ScifoSpace:
Signal processing — that’s a deep background, deeper than many people probably realize. I’m familiar with some of Shannon’s early information-theoretic ideas, but I’m curious how those concepts show up in modern signal processing and ML workflows today.
Thanks!
It was a long, long time ago—my knowledge there is surely obsolete by now. That said, it helped me build a structured way of thinking. Today it’s much less about old technical knowledge and much more about curiosity, iteration, and persistence...LLM are massive weapons to boost the way of thinking in just hours...My focus right now it is finance and quantum finance 
Insider buying definitely attracts attention. According to Finviz the filing hit the SEC at 7:07PM last night. FactSet still does not have the data. We could certainly try to process the data faster than Facset, but it's not really our business to compete with day-traders, and stock opened +7% anyway. It will be interesting to see what happens to MU in the next few months, and when we will get the data from FactSet
It is either planned or not planned (1-0) then the type basically says if it is buy or sell. If its planned it means the insider has a prearranged selling schedule. Hims basically just sells generic medicines anyone can get at costco or amazon but they sell them at high prices. They grew significantly because they had great marketing and a lot of convenience in terms of giving prescriptions.
Hims did extra well for a bit because they were compounding another company’s patented weight-loss drug, that was having a shortage which led to growth. The few insider buys took place amid these shortages. Now the patent-holder is fighting that and there is no shortage.
The types are just SEC form codes but basically it is buy and sell (P and S) for now. M just means theres an options component. Just click on the plus sign to expand the transaction and you see what it is. Usually just sale of an exercised options package when you see S M. You can see shares unchanged in your photo.
Correct. Amazon is quietly duplicating their model including telemedicine for ED and weight loss prescriptions with Amazon One Medical
Maybe of interest just to me but it turns out pink eye is the first diagnosis on their web site when I visit it for the first time (I think they treat about 30 diagnoses online). Pink eye is something that is self-limiting, has only symptomatic treatments and can be confused with herpetic keratitis, angle closure glaucoma or uveitis without a slit lamp exam. I guess I will just note things are changing in medicine.
Before either came along we had TDOC which started quite a while ago (2000’s I want to say) but they didnt have good marketing and I think they didnt have a pharmacy. That stock is not doing too well these days as Amazon entered the space. Interestingly, Amazon started it as a service for employees only and now it is greatly expanded. There is a physical clinic very close to me. When I need a medicine I just send them a text message via the app and gets shipped via Amazon pharmacy. Not bad! I am subscribed to the yearly plan and get free (included) consultations for most common ailments. I dont get sick too often so have not tried it too often, but like what I see so far
They treat those for free as part of the subscription but then you can pay for telemedicine and in-person appointments for anything else. They are doing a bad job explaining it. If they woke up and did a better job with this it would explode. I am very disappointed in terms of how they explain everything, but the potential is there
Well. What impressed me was the quantity of sells without any buys...Personally I traded successfuly for the past 2 years all the rollercoster of $hims selling puts and calls...it s was a nice wagon until volume and trending fade off. Tomorrow will report results...we will see if there some more room for fun...thanks for the explanation @SZ . I appreciate it
When will the new insider info be available to use in simulations?
Are there any updates as to when we can sim this?
Yes it's planned. Got sidetracked with several things. Adding this new data is part of the ongoing effort of adding fresh new dataset like the ones below, and tools for mining it. Few other datasets planned as well. I will try to get to it before we dive into Asian data since that will suck up a lot of bandwidth.
I came across this pertinent post:
sorry, link didn’t post well, but if you click through–interesting use of insider data that appears to validate the approach.
Not sure I agree with the methodology for that perplexity tool but does show the power of these AI tools. It is going to be a year for the history books
Thanks for that link. Gives me an inspiration for functions that can then be used to train an ML model or for a ranking system. I'm just throwing out some ideas. We have not started defining the more complex aggregation functions.
InsBuyers(days [, max_score])
InsSellers(days [, min_score])
InsBuyAmt(days [, max_score])
InsSellAmt(days [, min_score])
Calculate the win rate of the buys in a period. Can specify sector, industry, etc
InsBuy1YWinRate(future_days [, scope, max_score])
InsBuy3YWinRate(future_days, [, scope, max_score])
Averages the future performance of all the buys in a period. Can specify sector, industry, etc
InsBuy1YAvgRet(future_days [, scope, max_score])
InsBuy3YAvgRet(future_days, [, scope, max_score])
How many distinct bullish or very bullish (P123 score 1-2) in the past 30 days
InsBuyers(30, 2)
How many distinct very bearish sellers in the past 30 days
InsSellers(30, 5)
Sector future 1Mo win rate for insiders that bought with conviction in the past year
InsBuy1YWinRate(30, #sector, 1)
Love that we can customize the score on these examples
Hi Marco,
Another use case came to me recently. Using the insider activity at an aggregate, macro level to assess the overall market risk-on/risk-off balance, from the perspective of corporate insiders. For example, if I were to trend insider net buy/sell for all S&P500 companies, on a weekly (monthly?) basis, and I saw, for example, that in the last 3 months the aggregate selling neared a level akin to prior market tops, that’d be useful data for me to assess my own risk-on/risk-off behavior. Kind of like MACD of Advance-Decline Issues, but for insider trades. I hope this makes sense….I can go into more detail if the idea sounds interesting.
Do we pull the buys weekly or daily? Wondering since I tried pulling some CPNG ones but they have not registered
Is there an update when this will be available for use in ranking systems etc.? Maybe worth releasing the basic functions if it will take longer to engineer the more complex aggregations, as I imagine those will already be useful to most.