CYP Trading - post Fuji validation, page-168

"I can't be bothered about the manipulator theory, waste of time agree,..."

I mostly agree with this observation Sector but I think you've seen the youtube video Flash Boys (there's a book with the same name) that describes how high frequency trades and close proximity to the exchanges confer a real advantage based on pretty much the laws of physics (how soon a message can go over a connection) and based on computers seeing small time differences of milliseconds as still being plenty of time to react when humans can't think that fast.

Without being a conspiracy theory guy as I understand the term is usually meant - no tin foyle hats for me - I am trying to be mindful that computer technology particularly in areas like deep learning (a sort of AI, a machine learning) might be used in algorithmic trading to good effect at some stage. The share market is just a natural area to try some of the newer stuff it seems to me as there is an opportunity for a financial payoff.

Without reaching any conclusions yet I've been thinking about a couple of facts.

1) Most retail has to pay brokerage and the brokerage is a bigger or smaller percentage of their trades at certain parcel sizes.

2) Brokers (as members of the exchanges) don't have to pay brokerage - so they can put up irritatingly small trades that wouldn't be marketable parcels (wouldn't amount to $500). Sometimes the bot trades might be below a single dollar. All legal apparently despite being annoying especially to retail - see point 1.

So I got to thinking could an algorithm (perhaps by a broker bank like say Commsec) be written that would take advantage of their free trades and potentially high frequency trading power as well to try and extract more brokerage fees from the retail brokerage payers by drawing the trades away from the low brokerage payment points. For instance if a retail trader has to pay $10 for a trade up to $1000 on Commsec the retail trader should want their trades to be closer to $1000 all else being equal that say $500 because on the $500 they pay a greater percentage in brokerage.
(Which of course would be what Commsec would like).

So I figure Commsec (for instance) would have an incentive to try and encourage trades further away from the low commission points. Perhaps where machine learning or deep learning algorithms could be deployed would be in discovering how far away from the low commission optimums retail traders could be drawn if they were queue jumped by high frequency trades when they were holding out for low commission paying trades.

Not sure if that makes sense. But when I see lots of trades on a relatively high number of trades day in CYP that are below a marketable parcel I get to wondering about the state of the art in deep learning and algorithmic trading. Like a herbivore perhaps gets to wondering if those sleepy carnivores are just faking it perhaps.