Marco,
Will we be able to use the new functions UnivAvg(), UnivCnt(), UnivSum(), etc. as buy and sell rules in Sims & Ports like:
100 * (UnivCnt(“close(0)>sma(200)”)/UnivCnt(“true”))>50
I tried it but it doesn’t work now. I got the error: [color=red]Error near ‘UnivCnt’: Invalid command ‘UnivCnt’[/color]
Denny 
I’m getting the same error as Denny
Relllis,
WYNN’s Pr2BookQ on 11/18 went from N/A to 17K+ which throws off the average. I think that is the correct value for that ratio since their BookVal was close to 0.
We should probably add UnivMedian to make this type of series meaningful.
Or you can add your own sanity check to eliminate from the average the outliers. For example in “criteria” for UnivAvg use Pr2BookQ<20
Yes, please add UnivMedian, it would be very helpful in many cases like this one.
Very cool Marco!
I will love the see a contributing factor (weight) that can be negative(!) for each component.
An optimisation mechanism will be cool too.
Marco & All:
Is it possible to create a weekly series from daily data? If not, would you consider adding this functionality in the future. Weekly data (with its inherent filtering of daily fluxuations) is sometimes valuable. Thanks.
Bill
dmicsa, you referring to the series tool? what weights?
strader1, maybe. Inside the “formula” you have free reign. You have a specific example?
Can we please get more than 15 years of data for the series tool. For example, % of SP500 stocks > 200 day average. It is very limited with only 15 years/2 recessions. I would like to test some market timing strategies with a longer period of time.
thanks, Debbie
Marco & All:
Yes, I would like to be able to calculate / use Rsi(x, y, #Series) based on weekly data. Can I use "top-down" series to do this for say SPX? If so how? Thanks.
Bill
Marco, now you sum with factor 1 on all oscillators. It will be nice to can alter the weight of any oscillator we use. Some oscillators works well reversed and are not symmetrical.
A parameter defaulted to one on all oscillators will make it behave like now.
I know we can duplicate oscillators and they will contribute with integer multiple but still the one that contributes in a negative way are impossible to quantify.
The overal oscillator:
OverallOscilator = Sum(Wi * Vi) / Sum(Wi); where Vi = {0, 1}
now is done like this:
OverallOscilator = Sum(Vi) / n; where Vi = {0, 1}
Marco & Team,
It looks like Christmas is coming early this year. Thank you for this extraordinarily valuable upgrade to the platform. These new functions save so much time over the alternative (labouriously copy-pasting into a sheet / running series operations in sheet) that I am just about ready to shed a tear…
Best regards and keep up the great work,
D
Has anyone managed to create a series based on existing macro data? E.g. the Junk Bond vs. 10y treasury spread?
Great addition Marco. Will it also be available in screens?
Hi Marco
Thank you very much for this powerful tool. I have a couple of questions:
For example, I would like to plot the price performance differential for the last 20 days between the top 10% and bottom 10% ranked stocks based on Price/Sales, excluding N/As.
Is it even possible to use FRank in the Series? I receive an error message for:
FRank(“Pr2SalesTTM”, #All, #DESC) >= 90
ERROR: In Rule 4: Error near ‘FRank’: Invalid command ‘FRank’
Is it possible to do a certain calculation to create a series, and then transform this series to an SMA (series) to be accessed in macro charts? I see it is possible for the chart, but I would like to create an SMA series directly. For example, I would like to use the example series “SP1500 Vol ADC Pct” but create a SMA of the Highest (or Lowest) series values, or create BollingerBand around the “SP1500 Vol ADC Pct” series. The current macro charts backtest only gives the possibility to either set above/below SMA or a flip flop around the raw value. I would like to create a SMA series, and then use the flip flop on the SMA series.
Is it possible to use the LoopSum within Series?
Thank you
whotookmynickname
P123 Team,
I am just realizing how amazing a feature this could be. This will solve a lot of my problems. Really impressed how you can elegantly solve many solve problems with a single addition.
I am getting an anomalous value for the S&P 500’s earnings when I try to calculate the forward PE and PEG ratios.
I am doing it two different way, but both are resulting in anomalously large earnings spikes between 2006 and 2007.
Method 1 for Forward PE is as follows:
Method 2 is:
In theory, these graphs should be close, and they are somewhat. However, I get the following chart which shows that earnings estimates spiked and plummeted in a way that I think is clearly indicative of an error, either on my part or due to the data…
Please, advise.
Try excluding BRK.B, that’s what we do in the SP500 series
When the primary issue changed from .A to .B there’s a lag in the fundamentals that we do not handle.
Berkshire is notorious for causing a lot of headaches.
Thanks, Marco. I tried the rule UnivExclude(“BRK.A, BRK.B”) but the actual problem was that I ran the experiment on ALL stocks with fundamental data. Still, this is a sizable anomaly that probably deserves attention.
When I try to measure Corporate Profits on all stocks with forward estimates, I get an anomalous reading where profits seem to jump a few trillion $ during 2007. The approximate start date of the anomaly is Jan 19, 2007 and ended August 18, 2007.
Being ever diligent, I looked for firms with anomalous corporate profits in the screen. I found that during this period, Novelis Inc (NVL^07) & Manifold Capital Corp (NVL^07) result in gargantuan earnings anomalies. These could be fat finger errors passed to us by Compustat or failed currency conversions.
Whatever the cause, these kinds of buggy values are going to cause significant amounts of headache for us who want to create our own industry/sector/market aggregates. One bad value can throw off an entire aggregate. Anomalies in the data or inevitable, and all we can do is guard against them. Even with all the headaches about to come trying to clean & prep the data for aggregates analysis, I think the new series tool will produce a lot of fruit.
One last thing… It would awesome if you guys could give us tools to create aggregate robust to outliers and also tools to detect outliers and anomalies more easily. While medians are most robust to these outliers, they don’t always describe what we’re looking to describe. Some ideas to help with functionality of aggregate functions include:
FYSA:
One of the hidden feature of the series creator is exactly this: finding data anomalies
Looks like the culprit is MANF, not sure about NVL^07. I checked the EPS estimate and it is huge. I re-run the calculation and still comes out huge. Will have to investigate further.
A trim parameter was envisioned. It’s a bit of work since it involves a sort, then a trim. But with financial data it’s necessary.
In the meantime you can do your own trimming with something like this:
“#AnalystsNextFY >= 1 AND #AnalystsLTGrthRt >= 1 AND between (abs(CurFYEPSMedian) / abs( EPSExclXorA) , 0 , 10)”
Thank you, Marco. I can’t wait to see what you and team come up with. If I were you, I might by now be fed up with all my requests and suggestions. But I was just thinking that rather than fancy trimming methods, it might be a more elegant solution to have Date() and Year() functions so we can ignore certain tickers only during the time periods they cause issues. Ignoring BRK.B shares patently discards 1.85% of the S&P 500’s market cap. Rather than ignore $330 Billion of the S&P’s capitalization, I would sleep easier if I could just ignore BRK.B for the entire month of February 2010.
This custom universe which ignore only problematic tickers during problematic time periods could be created in the Universe Tool or, with the addition of new syntax, in the Series Tool.
Just for everyone’s awareness, the following are tickers I am trimming in order to get reasonable data for my Forward PEG estimate: AFLYY, NCMV^12, MANF, NVL^07, TROZ^09, BRK.B.
