I am building a system based on executed trades. Sometimes I see trades that are executed way outside the inside bid and offer. I was told that some of these trades could be so called "Late reports" of past trades. Is there a standard of how trades are reported, is there a way to distinguish these trades? With REG NMS, is it possible to have a trade outside current BID and ASk? When the trade is reported, is there a special way these trades are reported?
Trades can occur outside of the bid ask for lots of reasons, sometimes in violation of Reg NMS, and sometimes without violating Reg NMS. Reg NMS has lots of exceptions.
Maybe these are real trades that "walked the book". That means the trade was for a larger size than the bid or ask size and executed against the next limit order out. Have you compared the trade size to the bid/ask size to check?
Also sometimes trades are reported slightly slower than quotes. I am talking about normal conditions, not just late reported trades. It used to be a standard rule of thumb with NYSE data to match trades with the quotes from 5 seconds earlier to get around this.
Let me ask you one question. What made you think you had a print outside of the quotation? NYSE has separate feeds for trades and quotes, and it's quite tough task to bind stuff together. Also (as mentioned earlier), NYSE execution system is quite slow, so trades are reported sometimes 2-3 seconds later. Late prints have special trade conditions in the feed so you can easily destinguish them.
i believe late reported trades are maked with a T on the time and sales, right? i know i have a list of all trade markers at work. it's funny i see them all the time and i can't remember that damn letter for the late reported trades. i know almost all of them are NAQS prints too.
Actually... There are lots of big "outliers" in high/low/last price data... Especially in low volume stocks... That you did not see a year or two ago. Listed trades happen at ridiculous prices on a regular basis... Dollars away from the "market". It must some kind of collusion... Where erroneously entered trades are printed, like, $10 away from the market... And the colluding parties split the take. It could be almost anything. Quality data has to be stripped of "outliers"... Because they just distort everything... And cannot be exploited by a retail trader.
That is exactly what i see...like prints 60 cents or dollar away from LAST trade. If you have an automatic stop/loss, that print could take you out, if you do not specify that the print was errenous. This does not often that much...but it seems to happen most when it matters like around the open or during volatile periods where probably there is a surge in data. When watching tha data feed..it is easy to distinguish them...but i wonder how a computer program would do it. I was suggested to use a PRICE offset where if a print is X% away from the last print...just filterout. I still think it is a problematic and tricky issue.
I can see someone trying to un/load a large position sacrificing a few quotes so the B/A does not get affected.