Index | Recent Threads | Who's Online | Search

Posts: 373    Pages: 38    Prev 29 30 31 32 33 34 35 36 37 38 Next
Last Post
This topic has been viewed 18029 times and has 372 replies
InspectorSector
Re: FactSet beta site v1.0, NOW LIVE

If I shift the test back five years, the performance looks somewhat better. But I would get some clarification on how PIT the estimates are before getting too comfortable with it. And if the formula hasn't worked for the last 5 years I would be questioning the usefulness going forward.

Jul 16, 2020 2:18:18 PM       
Jrinne
Re: FactSet beta site v1.0, NOW LIVE

BTW, this is what Zacks does in their screener. I do not use Zacks but any reputation they may have is based their handling of earnings estimates.

Zacks determines which analysts give the best ratings and then they look for an increase in earnings estimates from those analysts specifically, immediately before the earnings report comes out. This info is used in the Zacks rating. We don't have this ability using P123.


Steve,

Exactly correct for "Zacks Rank." Zacks rank does use what you describe (and more).

They also have a screener that uses the formula being discussed.

Link here: Zacks' Screener

Best,

Jim

From time to time you will encounter Luddites, who are beyond redemption.
--de Prado, Marcos López on the topic of machine learning for financial applications

Jul 16, 2020 2:29:13 PM       
Edit 4 times, last edit by Jrinne at Jul 16, 2020 2:34:59 PM
Jrinne
Re: FactSet beta site v1.0, NOW LIVE

If I shift the test back five years, the performance looks somewhat better. But I would get some clarification on how PIT the estimates are before getting too comfortable with it. And if the formula hasn't worked for the last 5 years I would be questioning the usefulness going forward.


I do not disagree. Here is what FactSet says about their earnings estimates: https://www.insight.factset.com/hubfs/Resourc...1006-837288881.1589971006

Are the earnings estimates data different than fundamentals at FactSet with regard to being PIT? I ask this not having an answer as to whether one is more PIT than the other.

If there is no difference, I think I will drop EBITDATTM/EV first (in this market).

Best,

Jim

From time to time you will encounter Luddites, who are beyond redemption.
--de Prado, Marcos López on the topic of machine learning for financial applications

Jul 16, 2020 2:38:36 PM       
Edit 8 times, last edit by Jrinne at Jul 16, 2020 3:30:16 PM
InspectorSector
Re: FactSet beta site v1.0, NOW LIVE

That doc gives me a headache. The problem with the previous vendor was that they claimed that it was PIT but they didn't freeze the data until Monday morning, too late for P123 to process. So it may have been historically correct for them but not for P123.

In any case, the earnings growth formula does not represent growth. There are actually multiple problems with the formula that has been previously mentioned. But if you are OK with that then that is all that really matters so long as you are not trying to sell others on the approach. I don't really care what Zacks does. Lots of people misuse formulas and get away with it.

I'm also not understanding your point with regards to EBIT/EV. It is clearly not working so why even bring it up?

Jul 16, 2020 4:46:33 PM       
judgetrade
Re: FactSet beta site v1.0, NOW LIVE

Thank you Steve, also your blog is great!!
Andreas

Jul 19, 2020 5:41:45 AM       
judgetrade
Re: FactSet beta site v1.0, NOW LIVE

@Yuval,

"Abs(CurFYEPSMean) / Abs (CurFYEPS4WkAgo). This would rank stocks that had an EPS estimate of -0.13 four weeks ago and an EPS estimate of -0.01 right now very low. A stock that had an EPS estimate of -0.01 four weeks ago and an EPS estimate of -0.13 right now would be ranked extremely high. Is this what you want?"

hmmm, I get the logic and I understand the argument.

1) Factors do not work on their own, they work in relation to the whole ranking system. So I might capture good earnings development, when
both are positive, and I might get a great price with when both factors are negative, since the other factors keep things (quality, value, momentum, industry momentum etc.) in check. At the end I do not know. Judging one factor is difficult, in conjunction with 20 more factors I think its
to complex to really understand and explain.

2) Do we really now the reason why a factor works, or is is always a thesis that was produced in hindsight after the backtest?
o.k, there are factors where the reason is clear (earnings going up, higher stock prices). On other factors like small caps and momentum,
I think the thesis came after the backtest. I read a lot justifications on why momentum works (slow adaption to secular earning trends reflected in price...), none is really convincing to me: it might be one reason for one stock and another reason for the other.

One of my other hobbies is selling and I know what works and those factors have been researched by academics with a positive confirmation. But but the "why" academics find is usually very thin. It works, so you use it (for example ancoring a sales price to the customer, so you suggest first about 20-50% (depending on the country you sell) over a the price you might sell at the end).

I have a trading system not changed since 2003, its based on momentum and it picks stocks with a very high stock price (e.g. stocks that do not split).

There is some academic research, that confirm, that high prices stocks combined with momentum outperform, but its really thin.
I do not trade it since it is into big caps, but it produced a 17% performance since 1999.

What I am saying is: sometimes you found something and it has not be rationalized (and therefore it might be a good not found edge by others), you do not know why it works and you take the risk to still use it.

Still your argument is strong and my one is weaker, I am not sure and still thinking about it...

Best Regards
Andreas

Jul 19, 2020 6:32:24 AM       
Jrinne
Re: FactSet beta site v1.0, NOW LIVE

All,

Here is a study from the text "Modern Portfolio Theory and Investment Analysis that I think supports Steve’s ideas on the importance of growth as well as the importance of earning surprises:

"Kormendi and Lipe (1987) examined the effect of earnings information on stock returns. They introduce the notion of earnings persistence and begin to explore the notion of unexpected earnings…... This coefficient is often referred to as the earnings response coefficient, or ERC….. Kormendi and Lipe hypothesized and found that the ERC is positively correlated with earnings persistence; that is, the stock market responds more to the earnings news of firms whose earnings changes are more persistent. This indicates that the market understands the time series properties of a firm’s earnings, and stock prices adjust accordingly to earnings information.

According to the text, Steve has a point when he says that growth can be an important factor. But an earnings estimate surprise is something different. Growth is only one thing that may (or may not) indicate an earnings surprise.

There are additional factors that may be predictive of an earnings surprise (and may not be related to growth). Most writings relate the equations (CurFYEPSMean - CurFYEPS4WkAgo) / Abs(CurFYEPS4WkAgo) and (NextFYEPSMean - NextFYEPS4WkAgo) / Abs(NextFYEPS4WkAgo) to an increased chance of an earnings surprise.

But I cannot imaging that these would be the only factors related to earnings surprises--or necessarily the best. I am not going to debate what factors or equations are the better ones. Maybe you will use all of them or more than one of the good factors in a sentiment node.

Anyway, if the academic research is correct, you may want to follow Steve’s advice about growth and use whatever equations you think may be predictive of an earnings surprise as well.

That is unless you are going to use just one factor in your ranking system. Then you might want to debate the relative importance of each alone.

Best,

Jim

From time to time you will encounter Luddites, who are beyond redemption.
--de Prado, Marcos López on the topic of machine learning for financial applications

Jul 19, 2020 9:46:13 AM       
Edit 11 times, last edit by Jrinne at Jul 19, 2020 3:55:48 PM
yuvaltaylor
Re: FactSet beta site v1.0, NOW LIVE


1) Factors do not work on their own, they work in relation to the whole ranking system. So I might capture good earnings development, when
both are positive, and I might get a great price with when both factors are negative, since the other factors keep things (quality, value, momentum, industry momentum etc.) in check. At the end I do not know. Judging one factor is difficult, in conjunction with 20 more factors I think its
to complex to really understand and explain.

That's why it's even more important to only use factors that you really understand. If you think something works, it's a good idea to figure out why it works, and I mean "figure out," not "guess." If you use a factor that you don't understand, it will do things that you don't want it to do. And your money is at stake.

2) Do we really now the reason why a factor works, or is is always a thesis that was produced in hindsight after the backtest?
o.k, there are factors where the reason is clear (earnings going up, higher stock prices). On other factors like small caps and momentum,
I think the thesis came after the backtest. I read a lot justifications on why momentum works (slow adaption to secular earning trends reflected in price...), none is really convincing to me: it might be one reason for one stock and another reason for the other.

Yes, sometimes the thesis comes after the backtest, unfortunately. That's the academic approach: do a backtest, say that momentum or small minus big "works," then try to come up with an explanation. It's not a very good approach. Which is a better approach to curing a disease: testing a huge number of drugs to find one that might work or looking at all the symptoms for a root cause and then addressing that by engineering a drug that targets that cause? Both approaches are frequently taken, but one is clearly better than the other. It turns out that most factors have relatively simple explanations for why they work. Why does momentum work? Because, as I wrote recently, "Companies that outperform are more likely to continue to outperform than companies that underperform, and companies that underperform are more likely to continue to underperform than companies that outperform. Since companies that outperform tend to increase in price and companies that underperform tend to fall in price, momentum is a natural consequence of the persistence of performance." See https://blog.portfolio123.com/2020/02/24/why-momentum-works/ If a factor seems to work but makes no sense, it should be discarded without further ado. Sometimes you come across a factor that seems to work but you can't figure out why. In that case, it behooves you to dig deeply and understand it before putting it to use.


What I am saying is: sometimes you found something and it has not be rationalized (and therefore it might be a good not found edge by others), you do not know why it works and you take the risk to still use it.


Yes. Nobody knows why certain herbs cure certain illnesses. They seem to work, and so people use them.

Personally, however, I'd prefer to understand the factors I'm using to pick stocks. And if a factor doesn't make sense to me, I won't use it.

Yuval Taylor
Product Manager, Portfolio123
invest(igations)
Any opinions or recommendations in this message are not opinions or recommendations of Portfolio123 Securities LLC.

Jul 19, 2020 10:06:20 AM       
Jrinne
Re: FactSet beta site v1.0, NOW LIVE

Which is a better approach to curing a disease: testing a huge number of drugs to find one that might work or looking at all the symptoms for a root cause and then addressing that by engineering a drug that targets that cause? Both approaches are frequently taken, but one is clearly better than the other.


Yuval may be right about which is better in medicine—especially going forward. But more lives have been saved, to date, by accidental discovery of anomalies in medicine. Usually called "serendipity" in medical school.

It is not even close. Louis Pasteur’s discover of penicillin being the prime example. More lives have been saved by antibiotics than anyting else medicine has done. A few more examples:

Bacitracin (an antibiotic) was literally discovered by randomly culturing Tracy’s infected knee—daughter of the developer.

Digitatalis had no rational explanation when first purified from the foxglove plant (Digitalis purpurea) for heart disease.

The small-pox vaccine was a result of noticing that "milk-maidens" who developed cow-pox from milking cows had cross-immunity to small-pox.

Washing hands between patients used to be a controversial idea. Hungarian physician Ignaz Semmelwei discovered that washing hands between patients saved lives in the hospital. But since there was no theory of germs at the time (and no theoretical explanation) his idea did not catch-on at time.

Very recently, I had a patient who use Prialt for chronic pain relief. This was discovered from random testing of the toxins in cone snail venom. So this remains a successful method for the development of new medicines today.

But I agree with Yuval going forward. As an example: Artificial intelligence-created medicine to be used on humans for first time

I hope Yuval is right—about medicine anyway. Luck only takes you so far.

Best,

Jim

From time to time you will encounter Luddites, who are beyond redemption.
--de Prado, Marcos López on the topic of machine learning for financial applications

Jul 19, 2020 11:00:11 AM       
Edit 14 times, last edit by Jrinne at Jul 19, 2020 5:13:41 PM
InspectorSector
Re: FactSet beta site v1.0, NOW LIVE

"testing a huge number of drugs to find one that might work or looking at all the symptoms for a root cause and then addressing that by engineering a drug that targets that cause?"

Yuvol - This is a very controversial area that we could spend years arguing. In the case of Covid-19, the virus will mutate faster than Big Pharma can do double-blind trials. But they will claim victory none-the-less by demonstrating that the vaccine works against a milder form of the virus. There is also a bias for engineered drugs that can be patented. There is some clinical evidence that the the immune system response can be improved which doesn't require expensive engineered drugs. Such an approach would also eliminate the need for a seasonal flu vaccine.

Jim - I only work with sales surprises. I have totally abandoned earnings.

Steve

Jul 19, 2020 11:22:32 AM       
Posts: 373    Pages: 38    Prev 29 30 31 32 33 34 35 36 37 38 Next
 Last Post