lol - think youll find that most of these boxes are only doing 1 lots (futures). they all add up you know!!! grob has spoken before about the importance of getting your line off in blocks of the average size you see printing. last thing you want to do is get your 6000 lot position off in 1-3 trades - people will see you coming a mile off and youll loose the element of surprise. folk here have already talked about spoofing, bash the rat, and other fun & games being played - but who is playing, and who is the patsy? if you dont know, then its..... i also remarked about this in my post above where the idea of these things is efficient execution - the speculative part is still left to the human. the computer is just a labor saving devise remember (yet every one thinks they are getting creamed by an inanimate object with no conscience or power of original thought! what is the world coming too?)
apart from watching myself get pennied automatically, I can only contribute my story about "snakes." a fellow once told me that apps that look for mispriced orders float around and quickly execute against them. Seems kinda cool.
puffy - whack-a-mole or bash the rat.... either way, the object is to get one over on smaller mammals/less agile traders than our selves.
anyone identified any logic behind the randomness? Let's get the ideas flowing and hopefully find an edge that we can exploit! This was the original idea of the post....and i konw there are a lot of DOOMSDAY, PHD types that dont think its possible for anyone to make money out of anything ..sorry to ruin ur day..there are people..and they are makin a lot of money if you just wanna say (i dont believe, impossible, etc etc) please...go post elsewhere... i agree with the original Poster...if we can get some ideas flowing i would love to test them manually and live (not by hypothesis or cross flow mumbo jumbo) if you read my earlier posts,,and some guys have seen the same stuff i have ..there is a way to expoit these markets algorightmically someone earlier spoke about size...from one of the program guys i have talked to the Optimum order entry level is 3-400 shares believe it or not. they will have 1-2000 positions of small size makin them nimble and the slippage not as severe i also know of another box that adds to winners and elimantes losers (who would have thought)..how exaclty they do it ..i ahve no idea and what stocks, criteria.,.. i have no idea (i use this in my trading...it works..but not to this guys extent) i have heard another one exists that buys/sells Quick spikes on stocks trading above volume.... on volume sometimes you'll see a stock like rimm trading its high..then in 1 sec it will go down 15 and back up 15 c...no idea the criteria they are using..but im sure they are takin offers as turns back or biddin on way down (i guess a stock on high with high volume is lot less risky to buy a random dip ) on NYSE you'll notive in certain stocks ECNS are PEGGED and i mean PEGGED to a couple Cents below NYSE on bid and offer...obviously every stock is diff in volume, spread , etc...but that is one i have noticed for use (this game will be done in april though with new NYSE rules) its interesting to see Goldman getting out of Man based trading and moving more and more to fully computrized even on options.........seems liek the biggest and baddest on wall st would be doin that for a reason please...only positivity..i'd love to hear more constructive ideas what patterns u see in ur trading , I also apprecaite the poster who mentioned that his success/ failures with longer term vs shorter term boxes I have a longer term box i went live with 4 months ago and totally see the fluctuations ... sorry for rambling d
"randomness" pretty much assures that you can make money trading, as things trend between the bounds. you just have to pick your timeframe, and make sure you slippage doesn't exceed volatility.
This document approaches both the trade duration and the volume factors, in addition to a few others... http://www.uwasa.fi/~sjp/Conferences/fss2001/papers/Engle/Engle1.ppt
Well, I quickly read the paper and there is nothing viable in it related to building a system, thats for sure. It seems to be based on the premise that somehow certain informative type trades themselves are responsible for price changes but only simplistic models are sampled. Nothing approaching the complexity of an actual trading systems could be modeled this way ( unless of course you model the system as signal events, the actualization of both event (price, volume, etc) with a matching schema (your favorite indicator or your stop). Its akin to saying, well we see the effect but we can't identify the cause ( or a good entry / exit ) , so we'll just generate some results that reflect on heighty hypotheses about how the market works.
The idea is that the trader who hits your quote either detains directional information on the asset or not. If he does, you lose money. If he does not, prices revert to their previous state, and you gain money. The point is to trade on noise, and to get out of the market early when information shocks hit the market and are assimilated into prices. This powerpoint presents methods for predicting those information shocks, measured as increases in volatility. Ultra high-frequency GARCH and ACD models are easily implemented and calibrated to market for use in algorithmic trading (using matlab or similar software). The paper "Econometrics of ultra-high frequency data", on p. 30, shows the high statistical significance of inverse duration of last trade to explain the conditional variance of the next price movement (see z-stat of 1/dur impact on variance). Last paper demonstrates similar results, with different methodology. Engle is a nobel prize laureate (unlike grob109). I think we can trust his results. Institutionals already use this. Maybe so should you, before thinking of beating the market. (This is a constructive critic only). Volumes and durations in a ACD/ACM model will be much more helpful to you in identifying information risk, then any signal event you perceive as a predictor of when to exit. These will be my last words on the subject. If people are interested in continuing this in private, you can send me a private message so we can exchange emails. Good luck in your research and hard work! Your friend, cosine
Well, anything by Engle I take quite seriously. Heh. The presentation is fairly well written (but a bit sparse, I am looking up their actual journal papers now), I am actually thinking about applying their modelling technique into empirical observations, I have already implemented a variant of ACM in my system, but their presented modelling of noise versus duration is interesting, and may get me to make more "informed" strategies, since "informed about other informed party" is an important modeling factor.