFDs, UK/offshore betting sites and others. For a small trader, the available opportunities are more than enough to pay your bills or get your trading operation started. There's plenty of "platforms" that the retail traders can do which the institutions are not willing / legally unavailable for them. Yeah... I was planning on staying out of ET but this is just a good thread to keep out of... PS. Eric Chan is a joke.
In regards to pure High Frequency... 1. Rebate is somewhat accessible with BATS. They're capital requirement is a lot lower than the others. I have a prop. firm taking care of the clearing while I run the IT side of things with a help of a few friends. 2. My definition of a High Frequency model is a model that utilizes the B/A (Level 2, Market Depth, etc.) as the main source of data. The holding period varies considering that I monitor all products across the board. So I can be sending in 1000+ orders a minute, cancelling, scratching, replacing... etc.... but those filled orders may be open from 1 milli-second to a whole day. 3. Depth data OES/OMS/ and testing platforms are seperate. I only "store" data in memory for my OES/OMS, which is located remotely "in" the clearing firm. Historical data and other non-critical stuff are done in-office with me. My data storage is SQLServer 2005 and running tests using flat files. This has been my major headache, considering myself being C# programmer. I definitely need suggestions and advice for this.... (Other than learning C++ and using HDF5...)
Never say never. The India market gapped UP 17% earlier this year the day after their big election. Trading had to be halted after the first couple minutes. Here's the historical data from the India etf from yahoo finance (SYMBOL: INP): 20-May-09 47.82 48.64 47.34 47.45 1,237,400 47.45 19-May-09 47.95 48.27 47.50 47.51 1,934,900 47.51 18-May-09 49.21 49.41 47.51 49.03 2,543,500 49.03 15-May-09 39.75 40.00 39.20 39.47 616,800 39.47 14-May-09 38.53 39.13 38.19 38.87 300,700 38.87 13-May-09 38.65 38.80 37.50 37.82 745,000 37.82 12-May-09 39.58 39.74 38.54 39.16 804,200 39.16
Good post. I've heard that the exotic markets are being targeted and "raped" by quants these days. It's a 1-2 year old info. but I've heard that SAC (or DE Shaw) was opening up an office in Mumbai so that they can start exposing the Indian Market's inefficiencies, hence all the sell sides working hard to open up reliable DMA pipes to provide those markets. (And regionalizing existing ETS service like Algo orders). That's another list of things that the retail traders find it hard to gain access to.... (I don't trade exotics... and all I've mentioned in this post is 2nd hand info)
ummm... i don't know. that doesn't sound very general to me. kinda sounds like you're calling him out... questioning his analysis and suggesting he's not only a theif, but a sellfish one at that (which i thought was a nice touch). and maybe i don't read so good, but i coulda swore you're specifically discussing his EDGE. in fact, it's almost verbatim the argumentative tactic that other poster used on you... question ability, veiled insults, put on the defensive, try and pry out some info. btw, i didn't look up your posts. i have no interest in reading up on someone so weak as to bully a kid playing his first indy wad. i DID do some research on options arb last week and found that thread. after that last post of yours above, i thought, wait a minute, wasn't that the guy....? yup it was. i couldn't believe the same guy that was ducking and weaving against talking his edge, finally having to tell the guy to cut the shit and stop passively/agressive trying to get at it, would then turn around and do the same thing a couple months later. do unto others as they do unto you huh? lola's a good kid. quit being such a prick.
gann, i'm not sure if the headache is because you want speed. but if so, try binary flat files. it's pretty trivial... just write your in memory data structures direct to file. c# should be able to do this and will blow any rdms or string file out of the water. no parsing or db lookups. will put the bottleneck back at the cpu where it belongs.
I'm assuming that almost everyone on this thread thinks high frequency trading is better than long term trend following. This may well make sense if you are an institution with minimal per trade costs (I'm going with the assumption that microtrends are more 'predictable' than long term ones) but how on earth can you make it work with retail brokerage fees? When you're 'microtrading' you are capturing tiny price movements across those 50-100ms intervals. And this all assumes you are getting every trade right..
hey TsGannGalt, thanks for this informative post. You put forth couple ideas re high frequency and it was exactly what I was wondering about/looking for. Not that I will pursue it, I am not a high frequency guy, my edge is on the volatility side of things and I currently look into automating an option auto-trader based on nikkei index options, I am very happy Nikkei options traded in Osaka can now be settled in SGP as well. Also there are numerous arb opportunities between HK listed warrants and single stock listed options.... All the best for this week...
again there are lightyears between asking to share general info and pushing to reveal a specific strategy. I very clearly did the former. I got frustrated because LoLatency pointed to some other threads that had obviously (to him and me and all others) nothing at all to do with high frequency strats. LoLatency outed himself as someone who took others' codes and ideas without their permission, nothing I made up. Re Atticus's posts I never ducked, I simply showed him his limits when he thought he could outsmart a guy to give away his bread and butter. I think if you re-read what I added to Atticus's posts, what I wrote about China, and now here you will find out that most is based on logic and factual data, and I never made up stuff to bash anyone.
I guess you mean Ernest Chan, but I'd like to know what is your criticism of him? He is by no means my role model, but I have referred to his ideas in another post.