Looking for a partnership for AI algo.

Status
Not open for further replies.
robthequant said:
That's pretty cool. What did you use the t-90 for?

As far as drawdowns - my software minimizes underwater equity curve and maximizes profit per trade using multivariate genetic optimization - there's a great book on the subject ;) My equity curves on Bitcoin basically look like straight lines, there's very little draw-dawn in forward testing. Basically close to 0, it makes profit every day - you may think I'm crazy but it's true, if I saw this myself, I would think it's curvefit-bullshit, so I don't blame you if you think the same. But it validates on forward testing with real money, but there are drawdowns because of low quality executions, slippage and busy exchange server - not able to execute on their quotes during busy times. There are per second request limits etc...

You mentioned the "6 months" again :) This is not a "6 months" of corporate style work - I've worked for an investment bank in NYC, so I know how that is. I'm an entrepreneur and this is 6 months of 12 hours hard coding and hard study each day while loving what I'm doing... and including most weekends. One person can achieve amazing results.

About es futures I don't know... when I looked, I saw there are fees that may kill this. I need to be close to 0 or be a market maker getting rebates for limit orders. That's why I want to explore the other side. I can't do this on retail.
More...
Don't cryptocurrencies have fees too but you were successfully trading them live, right?
 
destriero said:
MarkBrown is broke. Never owned a cray. He's a voyeur.
More...

upload_2018-10-4_17-38-16.jpeg
 

Attachments

  • upload_2018-10-4_17-38-16.png
    upload_2018-10-4_17-38-16.png
    25.8 KB · Views: 10
This is the future! Every atempt is worthy in my opinion. I think you should move into fast moving futures that people will pay attention to like the es or nq and work a little more at home before you can post results. Keep posting the news. As you know every one that thinks ahead is criticized!
 
robthequant said:
Yeah I can make limited money, because I'm very limited by my access to markets. If my technology is applied on a pro level with the right access, I there's a huge potential. I just need to find a person who can see this, but I'm thinking this forum is not right for what I'm looking for.

I can go short, however there's an interest the broker charges. They say it's per day, BUT they actually charge it per trade! So if I do 1 trade that last 10 seconds, I'm hit with the FULL daily interest rate for that trade. Since I'm collecting about $1 per trade, shorts are not profitable. I would need to borrow bitcoins and pay daily interest rate per day, not per trade. That's not possible with the exchange I'm using.



That's because the system does very quick trades when the price is rising in the downtrend.



if / then statements, rules created by humans are very limited. Trading volume by bots is rising and more and more noise and randomness is introduced. Hedge funds are closing down because they can't extract alpha any more. Only those who are thinking outside of the "moving average/if then rules" will survive.

While you as a human can back-test and evaluate about 2-5 rules per day, my technology evaluates millions of "rules" per hour on a 96 CPU machine and can genetically evolve simple brain structures with neurons and synapses.
More...
Very impressive, how does it predict when to go long? What is it looking at?
 
robthequant said:
I'm doing my tests in the cypto markets because L2 data feeds and API trading access is readily available. I’m seeing some great results when I apply my genetically evolved nets on the order book and order flow. However order execution on crypto is bad, exchange servers can’t handle high load periods, and slippage and high fees are killing it. I
More...
I'm curious why you can't apply your strategy to equities. The market is fragmented and hft infrastructure more available. When I looked at trading Bitcoin, as a retailer, the fees were 2% of notional value and avg sizes were very very small (like .00001 btc or something). Some exchanges published book snapshots only. I can see why you would want to collaborate with hft firms ... I have a suspicion they don't use websockets to connect to crypto exchanges like the retailers do. And that is probably the reason they won't need any AI to make money ;)
 
brucelevy said:
Very impressive, how does it predict when to go long? What is it looking at?
More...

It's looking at order book and order flow in and out the order book, it measures 50 different features of the order book and feeds them to a neural net. It doesn't look at price history at all.

Seems like there's a huge gap between RETAIL crowd and the market making and HFT firms. Retail people are looking at the price history trying to find patterns in the past, they put indicators and draw lines on chart because that's what brokers and various trading "gurus" who can't make money trading teach them to do... But that's like driving a car forward only using the rear view mirror.

MM & HFT firms employ the brightest software engineers and scientist to apply the latest machine learning technology and AI to order flow and news and various other data sources. They are looking to the future rather than past. Retail people do not know anything about this hidden world, because it's a hard science and statistics that regular folks don't understand, and are stuck with charts and indicators because that's easy.

I'm looking for a person who can understand what I'm doing, but obviously this forum is close to 100% retail people.
 
Status
Not open for further replies.
Back
Top