Hiring someone to replicate an academic article

[metalztrader]

Interesting. Not to completely get off topic of the thread but what are the nodes and edges in this graph? I have done quite a bit with stock data and graphs but it has been pretty fruitless.

I just got thru turning an academic paper into a trading system. I used GPT professional and it took me about 3 days to get it like I wanted it. However it did come up with usable code within a few minutes, I just spent a lot of time messing with it.


Bounded Acceleration Shortest Path

 

[metalztrader]

I just read the paper. I would think to exactly replicate this it will not be worth the cost of the data.
All the work will be in paying for the data and the data wrangling. Even using pandas it will be non-trivial without building a proper database if you ever actually want to update it with new data(and pretty useless if you can't update the data). Excel there is no way.
I would imagine AQR already had this as clean data in a time series database.
Someone would have to do a ton of work to even know what to quote for the data itself. I agree that algorithm wise, chatGPT4 can write python code for this pretty easy once you have clean data in a consumable format.
I also don't think the idea is really that interesting either though.

 

[systematicinvesting]

I also don't think the idea is really that interesting either though.

Thanks for your response. I agree that the main work here involves collecting and then preparing the data for the back-test. The Python code shouldn't be too difficult to write. My problem is that I don't have the time or expertise to do either the data collection and preparation or writing the Python code myself, which is why I am looking to hire someone else to do this.

However, I wanted to ask for more clarity regarding why you think the idea isn't interesting. Can you elaborate? Assuming that a backtest does manage to replicate the article's claims, and further assuming that an OOS test from 2014 - 2023 produces decent results, what are the reasons you would still not invest in this strategy in real life for at least a portion of your portfolio?

 

[metalztrader]

I just find the whole idea of factor investing unsatisfying. Maybe from trading the Russel 2000, the classic Fama factor of small versus large companies is just wrong at this point.
The whole idea has been given an gloss of science because Fama won the Nobel prize in economics. Cliff Asness was a student of Fama so in that paper he is trying to show his teachers ideas are correct since he has invested so much of his own time in those ideas as well.
Without Warren Buffet, I don't think any would ever talk about "value investing" at this point as a valid mental model of the markets in 2023. Of course buying intrinsically undervalued stocks is a good idea. "Only buy a used car that cost 20% less than the fair market value" is a good strategy if trying to buy a lemon too.
Momentum I think is the worst of all though. Stocks don't have mass and velocity so they can't have "momentum".
If you started from scratch knowing nothing about the markets, no one would ever come up with these outdated categories in 2023.

Thanks for your response. I agree that the main work here involves collecting and then preparing the data for the back-test. The Python code shouldn't be too difficult to write. My problem is that I don't have the time or expertise to do either the data collection and preparation or writing the Python code myself, which is why I am looking to hire someone else to do this.

However, I wanted to ask for more clarity regarding why you think the idea isn't interesting. Can you elaborate? Assuming that a backtest does manage to replicate the article's claims, and further assuming that an OOS test from 2014 - 2023 produces decent results, what are the reasons you would still not invest in this strategy in real life for at least a portion of your portfolio?

 

[Concinnity]

I just find the whole idea of factor investing unsatisfying. Maybe from trading the Russel 2000, the classic Fama factor of small versus large companies is just wrong at this point.
The whole idea has been given an gloss of science because Fama won the Nobel prize in economics. Cliff Asness was a student of Fama so in that paper he is trying to show his teachers ideas are correct since he has invested so much of his own time in those ideas as well.
Without Warren Buffet, I don't think any would ever talk about "value investing" at this point as a valid mental model of the markets in 2023. Of course buying intrinsically undervalued stocks is a good idea. "Only buy a used car that cost 20% less than the fair market value" is a good strategy if trying to buy a lemon too.
Momentum I think is the worst of all though. Stocks don't have mass and velocity so they can't have "momentum".
If you started from scratch knowing nothing about the markets, no one would ever come up with these outdated categories in 2023.

I agree with a lot of this except the momentum part. Current overall momentum and the velocity of the bars created within that, does play a small to medium part in current action and an even larger roll in future action of how price responds later. Maybe I am wrong on exactly how or why what I am seeing works, am open to that possibility. However, my backtesting works almost identically to my forward live testing, so there's no hindsight involved in it and whatever I am doing is working consistently for some reason.

 

[Quanto]

Hmm. I'm not that convinced of that strategy b/c it has a very high MDD of -46.4% etc (cf. Table 2A on page 57):

 

[Concinnity]

Come on guys most of these places just collect a lot of funds and do whatever they can to turn a profit on it and get their cut. Why does that matter? because it really has little to do with being a full time retail trader. I mean if you have like million dollar portfolio's and/or throwing someone else's money around, than that's fine I could see where this interest you. In the end most of them are just using capital, time and/or natural bullish bias of Major US equities to turn a profit. Just like most people do, while claiming they don't.

 

[systematicinvesting]

Hmm. I'm not that convinced of that strategy b/c it has a very high MDD of -46.4% etc (cf. Table 2A on page 57):

View attachment 328930

That MDD is only for the Value style. What the paper recommends is to combine the 4 styles into 1 portfolio in order to benefit from diversification and the uncorrelated nature of the 4 styles. The portfolio that combines the 4 styles has a MDD of 18.5% (Table 6A).

 

[systematicinvesting]

I just find the whole idea of factor investing unsatisfying. Maybe from trading the Russel 2000, the classic Fama factor of small versus large companies is just wrong at this point.

As far as the "size" style is concerned, the paper doesn't disagree with you. On page 29, it says "There are two additional styles that readers may be familiar with that are not considered in our paper, namely size and illiquidity. The size style has not proven as robust as the four styles we focus on (for one thing its realized premium is significantly smaller than the others) and has had varied success in out-of sample tests".

If you started from scratch knowing nothing about the markets, no one would ever come up with these outdated categories in 2023.

Is there some link (maybe another paper?) that you can point me to which has found that these 4 styles - value, momentum, carry, and defensive (and more specifically, a portfolio that combines these 4 styles since any one style can certainly have long periods of under-performance before recovering) hasn't produced good results in recent years?

 

[murray t turtle]

In the end most of them are just using capital, time and/or natural bullish bias of Major US equities to turn a profit. Just like most people do, while claiming they don't.

ARE YOU sayin' thy ARE NOT telling you everything you KNOW!!??
WOW, thats close to a lie, but not exacty a lie.LOL
17%/maybe \maybe not next 10 years.IF it does it for 3 year I would keep looking also ; thats about 1.5% a month like credit card payment LOL