Silly question but I have some doubts as volume and liquidity as I saw conflicted information on the forum.
(1) What is the difference between “AvgVol(10)” and “avgdailytot(10)”?
(2) Are they in number of shares or $ quantum?
(3) Are they in million or units?
(4) Let’s say I want to limit to stock of at least a volume of $100,000 a day, is the following rule appropriate: (Vol10DAvg * price) > (100000/1000000) ?
(5) It seems the simulations are providing me different results if;
(a) I use “AvgVol(10)” vs “avgdailytot(10)”
(b) I include it in the buy rules or in the universe
I’ve seen cases on AAII’s screening results where average daily trading led to investing in low liquidity stocks. I’m talking days with no volume at all, but averaging out because of a few very high volume days. But having days of no trading volume in the recent past makes me suspect of liquidity in the future.
So I use a defined formula, “$LIQUID50”, that looks for the MINIMUM daily trading average over 50 days, instead of some AVERAGE daily trading:
Thank you.
My mistake. My question was the difference between “AvgVol(10)” and “Vol10DAvg” (and not “AvgDailyTot”).
Also by adding the same liquidity rule ("AvgDailyTot(10)>100000 ; to limit to stock below 100000 USD) on the buy rules or universe, the results are vastly different.
I do not understand as I would have expected that adding the same rule within the universe would have excluded the non desirable stocks from the universe.
Vol10DAvg is measured in millions and AvgVol is not. So the number you get from AvgVol(10) will be a million times bigger than the number you get from Vol10DAvg.
You’ll get different results putting a rule in a universe and as a buy rule. When you put a rule in the universe, you’re removing stocks from consideration by whatever ranking system you’re using. If you put your rule as a buy rule, the universe is going to rank all the stocks first and THEN exclude the ones that fail the rule. So your simulation is doing things in a different order, which can end up making quite a difference. It probably won’t make much difference if your rule just eliminates a few stocks, but if your rule eliminates a huge number of stocks, the ranking could be completely different. For example, if you were ranking stocks on R&D to market cap and excluded biotechs, the other stocks would get very different rankings depending on whether biotechs were excluded before or after the ranking was done.
This is not intuitive and many (even pros) have made this mistake in the forum. Fortunately, Marco and Yuval seem to thoroughly understand this.
What helps to confuse things is that you get the same results (using buy rules or rules in the universe) with single-factor ranking systems. Our confusion is related to problems we have visualizing things in more than 3 dimensions (2 ranking factors plus the returns equaling 3 dimensions). If you never liked to think about the Pythagorean Theorem in 2-dimensions, in school, you should probably just memorize what Yuval says.
Anyway, something that most of us (other than Yuval and Marco and a few others) have to remind ourselves of regularly.
Related and important to everything we do: What is a rank performance test? It is a projection of a multidimensional (hyper) space onto 2 dimensions isn’t it? Any problems (limitations) with that, you think? Like the above, your intuitions are likely to be wrong and you will not be able to visualize it.
This really affects—and depending on the answer, limits—everything you do. Not that it could be all bad. I probably would not be able to visualize anything without the Rank Performance Test (simplification).
Sadly, AI machines do not have my limitations with regard to seeing patterns in n-dimensional space. The Rank Performance Test is just an unnecessary (human) “crutch” to them.
I realized most of the stocks are sold as there is no more rank (no rank). I believe they were excluded from the universe and sold (not ideal as it was a universe defined by liquidity), hence the different result.