-
- Renaissance Technologies is the most successful quant fund.
- https://www.renfund.com.
- They’re ~300 employees, and are based in NY.
- Their medallion fund has the best record in investing history. It’s based off Baum’s models and has gotten 66% gross annualized return (39% after fees) for 30 years (88-18).
- They are a collection of scientists, not wall street buffs. They have PhDs in math/cs, not economics/finance.
- Disney jumped 5% from the earnings report, and will still have the disney+ launch next week.
- Chiro.
- Finished mindhunter season 1, starting 2.
- Real: ed kemper (the coed killer), montie rissell (the guy who wanted Big Red), benjamin barnwright (the fiance), jerry brudos (shoes), richard speck (ripe c*nts), david berkowitz (son of sam), dennis rader (btk killer, adt serviceman).
- Not real: dwight taylor (the grandma/dog killer), roger wade (principal, tickler).
- Made 5lbs maple peanut pecan butter. Tried without roasting. Definitely better roasted (and blends faster). Gonna try macademia, hazelnut, and pistachio next (individually). They’re all about $1/oz, which means a batch is ~$30. Peanuts are the cheapest, then tier 2 is pecans/cashews/etc. These nuts are tier 3, most expensive.
- Hazelnuts = Filberts. Nutella is made from these.
- Ordered 2lbs of each. $100.
- Smoked 20lbs ribs. That’s 43 sessions on this smoker, and it’s been almost exactly a year. Almost once per week!
- Made new batch of 12 shake jars.
-
- Unsubscribed from all bloomberg publications (I had maybe ~4 in daily digest). I find them redundant, half the good content is hidden behind a paywall, it’s verbosely delivered in paragraphs instead of bullets, and it focuses more on politics / international affairs than tech / business / software / finance.
- Supercontest. Made the leaderboard sort the EXACT same as the all-picks, except summed over all weeks and only for completed games. Before, it didn’t do the same “1 cover over 2 pushes” etc.
- Trimmed and brined 2 racks of pork spares (10lbs) and 2 racks of beef short (10lbs) for the smoker tomorrow.
- Today’s music: snow patrol.
- Bud light launching their seltzer next year:
- Doctor Sleep comes out tonight, Stephen King’s sequel to The Shining. It follows the son, Danny Torrance, who escaped the Overlook.
- Between IT, Castle Rock, and now this – I love how much mainstream attention King is getting. Been a fan for many years.
- datetime.fromtimestamp() should always be passed a timezone. Integers are naive, and must be made timezone-aware, otherwise it will have to assume. Some assume UTC, some assume the system-locale tz, you can never be sure. Be explicit instead.
- The UFC fight this saturday is in Russia. Zabit and Greg Hardy on the card.
- Logan Paul vs KSI 2 is also this saturday. Ugh.
- Resolved the 24 payment situation; it had reported confirmation but then rejected it weeks later. Paid for the past 2 months, ensured it was still on autopay with the card swapover. Dude at the front was nice, Joe.
- I’m liking the recent food changes:
- Oats and water raw instead of oat milk in morning shake.
- +Banana +Seeds for the morning for a more complete breakfast.
- Veggies raw in smoothie, rather than juiced. No waste.
- No more lab powders: whey, glutamine, etc. All natural: cinnamon, maca, etc.
- Remember it’s much easier in python3: super().__init__()
- Tradebot.
- Finished the entire framework. Now I just need to run tests to optimize. This step is crucial. Selection is easy, but overfitting is also easy.
- Another +1 for the backtrader package. Full-featured, and very well documented.
- You can add analyzers, including final sharpe ratio reports.
- Example output from my quick sma strategy on msft for ytd:
- Wow. Look how many indicators come built-in: https://www.backtrader.com/docu/indautoref/.
- RSI = relative strength index.
- Keep in mind that every single ticker would optimize around a different strategy with different parameters. You can lump markets together with decent accuracy, but things will even then change along a chronological axis. There’s no perfect approach for all situations for all time.
-
- Did some searching for spicy food challenges. Almost all are way north. None in south bay really.
- Supercontest.
- Updated the text in the payment section of the rules.
- Made the saturday email reminder 5pm, not 8pm.
- New banner. Standard wednesday deployment.
- Renamed all the files in the banner archive to include the buyer.
- New rule – you can pay incrementing $5 to take down the previous banner and put yours instead?
- Eg: Person 1 buys standard $5 banner. Person 2 hates it and pays $10 for their own banner. Person 1 gets their $5 returned. Person 3 hates them all and pays $15 for a third banner. Person 2 gets their $10 returned.
- Committed lines, submitted picks.
- Made the rankings grouped properly: https://gitlab.com/bmahlstedt-group/supercontest/issues/134.
- Abstracted the ranking cell color to a common layer, and added it back to all-picks (now that ranks are tie-grouped accurately).
- Final FF moves, got everybody except the defenses in both leagues (7/9 successful waivers).
- Upgraded the citi account in mint.
- Seaborn is a wrapper around matplotlib for visualization in python: https://seaborn.pydata.org/.
- So many people took advantage of the robinhood glitch. This is bad. All the SEC and public attention…wouldn’t be surprised if they get temporarily disabled for an audit, or shut down completely as malpractice, or even long-term bankrupted due to poor reputation.
- I’d have to rewrite all my logic around another broker’s API. Woohoo.
- I’d also start having to pay commission again. Woohoo.
- The glitch itself:
- With gold, robinhood gives you some margin based on the cash in your account.
- If you buy shares, and then sell them as options, robinhood would double count both in your total value, so it would allocate you more margin. That’s the bug.
- You could repeat this until you had infinite money. People were posting screenshots of >1M positions, starting with 1k.
- Went to Clippers/Bucks with Harner.
- Ta-eem grill on Melrose. Mediterranean kosher. Delicious.
-
- Tradebot.
- Read a bit more about backtrader, both for general strategies and specific backtrader syntax.
- Robin-stocks has a ton of useful information. http://www.robin-stocks.com/en/latest/functions.html#getting-stock-information. Basically everything that’s available about your account, holdings, and the entire public market.
- Added the full capability to transact remotely. Can login, view all account data, buy, sell, market, limit, etc. Closed the ticket. https://gitlab.com/bmahlstedt-group/tradebot/issues/2.
- Wrote the backtester components.
- Starting money.
- Quantity to buy.
- How and when to buy.
- How and when to sell.
- Commission.
- Online ingestion of yahoo finance data, current and historical.
- Reporting of portfolio performance: gross profit and % change.
- The buy/sell triggers right now are very simple, using basic analytics on the past couple days’ closing price. I’m off to the indicators section next, using SMA and then more complicated models.
- A single AA battery has 2000-3000 mAh, about the same as a phone battery.
- Lost both FF games and got 1.5 in sbsc.
- The columns in the user table for email prefs were non-nullable but had no defaults, so registration was broken.
- FF moves. 4 for ESPN, 5 for Yahoo. Lotta bye weeks.
- Under armour dropped 18%, under investigation for fluffing sales numbers.
- Muse again today. Lots to go through. Solid.
- Gonna try vegan protein powder again, from peas and brown rice instead of whey isolates.
- Bought Orgain. $12/lb, similar to my standard Optimum Nutrition bin.
- More carbs (all fiber). Slightly higher fat (couple grams). Slighly lower protein (couple grams).
- Bought sesame and chia seeds.
- Apple donated 2.5B (FB and Google made similar smaller donations) to make bay area housing less insane.
- Edge is the new Microsoft browser (RIP Explorer), and it launches Jan 15. It’s built off of Chromium, Google open-source Chrome. Weird.
- Transferred 10k from BOA to Robinhood for the tradebot to experiment with.
- Other options besides GTC: https://www.investopedia.com/terms/t/timeinforce.asp.
-
- McDonald’s fired its CEO because of a relationship with an employee. Yay accountability! I’ve seen the exact opposite in my career.
- TikTok is owned by a Chinese company, and is undergoing a security audit because of that implication.
- New season of silicon valley started last week.
- Lewy Bodies are deposits of the alpha-synuclein protein on the brain, causing erratic dementia.
- Jeopardy tournament of champions is back, including Holzhauer.
- Quick refresher on numpy. Matrix and linear algebra functionality.
- Quick refresher on pandas. Series is 1d. Dataframe can have more dimensions. Can group, sort, reshape, statistics, all the standards you’d expect.
- Sharpe Ratio is the metric for portfolio success, adjusted for risk.
- (return – risk_free_rate) / standard deviation of excess return
- Return in excess of risk, averaged per unit of risk.
- The higher the Sharpe Ratio, the better.
- Lol stefon diggs on family feud: https://www.youtube.com/watch?v=jFJaJTXvWl0.
- Tradebot:
- Business school applications are reducing in quantity. Good. STEM needs more.
- Ordinals, like datetime.toordinal, are integers that indicate the number of days since a specific epoch. They represent dates like YYYY-MM-DD in sortable fashion, or for regression, or for whatever else. They can be relative to the proleptic julian calendar, or proleptic gregorian, or others.
- Unix timestamps are the number of seconds that have elapsed since 1970-01-01 (midnight, the beginning of that day). Every ~31.5 million seconds is a year, so we’re at about 1.5 billion.
- Listened to muse most of today.
- A cat ran on the field during the MNF game and delayed it for a couple minutes!
- P/E is great for comparing companies apples-apples, because it’s normalized by share price. market value per share / earnings per share (eps). It can go negative, although this will usually just be reported as NA (losses, usually). Higher P/E is obviously more favorable for the outlook of the company.
-
- UFC 244, good all around. Prelims had some great finishes. Crazy to have the BMF title fight end with doctor stoppage – he might catch some flak for that. I like masvidal, he’s growing on me.
- Canelo fight, was losing but then got the knockout in the 11th.
- Love a good visual comparison between NYC and SF: https://medium.com/conquering-corporate-america/the-difference-between-living-in-new-york-and-san-francisco-3e8ae58832a5.
- Supercontest.
- DST ended. The app handled it well.
- Fixed the sort on pick percentage to properly count failures.
- The logo background-image was conflicting with the css background linear-gradient, so I have the color on the td and the team helmet on a div within the td.
- I really enjoyed the comparisons of amazon (sales %) vs kayak (pay per query), as well as orbitz (sell natively) vs kayak (redirect to sell), when doing my pharmaceutical modeling. Here’s a similar comparison between Ben & Jerry’s (slow) vs Amazon (fast): https://www.joelonsoftware.com/2000/05/12/strategy-letter-i-ben-and-jerrys-vs-amazon/.
- Watson had some incredible saves in his game against the jaguars this morning.
- The NFL schedule was nicely spread out this week: 1 thurs, 1 sun early, 6 sun morning, 4 sun afternoon, 1, sun night, 1 mon.
-
- Ufc data: https://gist.github.com/martj42/695f9a976b7beaa11617840467ac0352.
- Remember to run a bash command from ipython, just start the line with an exclamation mark.
- Tradebot.
- Added all reqs.
- Got live prices with yahoo_fin, at first. It’s an extremely simple package, basically just fetching data from yahoo.
- It was only about 400 lines, and was a poorly written package (didn’t even specify requirements), so I gutted and rewrote it myself. Condensed to about 100 lines. Delinted.
- Created a ticket to adjust risk based on ISM PMI later: https://gitlab.com/bmahlstedt-group/tradebot/issues/3.
- EPS = earnings per share.
- TTM= trailing 12 months.
- YOY = year over year.
- MRQ = most recent quarter.
- EBITDA = earnings before interest, tax, depreciation, and amortization.
- PEG = price/earnings-to-growth.
- Data fetch capability is complete.
- I now have a few decisions to make, as to the strategy/stance the bot uses to trade:
- Get the day’s biggest losers, run regression/ML on the historical data (how far back?) for all 100 to see who is trending positive, and buy at that low. Get the day’s biggest winners, and sell any you’re holding at that high.
- Use analyst/earnings/stats. Decide which ones, do an instantaneous fetch, and trade based on those numbers.
