At NASD or ARCX level, they still have lots of one-sided orders? I thought that the trading is so intense that there shouldn't be lots of one-sided orders. Currently the number of one-sided messages is about 1/3 of total messages. I thought they could always aggregate two opposite orders into one message and send that message out, because it's so liquid. Otherwise, does <1% of cancellation make sense for QQQQ?
always is different than 1/3... either you're subscribed to one of those new 'quantum' feeds, or you're not making much sense. <1% cxls on the QQQQ's? come on man, at some point you have to use your brain to analyze the data vs just using it to produce data! talking to traders who are telling you what's what doesn't seem to be helping. so go learn some basics on market microstucutre or even better go place some trades and see what happens! from there you should be able to tell what's what. good luck.
Depends on how you define cancel. The OP evidently defines differently from you (although I prefer your definition). With Nasdaq Level II, the top of each MMID is all you get. They could have 1000 bids below their best and only one shows. That's why they released TotalView...
No, my definition is the same as yours. Remember we are down to the atomic level. Those high level quote types will ultimately be translated into three atomic order types: market order, limit order and cancellation order. That's all. All the higher level fancy order types are a combination of these atomic order types. We are talking about low level order book data stream with arrivals/update messages.
In that case, you're going to have a hard time measuring cancels using Level II data. Or to put it another way, you'll be measuring the number of per-MMID-top-of-book cancels (since Level II only shows MMID top-of-book). You would need a full book feed such as TotalView to see atomic operations to depth beyond per-MMID-BBO. FYI, all this info is avail from the source at nasdaqtrader.com -- provided, of course, you're willing to spend a lot of time sifting through reams of documentation .
yes you're right, technically that's true. i suppose i'm being a bit 'legacy' in my nomenclature, but it's one of those marketing terms that have managed to stick around (ala xerox). for instance, my feed delivers full book for all exchanges, but rather than use the trade names (arcaBook, totalView, etc), 'L2' or just 'the book' are the goto jargon of choice when dialoguing with traders/programmers. more than a few data feed api's cement this as well, allowing subscription to the gamut of book feeds, but assemble them all under "level2" base classes and windows. i assumed, perhaps incorrectly, from those initial cancelation ratios he posted he meant full book when referencing level 2. anyway, 'level2' (heh) debugging is tough enough to do with a team of guys in an office, remotely is making me batty. take my advice mizhale, put some orders out with cxl's and watch how they affect the book, a very easy way to corroborate if your algo assumptions are correct.
Well I have all the order/quotes arrival messages that allow me to reconstruct the full orderbook. It's just the documentation wasn't clear about what messages denote the cancellation orders. That's why we made two assumptions about the identification of the cancellation order message. They result in two different cancellation ratio for QQQQ, one has a cancellation ratio of about 1/7 to 1/8. The other has a ratio of about <1%. I would like to know which one makes more sense?