r u guys kidding me? F#kn clueless. "Let me out smart them and scalp them a nickel by tricking them." LOL. By the time ur finger just figured that out they just put in 10,000 quotes.
Sorry for the REALLY late response on this thread! I just saw the info etc. OK, I will give you a bit of insight from somebody who worked with the "stuff." This ain't (sic) no bit player! You have to have considerable resources to punch down the numbers they are talking here! Even if you skip risk management...a decent number for (bids or asks)/server/second is say 300 (I should say - per FIX connection)! These guys are talking about sending down down THOUSANDS per second AND getting those orders accepted by SOME exchange - this is NOT a quick thing. You can see some serious data come down OPRA lines...but not FIX lines like they are talking here...this is somebody with substantial machinery and pipes! Think top 5 to 10 player! Just my $0.02 -gastropod Once again - if you need a market data administrator / FIX admin (Linux focused) let me know
the sawtooth pattern was detected here, at bottom of page; http://www.elitetrader.com/vb/showthread.php?threadid=83604&perpage=6&pagenumber=701
I've been thinking about this since the article came out. As g-pod said, the rates exhibited in this data scream wideband access by parallel means. The sheer muscle it represents makes me think about the reason(s) for these out-of-the-money bursty manifestations. The total rates would depend on how well the discreet parallel elements were synchronized to achieve the serial result. There is geometric/arithmetic progression here so this isn't the result of massive shotgunned unsynchronized quotes fired simultaneously. The rates an aggregate serial process could achieve might vary significantly diurnally as well as intra-day. If aggregate rate was important to my UHF strategy I might want to characterize just how much serial muscle I actually had on a given day or perhaps hour. These might be nothing but simple loopback tests to characterize/validate a massively parallel access schema. It reminds me of BER testing. Pretty damned interesting. I've got this rubidium-standard based nanosecond timetagger at work.... lol
Maybe everyone's trades are delayed randomly by 5 - 10 seconds seconds. That would perhaps take away the advantage the HFT have. I am sure there are some here who would whine and disagree, but these institutions are not helping the overall market. I see little value added by subsecond scalpers, and nothing but disadvantages piled onto smaller and home traders.
Here's my take: I think the intention is the pollute the limit book, and move the VMA of the bid/ask away from a true mean. This will trick other algos into buying/selling based on the manipulation.
the best trick is blasting the book w iso's. That really boggles the algos for the chase at critical areas.
Ya know, I thought about that idea when we had the "flash crash" and wondered whether we would all be better off if the exchanges built in say a minimum 200-500 millisecond delay on every limit order thats away from the inside bid and offer. That ought to have a great leveling effect, or would it??? The exchanges obviously benefit from these co-located computers -- there is a big monthly charge to hook up -- but they might profit even more from a level playing field if it enticed more traders to participate. Regardless I don't worry about it too much because I do OK as things stand now. These Crazies with the super computers piped directly to the mother lode don't seem to hurt me.