Dude! This is super helpful. I run a website with millions of huge text PDFs and thousands of users, and being able to implement this will literally save my users man years of time.
Interactive Brokers offers a cost plus commission structure, where your trades pay the exchange commission (or collect the rebate, if you are providing liquidity) and they just take on a ~$1 fee per trade on top. Very few "regular Joes" actually want that - what you don't realize is that exchange fees vary a LOT, each exchange easily has dozens of different rates depending on security, time of trade execution, the state of the order book at the time, and order type. Placing a market buy order when the book is sell-heavy? The exchange will actually pay you for that. Going the other way? That's a very expensive order. What if you take the order to a different exchange where the book is more balanced but the price is worse for you? It might make sense when you factor in the difference in commissions. This is why online brokerages exist, to abstract all of this complexity away.
I've looked into the XBRL data - it's closer to an XML version of each company's 10-K rather than a unified accounting/reporting system for all public companies. If one company reports their income statement using 20 lines and another uses 15 lines, the XBRL data will show both exactly as they appear in the 10-K and they would not be comparable (unless your software can understand that "Net Interest Expense/(Income)" for one is the same thing as the sum of "Interest Expense" and "Interest Income" for the other). To add to that I've found some companies are filing XBRL is a non-conforming XML standard (probably software issues) and some companies just include the last 3 years of data in every XBRL document (rather than just picking the same data that's reported in the 10-K), making the consumer figure out what is the actual new information in the document.
I run a similar free site and was looking at add ads. Google Adsense rejected it for not complying with their program policies. My data is on large US federal bankruptcies, so I really couldn’t pin point why but just a heads up that it might be more difficult.
I do a Bootstrap modal and blur the background for sign up forms. The right hand drawer looks great but what about mobile devices? A bootstrap modal appears centered on all screens
Carriers pay to port numbers to a neutral third party that maintains the Number Portability service. In the US, that used to be $6 per number port (which is basically just a DB row update operation) but a new vendor has reduced it by more than half. Carriers view that expense as pocket change as subscriber acquisition cost is well over $100 in total.
Without a legislative mandate and a third party integration partner that assures the port is successful, seems unlikely that a market leader like Spotify would ever allow them to exist.
Chargebacks are way harder in real life than people think. It took me multiple calls, emailing evidence/screenshots, and months to get a $300 charge back processed. That was for a completely fraudulent charge at a bar. The way it was explained to me, banks will only process a chargeback when you have provided evidence of trying to contact the merchant and they refused to provide a refund and couldn’t show a valid contract supporting the charge. So in this case you couldn’t just charge back NYT just because their cancelation policy is slightly burdensome. And they don’t stop that business from charging you again, so even if you can charge back a charge, it doesn’t solve the issue of canceling the subscription.
This doesn’t count the people who read the newsletter online or on the Bloomberg terminal (where I typically read it). I didn’t even know there was a newsletter till recently.
I’ve found Tika’s PDF to HTML parser to be pretty good. My only complaint is that in a double spaced document where there is equal amount of space between paragraphs and normal lines, it labels every line as a separate paragraph.
Thanks - I cannot get Camelot to run in parallel (I use celery workers to process PDFs), there is some bug in Ghostscript that SEGFAULTS. I’ll try using PDFPlumber instead! By the way Apache Tika has been the best for basic text extraction - even outputs to HTML which is neat.
I have a very similar project where I’ve extracted text and tables from over 1mm PDF filings in large bankruptcy cases - bankrupt11.com
I still haven’t found a good way of paragraph detection. Court filings are double spaced, and the white space between paragraphs is the same as the white space between lines of text. I also can’t use tab characters because of chapter headings and lists which don’t start with a <tab>. I was hoping to get some help from anyone who has done it before.
I love used tabula and recommend Camelot over it. The Camelot folks even put together a head to head comparison page (on their website) that shows their results consistently coming out ahead of Tabula.
My other complaint with Tabula is total lack of metadata. It’s impossible to know even what page of the PDF the tables are located on! You either have to extract one page at a time or you just get a data frame with no idea which table is located on which page.
Does the code have any factor investing features at all right now? I was just reading about alpha lens, another project focused on performance attribution of factors. To avoid reinventing the wheel, are there any other libraries that can be leveraged to add factors - so portfolios are constructed to go long and/or short a certain factor like sustainability or etc.
I wonder why they decided to start from scratch in Kotlin rather than just add the interactivity feature to plotnine. Now we have two competing projects...
True, and if I recall Apple is even a “small” LP in the Vision Fund. So Tim Cook could have bought his own stock but instead thought giving a few billion to Masa was a better trade! But hindsight is 20/20 and we don’t know what the fund is ultimately going to return. I have friends who work there - the sad part is the amount of sheer work that had to be done to keep the returns where they are today, restructuring old investments and squeezing every drop from the winners to make up for the total write offs.
They had 6 full exits generating $5B of gains, plus another dozen companies went public and the mark to market on the whole is a $3B gain. I don’t see that they’ve ever invested in Bitcoin. The remaining portfolio is down about $5B on $60B invested - I have a lot of questions about how they are marking those securities, but surely it’s similar to the VC model of “Mark to the last third party investment, even if we led the investment, unless something really bad happened”. That’s where the WAG and WeWork losses would be. So all in they still think they are in the black by $3B on $80B invested. Even if they are mis marking the private book by 20%, the fund would generate a 0.9x return or disappointing but not a complete disaster.
Thanks. There is one aspect of SPACs that is completely unmentioned in this write up that’s actually the most important - valuation. While it reads like Bill believes the goal of a “SPAC-off” is to negotiate the sponsor take, that’s not the primary goal. Each SPAC completes an acquisition, with a PRICE. That price is not set as $/share - that’s fixed at $10.00 - but in the number of shares the “seller” gets. I was involved in one transaction where a SPAC was desperate to complete a deal before it expired, and it vastly “overpaid” for a crappy business. I put quotes around that because not a single investor except for the Sponsor rolled into the deal at $10.00. They ended up having to raise outside equity at $8.00 per PREFERRED share, and even that wasn’t enough. Then they went to the bankers and issued them more shares in lieu of their IPO fees and then the sellers ended up contributing more money with a complex security. So the problem in those situations is the company is put into a bind - if they pick the SPAC that’s paying the highest price, then most likely there won’t be any new capital raised at all while the sponsor will still get their rake. If they pick a low valuation, then they can assure capital and negotiate down the rake, but then you are giving a below market valuation and incurring a rake plus paying IPO fees so why not just do the IPO to begin with. And because the sponsor’s rake is in shares (with a lockup), even if the shares go from $10 to $3, they will still get paid. Think about this incentive - the sponsor can announce the crappiest deal - buy WeWork or whatever - wait 6 months, watch investors lose 70% of their capital, and still make a profit by selling their sponsor shares (which are issued for effectively ~$1/share as the sponsor contributes 1-3% of the initial capital at $10/sh but then gets 10-20% rake at $0/share).
SPACs traditionally don’t have any guaranteed primary capital. 100% of their investors can redeem. So you are guaranteeing the sponsor their rake but they are not guaranteeing that you’ll get any money from the listing. Plus this article doesn’t mention the warrant coverage, which cannot be negotiated down, and IPO fees (when SPACs go public, the bankers defer their fees until the SPAC completes a transaction). I think this article researched the inefficiencies of the IPO process to the last detail, but somehow forgot to do the same on the SPAC process. Door #3 is just as rigged as door #1.
Source: invested in several SPACs transactions and PIPEs
You can’t calculate sharpe for a portfolio that includes private investments (PE or VC). Even if you trust the marks - big if - the returns are reported quarterly or at best monthly, and the marks are often unchanged creating a fake effect of a “zero vol” investment
I live in NY and created my own version of the case-Schiller price index just for Manhattan. The July 2020 index shows a 10% YoY decline in condo prices, which is worse than a 6% decline in Q2. Please see my analysis for backup: https://github.com/ivoytov/manhattan/blob/master/NYC_price_i...
There are many other claimants in the Purdue bankruptcy other than the Multistate Group mentioned here. Each of those claimants is incentivized to make their claim as large as possible to reduce the amount left over for the hospitals, Native American tribes, individual victims, and the many other groups fighting to maximize their slice of the Purdue pie. Full list of unsecured claimants is in this document: http://bankrupt11.com/dockets/documents/p_purduepharma414/
The fees are ridiculous. Google Scholar is free but it’s very difficult to use if you just want records for a particular case. In my attempt to solve the problem, I created the website bankrupt11.com which offers free access to federal bankruptcy records.
They are definitely still operational but they no longer compete with Ancestry for content like they used to. In fact they rely on Ancestry to digitize their own archives