Hi,

Is it possible to have industry group ranking like marketsmith http://www.screencast.com/t/YYxSFLxw7d

I don't know how they are calculated. Are they just based on the percent changes over specific spans of time?

If so, we could display the raw percentages and sort by these percentages, but there isn't a way to display the ranks as WatchList Columns.

Never used MarketSmith, but here's what Google says. From MarketSmith's FAQ...

The Industry Group Rank is derived from The William O’Neil + Co. 197 Industry Groups^® stock classification method that groups the stocks in our database into like categories based on their industry and tracks their intraday performance.  The Industry Groups are ranked according to their price performance over the past six month, allowing you to easily identify the best and worst performing industries. Studies show most top-performing stocks are found in the top quartile of groups and that group action determines at least half of a stock's performance.

You can access the Industry Group Rank by expanding Reports > Industry Groups in List Names, or by selecting the View All Industry Group Data button in the Related Information panel.

Source: http://marketsmith.investors.com/support/Topic.aspx?name=Industry+Groups&type=10#1

I suspect IBD uses something similar to the IBD RS Formula forstocks, which is

40 * C / C63 + 20 * C / C126 + 20 * C / C189 + 20 * C / C252 - 100

Since the previous post suggests that only 6 month performance is evaluated, something like

40 * C / C21+ 20 * C / C42 + 20 * C / C63 + 20 * C / C126 - 100?????

You could adjust to your own preferences

I personally use 

(.50*((C - C5)   / C5)+
.30* ((C -C13) / C13)+
.20*((C - C21) / C21))*100

which "evaluates" over a shorter time period. I also compare that value to the same formula 5 days ago to see how the ranking has changed, for better or worse.

If MarketSmith and IBD use similar methods, then bcochrane's approach seems like a good one. That said, the text quoted by UnclePennybags really doesn't say. It could just be a straight 6 month percent change.

100 * (C / C126 - 1)

Not really my thing, but I took a minute to dive deeper into MarketSmith's FAQ.

Looks like some linear regression might be involved but their tossing in the word "proprietary" tells me we're not going to find their exact secret sauce calculation anywhere online....

The Industry Group Rank is a proprietary number obtained by calculating the least-squares curve fit of summed prices on certain stocks within that industry. Another calculation is then done using all companies in the group. Separate weightings are used for different time periods. Groups are ranked from 1 (best) to 197.

The Industry Group Relative Strength Rating is presented in both numeric and letter forms. The Numeric Rating and Letter Rating are percentile-based versions of Industry Group Rank.

The Numeric Rating presents the Industry Groups ranked in value from 99 (highest) to 1 (lowest). The Letter Rating presents the Industry Groups ranked in letter grade from A (highest) to E (lowest).

Source: http://marketsmith.investors.com/support/topic.aspx?id=526138&Ntt=industry+group+rank

This is the most updated list as of today. http://www.screencast.com/t/JYlEK1EVa

Can Tc2000 experienced guys help me  reverse engineer it in TC2000

I really don't have a way to reverse engineer it based on the screenshots. The fact that it returns a rank actually obscures the values used in determining those ranks. And while the raw values might not be reverse engineerable, it would be a lot easier to do so with raw values than with ranks based on raw values.

Unless there is more detailed information available about the calculations someplace, you are going to get speculation and not reverse engineering.

Even if we had the exact calculations used, we would only be able to display the raw values. There is not a way to then convert these raw values into rank and display the rank in a WatchList Column in TC2000 v12/v17.

Some suggestions as to what to use so far based on the 6 month time span would be a weighted percent change covering different time spans.

40 * C / C21+ 20 * C / C42 + 20 * C / C63 + 20 * C / C126 - 100

Or a straight six month percent change.

100 * (C / C1 - 1)

Or a linear regression slope (this one is converted into a percent change over six months).

100 * (EXP(6 * (FAVG(LOG(C), 126) - AVG(LOG(C), 126))) - 1)

$149.95 per month? Holy sh*t ... no wonder he wants to reverse engineer it. 

raw values would work  Bruce. I tried using 100 * (C / C126 - 1) .  industry groups with those conditions are somewhat similar.

No matter how similar they are, if you use the raw value as a sort, the result will be the industry groups in a specific order (it would rank them even if it doesn't display a rank).