Ask HN: Anyone designs financial engineering tools as a hobby?
I enjoy doing this stuff but it's tough to find others with similar interests since I moved to the boondocks many years ago although it's super easy to find people looking for financial engineering jobs.
A bit different from what most financial engineers are interested in, but I'm in the process of developing a linear programming model for the US tax code so you can do long term tax optimizations based on your financial profile and goals. It was a lot harder than I anticipated but got a PoC working recently. Ugly landing page at https://smarterpath.co
Yup good ol' wordpress with the Contact Form 7 addon for the early access sign-up. Would like to do a redesign at some point to make the value prop more obvious, but getting a public beta out has been the priority recently.
We created a tool used in niche fixed income asset-backed securities (which are heavily engineered). We crawl EDGAR for ABS-EE filings which are posted in XML, clean and standardize them, drop them into a data base and then allow users to create instant stratifications of the portfolio (rather than having to download the 80k row loan file and run pivot tables).
This allows someone buying the B tranche of FORDO 2017-B to see how many people in Florida are 30-60 days delinquent, what model car they drive, the LTV, etc.
I'm the odd one out. I've been writing automated trading software since 2007 as a hobby. However, I'm not looking for a job doing that. 1) It's profitable enough to do it on my own. 2) I have a gig right now doing time series analysis for a company that isn't financial, which imho is just as fun (if not more), but has a bit more potential to benefit the world. 3) When I get a job doing a hobby it has the potential to burn that hobby out. It's not about making money, but enjoying life, and being a bit competitive is fun.
PhD in signal processing, amateur algo trader & economist and software engineer by day (not in the finance field unfortunately). I want to live your life!
For me, I like doing the impossible. If no one has done it before, I don't need to get bogged down by books or stackoverflow. (And, you don't need a degree.) I have to figure out new ways to solve a problem. That's what makes programming fun and that is what makes this sort of stuff fun.
Like just the other day I figured out how to compress arm movement down to, so far, 4 variables (all floats atm), for every reasonably large change of movement. (Though, for small tremors while walking and other kinds of movements I'll need to output some more data.) My goal is to increase the accuracy of the signal beyond the initial noisy accelerometer data. The pain is going to be proving it through a high speed video camera or something else (Got any ideas? lol) because I'm designing this to be accurate to real life arm movement, not accurate to initial data. (If you're curious, an ellipsoid, plus an axis for like looking at a watch, and possibly a magic number or a large radius for a straight line.)
Does HN do PMs? If you want to hang out we probably have a lot in common. Though, I'm untrained and just kind of unusual.
Oh btw, I'm not sure if it is obvious, and not like I've read a book on the subject, but as far as economics goes I would think meta macro economics would help day trading. But what do I know? I'm just a rabbit.
I made some resource management tools that helped me in MMORPG's (ex EVE, WOW) and they had potential for real-life usage. There was also a episode in my life where i tried to predict random using simple neural nets in roulette, lottery and some stock markets. Fun thing to do but time-consuming.
Resource management tool was made before git when SVN was a thing so im more than sure that it doesn't exist anymore.
That neural net thing may still be alive, but buried deep on my workstation. I'll look for it and publish it on git eventualy.
There is also particle swarm optimization lib i made for it here:
https://github.com/CreoOne/ParticleSwarmOptimization
Well, you motivated me to actually get started with something like this - doesn't mean I can actually beat my procrastination, but if, you definetely played a big part in it. So thank you!
And if you go back and modernise it, I'm still interested in Github links :D
At one on the previous jobs I wrote a tool which takes a list of bonds/equities trades and ties buys and sells in order to count PnL. The idea was any sell should be connected with one or several buys/repos. Like, I sell 100 bonds , so I should take them from 5 buys in correct order (FIFO).
It seems easy to do this but there are tricky operations like REPOs when you have to 'borrow' papers from a reversal REPO to perform a sell or REPO operation. Then you should close this hole with a next papers arrival. I wanted to attach some web interface to visualize these crazy buys-sells-repos chains but at some moment I left that company.
I want to rewrite that tool as a side project but not sure if it would be helpful for anybody.
I also had a previous job doing exactly that... but it was an billing / accounting system at a print company.
Certain large customers (a bank, an airline) had contracts where they would order large batches of printed forms, luggage tickets, sick bags, etc, at some negotiated quantity / price (taking into account the size of the print run, our capacity at the time, etc.). We (the print company) would then print it and warehouse it, and ship it directly to their individual locations in smaller batches as they ordered it.
The billing was done on some cost + shipping basis, so at the end of the month you'd have to look at everything that was shipped, work out which original batches going into inventory that came from on a FIFO basis, and then use that as a basis for the cost plus billing.
It was a horrible messy old codebase, and it was sometimes difficult to explain the results.
Anyway, I just thought I'd share this as an example use-case for a sufficiently generic / flexible / reliable implementation of your potential side project.
It is the interesting idea to make this tool as generic as possible! Like a framework for matching different types of events based on some rules + a visual explanation for any chain of matched events. Perhaps I should think about this idea. Thanks!
Yeah, and the visual explaination part is actually just as important as getting the right results. For any complex system with financial impact there will always be people claiming that the output is wrong, so you need to be able to quickly and easily justify everything.
I am developing Crypto analytic tool. It's like the fund, analyze your portfolio, show potential gaps and currency there you can invest. So you need to make an investment decision, the platform only shows information and analytics.
It's more a portfolio tracker than a financial engineering tool but I have been thinking about adding some correlation / variance measurement to optimize portfolios, see the blog:
I’ve been writing algorithmic trading software and back testers as a hobby for a couple years now, and what surprises me most is how it is possible to have decent profits without mad financial skills. The trick is good quality data and a couple spare hours a week! A rooky mistake I have made is losing too many days trying to hack myself a good source of data, so I can only recommend paying a few bucks for a reliable one. You can definitely have a good time (while making money) with basic swing trading algorithms and a bunch of historical end of day data like https://www.gourgane.io
I've had a similar problem to find proper data for cryptocurrencies while I was writing my trading tool. It was quite some work to set up a proper infrastructure to get the data. Though eventually I turned that into a full service. You can check it out on https://coinograph.io
I have been using Questrade for a while (I’m from Canada) but they are closing their API so I will have to switch to Interactive Broker...
My algorithms are mainly swing/long term, so making a few trades a month. I like to trade and back test ETFs like SPY and SPXL since they have high volumes and are more easy to « predict ».
I started with 10k and it didn’t feel risky at all since it was post 2009 :)
I'm a software engineer and I developed trading algorithms just by playing around and visualizing data. Once you have historical data, technical indicators and a simple back tester, you can test some trading strategies and get feedback quickly. Graphing the output of your tests to analyze the annual return over time also helps a lot to design algorithms.
I have limited financial knowledge but I would recommend reading about technical analysis and technical indicators (This one is short and sweet https://www.investopedia.com/university/technical/) just so you can grasp the basics.
I don't generally, but I've been hacking away at the infrastructure for high frequency trading crypto currencies for a while.
I originally just wanted to do arbitrage trading, so I built a simple tool to find opportunities using pre-existing services.
But as I slowly learned more about it, I had to expand it into monitoring the exchanges in real time with orderbooks and past trades to get around annoying false positives in my original system.
It's pretty interesting, but I haven't had the time to work on it recently. Plus the increased efficiency of the markets has made arbitrage trading a lot less desirable in 2018.
I would be happy to discuss it with like minded people.
Nice. Seen anything interesting with your tool other than the amazon observation? You should update your resume with this, scatterstocks.com. The autocomplete feature isn't working with Firefox Quantum(60) on Windows 10.
I guess it's a hobby right now but I hope it's not eventually.
Can i just say that I wish there was some kind of repository of information for drawing charts. And there are so many different kinds of financial charts that's it's a bit intimidating.
As someone completely oblivious to this world, is it possible to get an outline of what strategies are being refined by the tools mentioned in this thread?
I was sort of forced into double entry accounting as a youngin, and have been meaning to do a) a local sqlite general ledger/accounting system and b) maybe extend that db beyond local, into Apple grade territory with FoundationDB.
Hit me up if you're a VC funded by the US with huge amounts of cash.
This allows someone buying the B tranche of FORDO 2017-B to see how many people in Florida are 30-60 days delinquent, what model car they drive, the LTV, etc.
100% free: https://finsight.com/product/us/abs/ee
Note: this is not exactly a "hobby" but we offer all of these services 100% for free without restriction so thought I would share regardless.
For me, I like doing the impossible. If no one has done it before, I don't need to get bogged down by books or stackoverflow. (And, you don't need a degree.) I have to figure out new ways to solve a problem. That's what makes programming fun and that is what makes this sort of stuff fun.
Like just the other day I figured out how to compress arm movement down to, so far, 4 variables (all floats atm), for every reasonably large change of movement. (Though, for small tremors while walking and other kinds of movements I'll need to output some more data.) My goal is to increase the accuracy of the signal beyond the initial noisy accelerometer data. The pain is going to be proving it through a high speed video camera or something else (Got any ideas? lol) because I'm designing this to be accurate to real life arm movement, not accurate to initial data. (If you're curious, an ellipsoid, plus an axis for like looking at a watch, and possibly a magic number or a large radius for a straight line.)
Does HN do PMs? If you want to hang out we probably have a lot in common. Though, I'm untrained and just kind of unusual.
Oh btw, I'm not sure if it is obvious, and not like I've read a book on the subject, but as far as economics goes I would think meta macro economics would help day trading. But what do I know? I'm just a rabbit.
And if you go back and modernise it, I'm still interested in Github links :D
It seems easy to do this but there are tricky operations like REPOs when you have to 'borrow' papers from a reversal REPO to perform a sell or REPO operation. Then you should close this hole with a next papers arrival. I wanted to attach some web interface to visualize these crazy buys-sells-repos chains but at some moment I left that company.
I want to rewrite that tool as a side project but not sure if it would be helpful for anybody.
Certain large customers (a bank, an airline) had contracts where they would order large batches of printed forms, luggage tickets, sick bags, etc, at some negotiated quantity / price (taking into account the size of the print run, our capacity at the time, etc.). We (the print company) would then print it and warehouse it, and ship it directly to their individual locations in smaller batches as they ordered it.
The billing was done on some cost + shipping basis, so at the end of the month you'd have to look at everything that was shipped, work out which original batches going into inventory that came from on a FIFO basis, and then use that as a basis for the cost plus billing.
It was a horrible messy old codebase, and it was sometimes difficult to explain the results.
Anyway, I just thought I'd share this as an example use-case for a sufficiently generic / flexible / reliable implementation of your potential side project.
[1] https://www.backtrader.com/
https://www.cryptoport.net
Where you trade? What sort of trades? How much capital to start doing something like this?
My algorithms are mainly swing/long term, so making a few trades a month. I like to trade and back test ETFs like SPY and SPXL since they have high volumes and are more easy to « predict ».
I started with 10k and it didn’t feel risky at all since it was post 2009 :)
I have limited financial knowledge but I would recommend reading about technical analysis and technical indicators (This one is short and sweet https://www.investopedia.com/university/technical/) just so you can grasp the basics.
I originally just wanted to do arbitrage trading, so I built a simple tool to find opportunities using pre-existing services.
But as I slowly learned more about it, I had to expand it into monitoring the exchanges in real time with orderbooks and past trades to get around annoying false positives in my original system.
It's pretty interesting, but I haven't had the time to work on it recently. Plus the increased efficiency of the markets has made arbitrage trading a lot less desirable in 2018.
I would be happy to discuss it with like minded people.
edit: sorry, should point out I've been doing the same for a while and have hit some challenges I'd love to discuss.
scatterstocks.com
Firefox on windows 7 IE does not work either
There's also one with a headline "Tesla stops Model 3 production" or something, and Ford, GM and Chrysler go up
https://github.com/danielecook/iex-api-python
(work in progress)
I plan to integrate it into a larger project.
I'm coming from a biology/bioinformatics background - so I don't know too much about finance but this is my effort to learn about it.
https://github.com/DutchKevv/TradeJS
Under heavy development and simplified. Removed the trading bots part and focusing on realising 'social overview + price alarms' for now.
Only check on chrome svp
Backtester by author of quantstart: https://github.com/mhallsmoore/qstrader
Can i just say that I wish there was some kind of repository of information for drawing charts. And there are so many different kinds of financial charts that's it's a bit intimidating.
Hit me up if you're a VC funded by the US with huge amounts of cash.