Quote from trom:
This is one of those areas where the people that know how to do it are not going to share it with you.
Bingo.
Best I've ever seen, was a day where NSDQ declared self-help against EDGX and most firms couldn't see EDGX quotes, including mine, and had about a 15 minutes window where any stock with a live stale EDGEX order was getting traded through. I believe he made about $500 arbing (lifting the edgex offer, hitting the bid on NY or other ECNs 5-10 cents higher instantly). After that, EDGX operated normally and other ecns started routing to it.
I believe the Hamilton Swift branch found a great gig a few years ago where they were doing some darkpool manipulation/arbing, but that is long since dead.
Here's the only idea I can think of (this is me being nice and sharing what I'm pulling out of my ass)... reg NMS only requires orders get routed to the inside market, after that whichever venue you send your order to, you're sweeping that venue's book.
People do send really stupid panic limit and market orders that only sweep one venue's book. If you can develop an algorithm that separates with some degree of predictive value BS orders deep in the book from legit stale size (I think timestamp analysis should do the trick), perhaps when that stale size is 2SDs away on the bid, you could put a Bid in a different book behind it and hope to get filled by a freak occurrence sweep. At the beginning of the Hybrid, this would have been a killer strategy. Now, most likely, unless you're really clever the stale size will pull, and even when it doesn't, there are a lot fewer retardo market orders that only sweep one book than there used to be. One problem is that all the other bots trying to do this shite (lots of them) add a clutter to the book, but again timestamp analysis should work for that. However, since every quant fund and box writer and their mother wants to take advantage of some variant of this strategy (get filled on size away from the inside market on a 2-3SD inefficient booksweep), the margins may get killed. Maybe going the other way is where more money is - finding extremely, extremely high probability situations where a large reserve buyer or seller comes into a stock, and then sweeping all that liquidity that's enveloping the market trying to benefit from mean reversion... where your edge is basically that those boxes aren't aware of when that specific buyer/seller comes in, thus they don't realize it is non-advantageous to provide liquidity against the direction of the main player in that stock, even if it's 2SD's from the inside market.
What it comes down to is you have to find your own edge. From your post, it sounds not like you have one.
