Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Neural Net Side Project Makes $3500/mo Trading Stocks

indiehackers.com

Post reply on HN
Print view View original

71–80 of 122 posts

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#71

Post by creeble »

tunesmith wrote: post #58
brianwawok wrote: post #27

Earlier quoted context omitted.

Backtesting does not mean much. It's easy to make an algo do well yesterday. Very hard to make it do well tomorrow.

Does backtesting work better if you, say, start with seven years of data, use the first five to judge the strategy, and then test it against the last two? Is that a good way to avoid fitting the data?

It very much "depends". I'll give you an example.

A friend had a technique that worked very well on the NASDAQ 100 with roughly those BT criteria. One problem was overlooked: The NASDAQ 100 changes over that time. So you aren't necessarily investing in the same securities in year 6 as you were in year 5, even though you're investing in the same "top 100". IOW, backtesting in this particular case had future knowledge that wasn't easy to see.

When he started (paper) trading in the real NASDAQ 100, the returns were negative. In that he was lucky: They could have been positive for some time, until the constituent stocks changed.

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#72

Post by brilliantcode »

alva wrote: post #52

Reposting my comment on the first HN submission of this page "The claims made in this interview are extremely suspect, it just not make sense. Absolutely no relevant details are included. The developer claims he was able to built an AI trading strategy that is profitable 95% of the time. No technical details about the strategy or platform for trading is provided. A few trading buzzwords thrown in a few places. The re…

You fucking nailed it.

Articles like this where proper due diligence verifying the authenticity of the poster isn't in place raises strong questions about how reliable and factual IndieHacker is.

I really feel like they fucked up by posting this article. The other article seemed alright but now I ask the same questions.

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#73

Post by csallen »

minimaxir wrote: post #56

Votes on this submission may be manipulated with an attempt to bypass the voting ring detector via linking to /newest: http://i.imgur.com/08pAFOw.jpg And another one by Indiehackers founder csallen after it already hit the front page: http://i.imgur.com/rxWtsWJ.jpg

As far as I'm aware, it's perfectly fine to link to discussion on HN, but not okay to explicitly ask for upvotes: https://news.ycombinator.com/newsfaq.html

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#74

Post by sebyddd »

svantana wrote: post #59

I may be way to cynical for my own good, but there's no proof that this is actually real, right? Besides a couple of pretty bizarre screenshots [1]. I'm skeptical because a) these types of gains in public markets are pretty much unheard of, and b) faking a story like this would be a fairly easy way of getting lots of attention for your freelancer business (which this guy advertises right at the top of the article). […

Appreciate your point of view. First of all, there seems to be a lot of skepticism around this project and I found that surprising. You're saying that if you would get incredible returns from trading you would keep quiet. And therefore you are contradicting yourself by accusing my lack of transparency.

As I was invited to share my story on Indie Hackers, my goal was to provide as much insights into the project, without blowing up the advantage it gives me over a very niche market (NSE). I have absolutely no intention to 'advertise my freelancer business' or 'sell my tech', as someone mentioned. My goal was strictly to send an encouragement message to people thinking of/building something similar, by sharing as much as I can of my project. Faking stories online would be the last thing on my agenda and I find it somehow amusing that some people believe that.

PS: Having to cat a lot of text + csv files while developing this, it was much easier to alias my cat function to use csvtool on csv files and it's regular implementation on other files

PS2: The 'minimalist, ultra-stylish AWS window' is a quick app I cranked in a few hours to see the status of my instances. Not sure what's suspicious here. The window or the the fact that it's stylish?

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#75

Post by csallen »

brilliantcode wrote: post #68
svantana wrote: post #59

I may be way to cynical for my own good, but there's no proof that this is actually real, right? Besides a couple of pretty bizarre screenshots [1]. I'm skeptical because a) these types of gains in public markets are pretty much unheard of, and b) faking a story like this would be a fairly easy way of getting lots of attention for your freelancer business (which this guy advertises right at the top of the article). […

I also felt like this article was very strange, almost too good to be true, not really discussing downside risks associated with each position. It's possible to generate a highly profitable and successful trading system but it's the expected probability factoring in the worst case downside scenarios that really matters. After all, you could win 9 out of 10 times but that 1 time is where you blow up. Right now, I'm pu…

> After all, you could win 9 out of 10 times but that 1 time is where you blow up.

In the article he discussed exactly this happening.

> I also question if it's really HFT, the bar is really high to get low latency connection but he's only making 3.5k/month...it just doesn't make sense.

Again, in the article he states that he abandoned his HFT approach.

Most of these negative comments seem to come from people involved in trading who didn't read the full article but are extrapolating based on their knowledge of finance.

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#76

Post by alva »

csallen wrote: post #67
ucaetano wrote: post #60

Earlier quoted context omitted.

Do you have a link to the original version? It looks fake enough as it is, I wonder how much bad it could be. But hey, maybe this guy really cracked the market, and a Nobel prize and billions of USDs are on the way...

The changes were to add details explaining the 95% "success rate", the sources of data, and the exchange that he was using. I think you guys are both creating a false dichotomy between "genius who cracked the market" and "liar who is fabricating data", when there is plenty of room in between, e.g. "someone who made a few lucky trades in a bull market over a short period of time," which would describe probably the vas…

Hey. As we spoke about this earlier today, could you please post the original before the edits post our conversation?

edit: I rang the bloody alarm bells about this submission with you. I hope I am wrong, but it looks like you have been had.

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#77

Post by minimaxir »

csallen wrote: post #73
minimaxir wrote: post #56

Votes on this submission may be manipulated with an attempt to bypass the voting ring detector via linking to /newest: http://i.imgur.com/08pAFOw.jpg And another one by Indiehackers founder csallen after it already hit the front page: http://i.imgur.com/rxWtsWJ.jpg

As far as I'm aware, it's perfectly fine to link to discussion on HN, but not okay to explicitly ask for upvotes: https://news.ycombinator.com/newsfaq.html

Then why did you include a picture of the submission in that tweet?

Also you ignored the first picture which clearly shows your account.

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#78

Post by brilliantcode »

hmate9 wrote: post #47
brianwawok wrote: post #27

Earlier quoted context omitted.

Backtesting does not mean much. It's easy to make an algo do well yesterday. Very hard to make it do well tomorrow.

I completely disagree. Backtesting matters, especially in HFT. What would you rather have: an algorithm that has performed well on 5 years worth of historical data or one that has performed well in the past 2 weeks.

He's not wrong, backtesting really is not the same as testing on real live data. You are fitting your models on data that has already been done, without any of the risks factored in. Even if you do find a system that picks every ups and downs in the backtest, it will do poorly with live trades because the volatility and risk is something that changes based on fundamental data.

However, in this case, the fact that no backtest or any sort of risk/reward ratios were published suggests high level of skepticism.

tl;dr: past prices can't predict future prices.

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#79

Post by csallen »

alva wrote: post #76
csallen wrote: post #67

Earlier quoted context omitted.

The changes were to add details explaining the 95% "success rate", the sources of data, and the exchange that he was using. I think you guys are both creating a false dichotomy between "genius who cracked the market" and "liar who is fabricating data", when there is plenty of room in between, e.g. "someone who made a few lucky trades in a bull market over a short period of time," which would describe probably the vas…

Hey. As we spoke about this earlier today, could you please post the original before the edits post our conversation? edit: I rang the bloody alarm bells about this submission with you. I hope I am wrong, but it looks like you have been had.

Sure, they were additions rather than edits:

Addition #1: "For proprietary reasons I will abstain from publicly discussing a lot of details about the technical implementation. Although I get many requests to open-source the project, I believe that disclosing deep details of the models or prediction approach would hurt the advantages that this solutions has over the other existing bots. However, for anyone willing to learn more about that, I would be more than happy to discuss in private, to some extent."

Addition #2: "...worked around 95% of the time. Now this is not by any means a reliable metric, and there are many factors that affect it. The bot has not been tested enough to guarantee that this isn't just a fluke (it might as well be). Large investment management companies would do anything to achieve those statistics, and I'm sure I won't keep up that amount of success in upcoming trades. The success so far was also greatly impacted by the favorable market conditions, chosen stocks, and the fact that the bot was running intermittently."

Addition #3: "For trading I recommend Kite, mainly for their stable Connect APIs and the low bandwidth. Their limitation is 3 requests per second, and this was more than enough for my new strategy. Getting solid historical financial data isn't cheap, and with so many people hitting the providers to scrape and download data, I don't blame them for limiting the offered information. Intrinio is a good provider for real-time stock quotes at very inexpensive prices. However, getting access to more in-depth data would always yield better results."

Re: Neural Net Side Project Makes $3500/mo Trading Stocks

#80

Post by alva »

sebyddd wrote: post #74
svantana wrote: post #59

I may be way to cynical for my own good, but there's no proof that this is actually real, right? Besides a couple of pretty bizarre screenshots [1]. I'm skeptical because a) these types of gains in public markets are pretty much unheard of, and b) faking a story like this would be a fairly easy way of getting lots of attention for your freelancer business (which this guy advertises right at the top of the article). […

Appreciate your point of view. First of all, there seems to be a lot of skepticism around this project and I found that surprising. You're saying that if you would get incredible returns from trading you would keep quiet. And therefore you are contradicting yourself by accusing my lack of transparency. As I was invited to share my story on Indie Hackers, my goal was to provide as much insights into the project, witho…

Post your current and historic positions, opened and closed and people might believe you. Currently you have offered absolutely no proof of your claims and the interview suggests you are extremely naive of trading.
Post reply on HN