Chief Scientist & CEO QIM. UVa School of Data Science (Founder/Funder). Embrace Simplicity, Beware Complexity. Coding at CODE. Honesty. Sleep. Squash.
Charlottesville
- Tweets 8,060
- Following 5,042
- Followers 7,089
- Likes 1,486
In 1987 Warren Buffett predicted that quantitative trading would not be successful.
quora.com/What-does-Warren-B…
17
43
123
AI startup {X} raises $$ to do {Y}.
Reality 99%+ of the time:
Startup {X} exploits AI buzzword to raise $$ to do data analysis on {Y}.
5
23
99
A computer, price data, and the conviction that total randomness in human activity is a ridiculous proposition. nitter.app/colemancrussell/status…
1
15
78
No, the biggest embarrassment to the EMH is the spectacular success of quant funds such as Renaissance, TGS, DE Shaw, Two Sigma, & PDT.
Fama on momentum: "Its the biggest embarrassment to the EMH...I'm hoping it goes away." #CFAChicago
10
25
63
Breathtaking arrogance.
Fama posits that if he hadn't invented EMH, that neither EMH nor BF would exist.
marginalrevolution.com/margi… nitter.app/jarallahalbrake/status…
3
9
41
Excellent thread about cognitive dissonance’s role with continuing to support Donald Trump.
Cognitive dissonance refers to the psychological discomfort that exists when there is some type of conflict or contradiction between people’s thoughts, beliefs, feelings, values, behavior, and/or a specific situation with which they are involved.
Let's unpack.....1/
1
9
30
The arrogance & stupidity of Fama's EMH has always astounded me.
I've enjoyed a long career investing in EMH-defying price inefficiencies.
Eugene Fama has long insisted that asset bubbles don't exist. Greenwood, Shleifer, and You say: Think again, Fama. nber.org/papers/w23191
5
12
26
I am looking into why local vaccination urgency certainly appears pathetic. I am acutely aware that this is a nation-wide issue, but I am certain we can do better.
Coordination of local government & healthcare pandemic response is not in my skillset, but I will try to learn more.
3
22
I use them to bucketize users by different important topics. They work brilliantly for me. Twitter downplayed them forever; apparently they are still unused by most Twitter users.
It is annoying to hear somebody arrogantly dismiss a feature which some find extremely useful.
Iis there any software feature more existentially useless than the Twitter List? Why does this still exist?
4
1
24
Public policy discussions are for government spending. I am simply spending my own money to build an office building and I invest in small local businesses. I am not trying to remake Charlottesville and I certainly don’t want our small city to become the next Palo Alto or Austin.
Actually, bureaucratic entities and policy-making bodies are two different things. One follows rules and guidance, issues permits, etc., the other one allows for public debate and direction (hopefully). Joe is asking a very important question about how you wield your power.
5
3
19
“This helps to explain why momentum strategies have beaten value very convincingly this year, by 22.3% - on course for the strongest year since 1999, (which was followed by a crash for both momentum and for the market as a whole).”
ft.com/content/f8a421f2-c99b…
1
10
25
This is just a terrible take. In fact, we should use AI to carefully re-craft and simplify our laws so that understanding, awareness, and enforcement are far more transparent and fair. The concept of 1000x more complexity in our legal system is beyond bonkers.
1
21
862
But Musk’s posts, and the lack of scrutiny they receive from his company’s own fact-checking system, lay bare how Musk has turned X into a pro-Trump machine.
cnn.com/2024/09/03/media/elo…
2
9
22
2,651
NIPS 2015
10 of the 40 sponsors are trading firms.
AHL
Citadel
DE Shaw
Ketchum
PDT
SAC (Cubist Systematic)
Two Sigma
Voleon
Vatic
Winton
2
11
23
Today is big for my 14yr-old son Piers:
His first official single dropped today.
[collab with GREENDRAGON]
piped.video/watch?v=Cfua3RfM…
📷
1
1
18
2,837
Eugene Fama tried to beat the market, failed, then declared it impossible by introducing the Efficient-Market Hypothesis. #EMH
1
4
17
Yes, week-starts still get me amped. As a teenager it really concerned me that most people yearned for the week-end and did not enjoy work. Coupling my passion for prediction with clear daily feedback has kept me remarkably engaged. Learning continues to be an absolute joy.
Anyone else love their job so much that they get a giddy excitement for the week ahead on Sunday nights?
1
18
3,326
Over the years as I have learned about AI & data science, I have come to recognize that we've been doing machine learning for time series./3
4
2
17
.@Noahpinion
For 30 years, I have found EMH (even weak (prices only) form) to be utterly wrong. The best inefficiencies persist for decades.
3
16
.@mpawlo
As somebody whose business would have been destroyed if the markets were rigged as described, consider that Lewis is simply wrong.
4
14
We don’t correlate to Crabel, so we’re using different stuff.
We certainly measure volatility in various ways, but we don’t filter by narrow and wide range days. I’m not actually sure what that means.
My method is self-taught machine learning prediction with price data. nitter.app/m_mcintosh113/status/9…
4
1
17
.@LaurenBerson
A different wise man who doesn't believe in EMH once said the following about "quant" hedge funds.
Both wise men were wrong.
3
16
Thanks Peter!
My reasons Sharpe Ratio works for QIM:
1) No stops
2) No options
3) Not mean-reversion
4) Targeted vol
Last word (I hope) on Sharpe. I understand and have great respect for @JaffrayW and the reasons why Sharpe ratio is appropriate for QIM
1
5
14
We're never going to return to the days of slowly disseminated news, so markets will legitimately move extremely quickly at times. #HFT
1
6
13
This comment should have been in the article, wisdom on wealth & innovation.
The Economist on @elonmusk.
economist.com/news/business/…
3
15
It's surprising, but those w/o unique data can still do remarkably well in the ML arms race with excellent domain expertise & creativity./5
2
3
13
The best hedge funds have high Sharpe ratios with very low to negative correlation to the market.
@LadyFOHF @michaelsantoli @Chris_arnade
3
6
14
A non-HFT trader reviews Flash Boys:
"The whole damn book is hyperbolic sleight-of-hand." mercenarytrader.com/2014/04/… #FlashBoys
2
7
12
A key component of real machine learning is hierarchical and automatic (as opposed to manual) feature engineering (--> deep learning)./4
2
2
14
Warren Buffett believes that HF managers that only charge incentive fees on alpha (zero management fee) don't exist.
bloomberg.com/amp/news/artic…
1
5
13
Starting at age 11 (1980), I read everything I could find on sabermetrics (baseball data science), most notably all of Bill James’s Baseball Abstracts.
In 1987 I dropped baseball and have been using my own prediction algorithms to trade the markets ever since.
4
14
Excellent overview on how OTC derivatives are used in numerous nefarious ways, worldwide.
nytimes.com/2013/06/28/busin…
/cc @FrankPartnoy
2
9
13
My son PierZ's second official single release:
PierZ & M1CK3Y - Before
piped.video/watch?v=vNw82e0B…
1
2
13
2,158
This is just a terrible take. In fact, we should use AI to carefully re-craft and simplify our laws so that understanding, awareness, and enforcement are far more transparent and fair. The concept of 1000x more complexity in our legal system is beyond bonkers.
AI is going to enable the creation and enforcement of laws 1000x as complex as we have today. This is mostly a good thing.
2
1
12
3,129
I find it very strange that this guy isn't setting up a hedge fund. His pitch is that he identifies great trades!
bloomberg.com/news/features/…
6
3
12
Amazing quote about how most of us separate our work lives from the most important aspects of living fully.
1
15
12
I have never used trend following in 31 years of investing in stocks & futures.
Although trend following in stocks has exhibited a bit of alpha in the past century, I have always had zero interest in chasing trend following alpha (crowded trade) when stronger alpha is out there.
Google search for the paper “Does Trend Following Work on Stocks” by Crittenden and Wilcox. They address the structural issues that make testing and “proving” IF it works difficult.
1
1
11
Humans do these things all the time.
"hallucinates, defames, and leaks private information"
And LLM progress is reducing these issues very quickly.
"Perfect is the enemy of the good."
-Voltaire
5
11
1,271
May I please get 2 friends to copy and repost? I am trying to demonstrate that someone is always listening.
#SuicideAwareness
1-800-273-8255
Just two. Any two. Copy- not retweet.
6
2
13
Since you consider the momentum factor crazy ...
I'd love to hear your take on managing money with price-based machine learning. nitter.app/CliffordAsness/status/…
1
11
Patterns in the final digits of prime numbers.
For base-19, the patterns are much stronger than base-10.
3
8
12
Fanatastic idea.
Combats humankind’s natural tribal tendencies to cluster together around positions and ideologies. Would almost certainly shrink the seemingly ever-expanding partisan gulf between Democrats and Republicans.
11
I see your point vis-à-vis discretionary market timing, especially in this environment. I also wouldn’t ever go near it!
However, systematic market timers can potentially thrive in markets in which crowd behavior is less rational. nitter.app/cliffordasness/status/…
3
12
Using unnecessary complexity to sell things is a primary part of Wall Street's business model.
@egrefen @tyillc @MikeSigner @mcuban
3
3
8
Seriously Dalio?
"Consumers will have to be encouraged to spend, which could include sending cash directly to them."
bloomberg.com/news/articles/…
3
11
Functional coding for me:
Functions as parameters
Striving for purity
Clean refactoring
Immutable data
Algebraic logic
Simplicity
Brevity
F#
1
2
9
If CAPE were well-designed, the algorithm would be time-weighted, and there would be no “quirk” caused by weighting 10-year-old information the same as current information.
Not time-weighting is either a curve-fit tell, or an opportunity to improve the CAPE concept.
There's an odd quirk to the CAPE Ratio. Since it's based on 10 years, we're currently dropping the terrible earnings from 10 years ago. Hence the earnings lining (red) is rising. Blue is regular earning and black is the S&P 500.
3
4
10
Huge fan of your essays here, …. but, seriously, …. what in the world has gone wrong with your bs detector?
1
10
847
Being adversely selected is literally the act of providing liquidity. HFTs' desire to avoid adverse selection = evading liquidity provision.
6
10
Seems your premise is mistaken.
I want to speak out.
I want to understand the situation.
@mtgillikin gently chided me into saying something.
I continue to be horrified by the *overall* pandemic response, & if there is more I can do locally, then I would like to step up again.
9
Severely under-followed on market structure:
@mkt_strctr_guy
@garybasin
@stucchio
@pennyspread
@TrilliumMJF
@etfsamlee
@SniperInMahwah
1
9
The cause is the upcoming French election.
Check out the European VIX term structure. 10 times as severe.
Looking at the $VIX term structure with weeklies, the major jump occurs between Mar 21 and 29. 9% weekly contango! Explanations?
9
If an NLP expert (I am definitely not one) can convince me that they have a robust @kaggle competition design for conference paper-reviewer matching, I will pay for it.
@GaelVaroquaux
@tdietterich
@Plinz
@dvogel20
Assessing the significance of a paper requires to understand it and to know the literature. Therefore, the most pressing thing to improve in order to improve review quality at NeurIPS scale (> 10,000 submissions) is automated paper-reviewer matching.
1
3
9
I am a huge fan of Tesla vehicles, but it was very foolish to use the term "AutoPilot" to describe the current tech. nitter.app/ivanthek/status/751873…
2
10
Worldwide QE has reached its perilous endgame.
Future demand available to pilfer has almost completely dried up.
1
3
9
Listening to 1991 Dave Matthews Band is maybe the most addictive thing ever.
"Don't throw away your playful beginnings"
#DMB1991
@Copperpot5
2
7
"...the number of investment professionals actively engaged ... has risen, from 5,000 to well over 1 million today."
institutionalinvestor.com/ar…
1
3
9
Comprehensive explanation from @liquidmkt eviscerating @nanexllc's $F 20,000 share "market is rigged" example. medium.com/@liquidmkt/what-r…
1
10
7
"Programmers are like chess players trying to play with a blindfold on — so much of their mental energy is spent just trying to picture where the pieces are that there’s hardly any left over to think about the game itself."
medium.com/the-atlantic/the-…
3
1
7
Stock market microstructure has its issues and can certainly be improved, but it is the corporate bond market that is a complete mess. #HFT
3
4
8
I am looking into why local vaccination urgency certainly appears pathetic. I am acutely aware that this is a nation-wide issue, but I am certain we can do better.
Coordination of local government & healthcare pandemic response is not in my skillset, but I will try to learn more.
7
With such track records that only report compounded return, there are several significant items that probably tremendously limited the eventual compounded profits:
Lower-capacity strategy
Taxes
Trading operation costs
Spending profits
@PeterLBrandt, what’d I miss? nitter.app/longshorttrader/status…
2
1
7
Moving the character limit from 140->10,000 is a disaster unless large incentives exist to minimize chars.
@Twitter
Sources: Twitter's feature allowing tweets over 140 characters slated for Q1 launch recode.net/2016/01/05/twitte… techmeme.com/160105/p13#a160…
1
9
.@stevelevine @jeaneaglesham Absurd to imply that Tesla vehicles have no value.
Try driving one.
They are the best cars ever made.
1
8
As a completely bootstrapped startup, we at QIM generally described how we make investment decisions as statistical prediction. /2
3
1
9
Jamie Dimon's foolish one-in-3-billion comment on risk is exactly what one would expect from a too-big-to-fail bank.
bloomberg.com/news/articles/…
3
6
9
I was trading DMB tapes locally in Charlottesville in the early '90s, then starting using email discussion groups to arrange snailmail trades. Generally would agree over email to mutually mail 2-3 chosen tapes to each other. I had collected over 300 DMB tapes by 1996. Fun times!
1
6
Some very important inventions have come from folks brainstorming outside of their fields. Why give a group of people a hard time for trying to contribute their talents during a disaster? This is actually a great example of people coming together trying to help.
This is a more sophisticated & articulate Trumpism: magical thinking that smart rich people will save us.
wsj.com/articles/the-secret-…
8
.@seb_g @bradneuberg
A creative data scientist at Twitter could generate an excellent rating algo for news links. That's my strong opinion.
1
1
7
IOW, extreme sub-sampling then ensembling often works very well for machine learning problems with large data sets.
github.com/glouppe/phd-thesi…
8
Eugene Fama is again busy disproving his own Efficient Markets Hypothesis, although he doesn't see it that way. #EMH
papers.ssrn.com/sol3/papers.…
1
2
8
Nate Silver gets absolutely demolished in the replies to this horrible tweet.
Some of those stories don't click well because, believe it or not, the audience isn't interested only in stories that say Orange Man Bad, Don't Worry Harris Will Win or Liberals Are Right About Everything. But there are other outlets you can go to if that's what you want.
7
968
It seems that some world-class programmers (such as the aforementioned @karpathy) can code a lot more quickly with these new “AI-based” tools. I hope more pedestrian programmers (such as myself) can benefit as well. I will try out CoPilot soon.
1
7
2,247
The definitive book on Machine Learning on .Net (F#-focused) by @brandewinder now on Amazon.
amazon.com/dp/1430267674
1
8
This ranking doesn't include quantitative hedge funds such as Renaissance Technologies, DE Shaw, and Two Sigma. Why?
Ray Dalio of Bridgewater trumps Soros as the man behind the most profitable hedge fund bloom.bg/1lOcERj
5
3
7
None of these things fully prevent virus spread.
Rapid antigen tests speed the identification and hopefully then the proper quarantining of Covid-contagious individuals, thus reducing spread.
I am confident you recognize the value of the diversity of one’s tools.
1
4
This take on the stock market, coupled with your unparalleled expertise in Deep Learning, must make predicting the stock market appear easy to you.
Plus you can have the fun of working on a new dataset that is sure to stretch your ML toolset in new ways.
I guess it will be a real surprise when the market dips again then. Much like no one could have possibly foreseen COVID-19 becoming a problem until... March.
If market movements were informed you would never see big swings. Major events don't become abruptly more or less likely.
1
2
7
Public policy discussions are for government spending. I am simply spending my own money to build an office building and I invest in small local businesses. I am not trying to remake Charlottesville and I certainly don’t want our small city to become the next Palo Alto or Austin.
And it also means those jobs need new physical spaces to exist in — otherwise they'll spread outward and displace others, just like high-end residences sprawling into existing nbhds.
But yes, there needs to be a real effort (by the city!) to help everyone access that ladder.
1
7
Engle was specifically asked about a liquidity measure and fails to be clear that the Amihud ratio is actually an *illiquidity* measure, which is quite confusing.
Furthermore, this ratio is terrible on days with near-unchanged daily price moves (a frequent occurrence).
Nobel winner Robert Engle makes 2 bad missteps describing a simple liquidity measure:
“Engle: We used the so-called Amihud (2002) ratio, which is the absolute return on a day of some particular asset divided by the dollar volume that's traded on that day.”
aqr.com/-/media/files/words-…
2
6
Best analysis I've seen of the false rumor sell off today. Markets handled the situation very well. lunchboxtrader.blogspot.com/…
3
7
7
I have been warning about circuit breaker issues for years.
Circuit breakers cause many types of frantic trading.
2
5
6
Severely under-followed on market structure, part 2:
@ChuckHill137
@PZ_SL
@nick_animal
@pauld44
@rbattali
@MarcosCarreira
@uglylukes
3
1
7
The author made the point that the Case Fatality Rate (CFR) will end up being more in the 0.5-1.0% range, which is totally wrong.
He is confusing Infection Fatality Rate (IFR) with CFR. The two differ enormously by the ratio of infections to total cases (est. around 5-20x).
One of my favorite writers/thinkers on healthcare hit a home run with this post. It is truly a "must read" Covid piece. "Treat the Patient, Not the Virus" by @david_goldhill link.medium.com/kOedvKvK55
4
I wholeheartedly agree with @ritholtz's take on the ratings agencies' involvement in the financial crisis.
bloombergview.com/articles/2…
3
7