header-langage
简体中文
English
Scan to Download the APP
Partnered With:

SBF: stop abusing useless data

Read this article in 4 Minutes
The availability of data everywhere has, on the contrary, become an obstacle for decision makers.

Original author: Sam Bankman-Fried, founder of FTX
Original compilation: 0x22D, Rhythm BlockBeats


< p>This article is based on the views of FTX founder Sam Bankman-Fried (SBF) on his personal social media platform, which is compiled and translated by BlockBeats as follows:


Lies , Damn Lies and Data: My Uneasy Relationship with Data


I went to MIT, then became a quant trader, then became Founder of a Fintech company. Outside of work, I'm an effective altruist. At the heart of this proposition is maximizing the positive impact you can have. So, you might think I'm a sucker for data. But in reality, I think most people are overusing and even misusing statistics.


There are so many such situations, and many people would be better off if they ignored some useless data and focused on what they were doing. It took me a while to accept this.


Here are a few examples:


a) Bob is running a Fintech company. He counted the multiples of transaction fees and B2B subscription fees and found that they were 20x and 80x, respectively.

It was this data that made him decide not to do mobile apps, but to focus on B2B services.


b) Alice works for a venture capital firm. She looked at the correlation between the number of employees and market capitalization of the companies in her portfolio. After controlling for many other factors, this correlation is +75%.

In the next round, she decided to primarily fund companies that were rapidly expanding their headcount.


c) Zed needs to choose between a Super Bowl ad or a Facebook ad.

He looked at the impressions per dollar for both ads and found that the latter cost less, so they dropped the Super Bowl ad.


These scenarios vary, and I don't necessarily know the right answer for all of them. But in these few cases alone, statistics are reasonable in (a); correct in (b); and detrimental to their decision-making process in (c).


The point is that you don't necessarily have to choose between relying on statistics or just picking at random. In addition to this, there is a kind of prior, which is based on your intuition and critical thinking. The question is whether the statistics are more useful than your priors, and whether you have combined the data with reality well.


In Bob's example, his data is technically correct. But there are two core issues:


a) The revenue in these two cases may be different; maybe the mobile app brings in 4 times the revenue of the B2B product above.

b) Valuation is not the most important thing! I prefer to look at income.


Bob would have achieved better results if he had just said, "Let's build the business that looks the best" and ignored these valuations .


In Alice's case, she may have misinterpreted the data.


Yes, there is indeed a positive correlation between having 10,000 employees and success: you can only hire 10,000 employees if you do a good job. Correlation is there, but the direction between cause and effect may be wrong.


And what about Zed?


What exactly is an impression? Super Bowl commercials have one important property: They're talked about over and over in so many difficult-to-trace places. If you only look at it from an intuitive perspective, you will greatly underestimate its impact. In this case, a simple gut check might have led Zed to realize the significant impact of Super Bowl commercials, much of it in small talk.


Not all data is useless


There are many ways to use data badly, but that doesn't mean it's useless, there are certainly ways to use data well! But if you're not very good at using data, it's just noise that distracts you from sound judgment. Only a high standard of statistical analysis is really useful! As data becomes more abundant, using and referencing them becomes more convenient and commonplace, so does the misuse of data.


I see a lot of people fall into the trap that most of the stats they cite are useless. Statistics certainly have the opportunity to make a big difference, but that's when solving a very specific, deliberate problem. If you are faced with a decision that you have to make, and you really understand the various factors in the problem, and know which factors you are not sure about, then statistics can be "very" helpful at this time! But generating data aimlessly is just a distraction.


Similarly, there are some traps that interviewers prepare for candidates, especially those with a deep mathematical background: giving a tricky, messy problem, they try to solve it exactly. But if they can't, they doubt themselves.


The opposite of overfitting and outputting irrelevant data is Fermi estimation. While it is difficult to try to estimate certain quantitative factors without knowing all the relevant data, you can at least get reasonable bounds that can be extremely useful.  


Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia

举报 Correction/Report
Choose Library
Add Library
Cancel
Finish
Add Library
Visible to myself only
Public
Save
Correction/Report
Submit