huge feature on quants taking over wall street in WSJ....... and in it one professor says he tells his students to learn R and python..

if you know quant, then WSJ artilces are nothing new except that quant is red-hot these days and i would think is the future.

Quants took over Wall Street like 15 years ago.

Have you looked at any of the Data Science/Machine Learning courses on edx.org? I went through DAT210x recently and it was pretty good. DAT203 also looks good. If you need to learn Python, then do the MITx 6.00.1x and 6.00.2x. The 1x course begins in another week.

Is quant really any "hotter" than it was 5-10 years ago? I don't doubt that quants are very much taking over, but from what I understand it's more that they're just eliminating LOL trading etc. jobs than massively increasing quant jobs.

"Quant" also has (and had) a bunch of different meanings to different groups. To some it was simply sorting on BE/ME and momentum. For some it's "stat arb" (a term which has also had some evolution). Others conflate it with HFT. To the extent that it's "hot" in 2017, usually there's some mention of AI and ML (although such was also the case 5-7 years ago).

i agree with pretty much everything said in the last few posts.

quant is alot of different things........

alot of it is just automation of what highly paid labour formerly did.

i also think often quant is simply better collection and analysis of fundamental data. i remember an analyst telling me many years ago that all the proprietary data that he collected back then - like writing down numbers from an industry journal each month - were being made available online - and now we're many years later.

the kind of quant i'm interested in starting out with value, momentum, sector rotation type models....... but now i'm not sure why people can't construct pretty sophisticated models of oil supply/demand or the same for copper and submit them to rigorous analysis/testing.... basically i think the amount of data has just exploded the last 10 years and things like SSRN has really helped to spread info.

firms like fidelity i'm sure had amazing insights about different sectors with all the great people they have but it was so expensive and they manage so much money that i'm not sure how much they can act on it that much to make a huge difference... but now i think most of the great insights they had are basically out there in the market.

unless you are doing really detailed analysis of a limited number of situations, i think fundamental will continue to morph into quant. even fundamentally how much can quant not capture unless you are going so far as to do old analyst work like calling supplier/customers/competitors/etc... and i would presume lawsuits/sarbanes-oxley/etc. has put a damper on those sources anyway

this article is such a buzzkill.... i'm shocked these guys replicated 447 anomalies....

i always thought that the quant industry's dirty secret was that alot of these stratagies work alot better on stocks that don't trade that much.. 2 problems: 1) can't implement it with much $$$$; 2) are the historical results even accurate/feasible? i.e could you have traded at a bunch of closing prices for illiquid stocks?

https://papers.ssrn.com/sol3/papers....act_id=2961979

Abstract
The anomalies literature is infested with widespread p-hacking. We replicate the entire anomalies literature in finance and accounting by compiling a largest-to-date data library that contains 447 anomaly variables. With microcaps alleviated via New York Stock Exchange breakpoints and value-weighted returns, 286 anomalies (64%) including 95 out of 102 liquidity variables (93%) are insignificant at the conventional 5% level. Imposing the cutoff t-value of three raises the number of insignificance to 380 (85%). Even for the 161 significant anomalies, their magnitudes are often much lower than originally reported. Out of the 161, the q-factor model leaves 115 alphas insignificant (150 with t < 3). In all, capital markets are more efficient than previously recognized.

btw, i had read the article before today but i saw that CXO posted in the last day or two.

reading it in a few minutes it was hard to figure which anomalies passed their t-tests but some did.

tnx...don't know how I missed this...
will read later...

looks like many of the anomalies that survived and have the best returns, t-stats and alpha (after adjusting for a bunch of other factors) are mostly momentum and pure value.

alot of the momentum is very short term oriented.

pointed me to at least one very interesting paper i hadn't seen... one of the best things about academic papers is the bibliography.

A lot of anomalies vanishes as you value weight as their concentration is predominately in the lower cap space. Obviously researches identify anomalies, not necessarily to trade them effectively. (lol academia :P)

Also some notion on why equal weight > value weight in absence of anomaly...
http://docs.edhec-risk.com/mrk/00000..._Portfolio.pdf

But still, yea, there is a lot of p-hacking nevertheless. I think there was a paper back that showed that due to hacking literally only momentum and value cleared adjusted p-value significance.

Still this is a very good paper...probably the best in a while in painting a true picture...

Rikers, thanks...

i think there's two things,

there's EW vs. VW of the S&P 500 for instance.. you are still buying very very large companies but you are underweighting exxon, wal-mart, microsoft, GE etc.. i think it makes sense although i don't think of many huge companies as having very high p/e's

then there's doing EW portfolios of momentum stocks within an 8,000 stock universe....... so you end up with a bunch of stocks with mcap less than $100MM, maybe alot less... put another way, if the strategy chooses S&P 500 or Russell 2000 stocks then i can picture that. if it's EW of huge universe it feels like they are picking stocks from some "cloud"

i rolled my eyes every time i see a study where they are equally weighting tiny stocks to demonstrate their anomaly. not this study semi-confirms my instincts.

here's a conceptual "thinking out loud" question,

say for strategies like "dogs of the dow" or "mean reversion" that perhaps work really well over the long term and most years....

is it really weak to take a backtest and say "well, i wouldn't do that strategy during the credit crisis"........ in fact, i'm not sure your boss in almost job situation would let you do these strategies fully and aggressively in october 2008.

dogs of dow would have gotten slaughtered. C and AIG. maybe one more too. surely you could apply some max yield or max volatility filter although i realize that's selective memory and/or look-ahead bias

mean reversion - buying weak days, multi-days or week - would have actually hung in pretty well as long as you sold on some strength..... not sure if everything from a day to week measurement of weakness worked but i am actually surprised from my work how these strategies didn't work that badly in credit crisis.. buying strength got killed.

Although a little farfetched for BFI some notable - practical paper on quantitative understanding of algos is below. Pretty epic stuff (although not so epic in hindsight)

"Why Should I Trust You?": Explaining the Predictions of Any Classifier
https://arxiv.org/abs/1602.04938

I've run it on some of mine algos, and locally it telly you exactly what went into decision and how it was scored. Kinda to give back for the previous paper.

[QUOTE=rivercitybirdie;52305759]
dogs of dow would have gotten slaughtered. C and AIG. maybe one more too. surely you could apply some max yield or max volatility filter although i realize that's selective memory and/or look-ahead bias

turns out AIG didn't have much of a dividend yield b4 credit crisis. and then when it got absolutely killed it may have suspended its dividend by that time. plus, generally the dow dogs is judged by annual picks and almost certain AIG wouldn't have made portfolio in january 2008

JPM was there too.

so if you did annual "dogs of the dow" on january 1 2008, i'm guessing C was chosen - say for 3 or 4 stock portfolio - but not necessarily JPM and probably not AIG.

of course, there are about a million reasonable filters you could have applied to make sure you don't end up with crazy dividend yielding - in theory - AIG, C and JPM are your 3 picks...... this presumes say you are adjusting monthly or quarterly.

do the big fundamental index rebalancers eventually flag things like Bear Stearns and Lehman and say "these earnings or dividend trailing figures aren't representative of the situation"?....... otherwise you would have been buying more and more Lehman each day. never to a big weight or anything but constant buying based on earnings figures that are essentially meaningless - at least in hindsight

Another issue I ran into with black box algos is some of the algos produced trading strategies that ran afoul of the SEC. I had a problem with an algo that produced trading patterns that SEC really didn't like (looked like banging) and hilariously, the person that came up with the algo basically responded: "I was wondering why it worked so well."

I do think the quant trading strategies are gradually getting out of hand for two reasons:
1. there is quite a bit of group think
2. a whole lot of "quants" have no idea why the strats work and are unintentionally exposing themselves to a lot of risk, some of which is regulatory. Put differently, ironically, traders are increasingly financially illiterate.

At the same time, the algo traders obviously have injected liquidity into the market and narrowed spreads on almost everything exchange traded so curtailing them too much is almost certainly a bad idea.

I honestly don't know if there is anyone who understands the issues well enough to know the right levers to pull/buttons to push to eliminate the risk of another unnecessary flash crash and still let the algos fight over the crumbs and narrow spreads.

what's banging?

is that when a stock promoter tries to look like there's action/breakout in his stock by using multiple accounts? or just pump/dump which is similar? or something else?

@Maverick93 he's made enough from 25NL poker for this claim.... I'm guessing this is easily in the $150,000 to $250,000 range depending on where he goes.

I don't think I've ever heard of anyone crushing 25NL for $150,000 profit while still in college and pursuing graduate studies. Have you? Because Maverick93 is claiming such