Thank you Georg,
Andreas has anticipated me, evidently he had the same difficult to understand your suggestion. Now is clear…I tested with my 10 stocks Sim, and the result is worse.
May be because in my Sim I use a RS with Percentile NAs Neutral. If I try with RS set to Percentile NAs negative the result is worse than neutral, but improves using your suggestion.
I’ll try on other Sim but at least I learnt a new thing. 
Andreas,
is this with variable slippage?
I added this to my ranking system, performance was better, but only the last 10 Years
new →
(CurFYEPSMean-CurFYEst8WkAgo)/Abs(CurFYEst8WkAgo)
(CurFYEPSMean-CurFYEst4WkAgo)/Abs(CurFYEst4WkAgo)
(NextFYEPSMean-NextFYEst4WkAgo)/Abs(NextFYEst4WkAgo)
since when is this estimate data incooporated in the database?
instead of (old -->)
EPS%ChgPYQ eval(EPSExclXorTTM>0 and EPSExclXorPTM>0,EPS%ChgTTM,NA)Let me ask my question again;
the estimates data, since when is it in the database, since 1999 or did it come later, Thank you
Andreas
Andreas, the reference index says estimate data starts in 2000. Don’t know if there’s more detail though.
https://www.portfolio123.com/doc/doc_detail.jsp?factor=EPSEst&popUpFullDesc=1
2000 looks reasonable. Here’s a series of UnivCnt(“NextFYEPSMean=NA”)/UnivCnt(“True”) for the Fundamental Universe;
Thank you!!!
@ wwasilev → so do I get this right: there are less and less stocks that have and NA on those estimates?
If this is true, I have an explanation, why the new ranking system works better the last then years.
Thank you all very much!!
Best Regards
Andreas
That would be my interpretation. It appears that the number of stocks covered by NextFYEPSMean estimates is increasing over time . Maybe P123 can comment, too.
Walter
PS It also appears that, depending you what’s in your ranking system, being limited to 15 years of data isn’t a big deal. For many of my systems, I start optimization sims from about 2006.
The number of stocks for which we have estimates has remained pretty steady. See below. The reason that Walter’s chart slopes downwards is because fewer and fewer stocks are in the All Fundamentals universe. The chart below was generated with UnivCnt(“NextFYEPSMean > -100000”).
Agree with Yuval; the chart for NextFYEPSMean!=NA is similar.
Is there a method for listing the companies that fell out of the Fundamental universe? I used short price history as a proxy for newly listed companies but finding dropped companies is more difficult.
Walter
Andrea, you kindly gave us 5 out of 8 of your RS macro clusters (see attach).
Can you tell us the difference between “Stock momentum /Volume/Shape” (that you gave us) and “Stock Momentum”?
Fabio
Open a screen. Use “All Stocks” as your universe in the settings tab. Type
Universe(allfund) = 0 and FHist(“Universe(allfund)”,52) = 1
in a rule. Run the screen and you’ll see the stocks that fell out of the All Fundamentals universe in the last year. If you’re also looking for tickers that no longer exist, you can use the Advanced Stock Search and in the “Symbol” box type in ^19 or ^18 or whatever year you’re looking for. You’d then copy those to a list to see if those stocks were in All Fundamentals or not before they were delisted.
Perfect.
Andreas,
Thanks for sharing your system on this forum and Youtube, I have to say I like it so far!
I’ve made some assumptions on your remaining RS nodes and buy/sell rules, but staying within your criteria of low volume small caps, here is what I’ve come up with (consider this ‘base case’):
Variable slippage, w/ market timing, buy rank>80, sell < 70, all RS nodes you provided are unchanged:
Very similar results, but slightly higher turnover.
I have variations of a buy rule that screens for high ranking EPS, results vary for different time periods.
I’ve tried different tweaks on the ranking system with varying minor improvements, feel free to try some, sim results will vary depending on how our buy/sell rules differ, other elements of the ranking system, etc:
• EPS Growth
o Negative values can be problematic, the rule for TTM is nice as it screens out any -ve value EPS’s.
o For quarterly, can try this:
o (EPSExclXorQ - EPSExclXorPYQ )/max(0.01,abs(EPSExclXorPYQ))
• Quality
o Marc provided some suggestions on (a-b)/abs(b); I tried this as well, but with little impact.
o ROE% 5y AVG will result in NA for any companies less than 5 years old, which may skew results. On the other hand, this is also like a screen to give credit to any firms with a minimum 5 yr history. Could also tweak formula to rank NAs negative for this formula (haven’t tried this yet)
• Accumulation/Distribution & Stock momentum/volume/sharpe
o Can try using median volume instead of average volume in calcs. This will eliminate some of the outliers in average values.
• Rank industry/sector specific factors on “industry” or “sector” rather than “universe”
o Profit margins can vary by sector/industry, try ranking by sector/industry
o Same with EPS factors
Let me know if these work out for you.
Cheers,
Ryan
Hi Ryan,
I like the way you are using 25 stocks and variable slippage to help get realistic simulation results.
I’m concerned, however, about the No OTC Universe since it has a lot of illiquid stocks that are difficult to trade. Did you use a liquidity buy rule. Here is an example.
MedianDailyTot(10,0)> 200000 and MedianDailyTot(10,10)>200000and MedianDailyTot(10,20)>@adt and MedianDailyTot(10,30)>200000
This helps to avoid buying stocks that might temporarily have a decent volume in the past week or two but could become illiquid. This rule does not guarantee that there will be good liquidity when it comes time to sell, but it does help.
Brian
Hello Brian,
Yes, there’s a buy rule of mediandailytot(20)>100000.
Thanks for the tips on multiple volume rules.
For comparison, if I add the multiple offsets as you suggested, there is some minor change. Performance drops slightly in the overall 20 year test, however is about the same CAGR in last 10 years, but a smoother curve :-). Turnover increases in both.
See screenshots below.
BTW, What is the rule “MedianDailyTot(10,20)>@adt”? Median volume must be greater than average daily volume for the given period?
Thank you Ryan!!!
Very, very good results, mine are just slightly higher!
Will put out more vidoes, got some stuff on big caps too, but much lower results, but decent capital curves…
Lets keep those strategies and look how they do in realtime.
My realtime results are here: https://twitter.com/GfI_Himmelreich/status/1076219796381937667
Unfortunetly not since 2011, but good enough 
Best Regards
Andreas
One more question: Are you going to trade this Ryan?
Best Regards
Andreas
