Congrats, you're onto something. Too bad you're a student and you'll probably never get a shot at testing your hypothesis
it is hopeless and any retail who tries to compete with hft is just dumb. even if you learn something if you cant make money from it then what good is it in the financial markets. these are selfish mkts they help no one emotionally or with their health. there is no room left in the hft space. THE FIRST THING THAT MUST BE DONE ON ANY TRUE PROPER STUDY OF HFT WOULD BE OF THE TRADE ENGINE AND ITS FILLING ALGOS IN THE MKT YOU ARE TRADING. THEN THE SPEEDS OF THE TRADE ENGINE AND THE ELECTRONIC JITTER AND WHICH GATEWAYS TO USE. THEN YOU WOULD NEED TO LEARN ABOUT COLOCATION AND HOW MUCH SLOWER YOUR HARDWARE IS COMPARED TO YOUR COMPETITIONS THEN YOU NEED TO DO THE SAME IN THE CASH MARKET OR ANY OTHER MARKETS OF THE SAME CONTRACT. LIKE SOYBEANS IN BRAZIL OR OTHER EXCHANGES THAT INCLUDE USA STOCKS AS ADRS. YOU MUST ALSO INCLUDE CURRENCY FKUCTUATIONS AND INTEREST RATES. THIS IS ALL BEFORE EVEN STARTING TO LEARN HOW TO OR WHEN TO GET PRIORITY ON THE CLOB. MANY HFT NOEMW ARE AGGRESSIVE AND ONLY TAKE LIQUIDITY. THIS IS JUST THE START. IT TAKES 20 TO 50 MILLION A YEAR IN TECHNOLOGY UPGRADES AND TRUE DMA FEES N COLOCATION JUST TO SHOW UP AT THE RACE.
if you are so good at math then break down 1 second in time into milli micro nano pico and then equate it to days and show us how many weeks months years of time they have to calculate and cancel or add liquidity compared to even 1 or 2 mouse clicks. even right now what is the conversion time of your trades from Mit to mkt. Or how long for a limit order to be risk checked. etc. this is futile.
Certainly I will test it in practice (it's the most important section in a paper for me). I've traded using NinjaTrader, Thinkorswin, MT4, MT5, Tradestation, "TastyPlatform" (lol) and a dozen more outside the US. Now I know just what I need and I'll probably use some nice open source code base, like QuantLib, and maybe I will hire programmers to help me in boring parts, like routing protocols.
Well, at least you are starting with SOMETHING. When Jim Simons https://en.wikipedia.org/wiki/Jim_Simons_(mathematician) of Renaissance fame started building automated trading models, they had nothing and ended up writing over a million lines of C++ code. Keep in mind that you are tackling an enormous project of building a customized trading platform. I, for one, am a user of Multicharts which is written in C++ and is very efficient. In fact I know of one HFT firm that uses it successfully in a co-located server environment. Instead of a million lines of code, their trading model was written in a couple hundred lines of Powerlanguage.
Unfortunately, there're just a few (retail) trading platforms where I live. It's a shame. Firms try to "import" platforms like NinjaTrader, MT5 etc, but they fail to integrate them properly (stability issues) with the local infrastructure. But I see that as an opportunity. If I have the chance, maybe I'll try a more robust solution (=spend more money).
There is a guy at AQR who was the Wall Street jock on this stuff and has extensively published his work - Kukanov. May still be around, but search his work.
Dude - this may possibly be one of the most incoherent posts this year in ET. Do you want to clean this up ? How about some links ? Who's AQR ? Who's Kukanov ?
This is gold! His PhD thesis is the kind of work I'm looking for. The subscription fee of this forum is finally paying off!
Phew... I always think I'm missing something... some slang, something I should know. I had to google everything.