About automatic points

What are automatic points?

Automatic points are points that the system gives automatically (each night) to analysts according to certain predefined rules. Automatic points rank analysts (or their analysis/models) e.g. according to how their recommendations have succeeded, how accurate the estimates have been, how completely they have offered the necessary information etc.

Automatic points subcategories

At the moment automatic points are in a very early (development) stage, so they currently cover only one subcategory (recommendation accuracy). Later on automatic points will be given automatically at least in following subcategories:

1. Recommendation accuracy
2. Estimate accuracy
3. Adequacy and validity of data
4. Comments in different categories, answers to customers (available or not)

Different weights with each subcategory

Different subcategories will get different weights so all subcategories are not equally important. The weights of different categories and their subcategories etc. are shown at Points pages. In the future is is also possible that customers form their own weights (and thus also total points) so that they can emphasise those things that they see important themselves: some emphasise recommendation accuracy, other estimate accuracy and so on.

Description of individual Automatic points subcategories

1. Recommendation accuracy

Principally recommendation accuracy subcategory ranks the analysis based on how the recommendation has succeeded in the past:

• If an analyst has had "Buy" recommendation and the company share price has increased, or if an analyst has had "Sell" recommendation and the company share price has decreased, then the recommendation can be considered as successful.
• On the contrary: if an analyst has had "Buy" recommendation and the company share price has decreased, or if an analyst has had "Sell" recommendation and the company share price has increased, then the recommendation can be considered as unsuccessful.

Accuracy measured each month, the latest months weight more

For each month each analysis gets a percentage which describes its relative position on average. So if an analysis gets a percentage of 76%, then that tells that it has been better than 76% of other analysis.

After the relative position of each analysis has been measured for each month, then the final accuracy is defined by calculating the months together. The last months (1 month ago, 2 months ago...) weight more that months further in the history (5 months ago, 6 months ago...). So we do not punish from "old mistakes" as much as from "new mistakes" which makes it easier for the analysts to improve their rankings. Even the best analysts do make mistakes but furthermore good analysts tend to learn from them.

Categories and limits inside one month

In practice the system ranks analysts with recommendation accuracy inside each month in four different categories (in rank order):

•  Successful (points: +1)
•  OK = not successful nor unsuccesful (points: 0)
•  Not available (points: -0.1)
•  Unsuccesful (points: -1)

Categories and limits (monthly)

Example: Recommendation is Accumulate

If the share price increases 5% during the month, then the recommendation is regarded as "Succesful" and the analyst gets to the best category for that month (1 point).

If the share price decreases 1%, then the recommendation is regarded as "ok" and the analyst goes to the second best category (0 points).

If the share price decreases more than 2%, then the recommendation is regarded as "Unsuccessful" and the analyst goes to worst category (-1 points).

If the analyst has not had any recommendation for that month, then he goes automatically to second worst category (with -0.1 points).

How are the percentages actually formed inside one month? (details)

As all analysts have gained some points for month X, then they can be divided to 4 categories. An example with numbers describes best how the percentages are formed in practice:

In one month 38% of analysis have been "Successful", 32% of analysis have been "OK", 11% of analysis have been "Not available" and 19% of analysis have been "Unsuccessful". In that case we can also say that those analysis that have been successful have on average been better than 81% of all other analysis: They have actually been better than 62% of other analysts who have not been "Succesful". Furthermore we also have to count in that as they all belong to group of 38%, then on average they are better than 62% + 19% = 81% of analysts. So we get an average (19%) of their own polulation of 38% (i.e. 38%/2=19%) in order to get a percentage that describes best their placement in the whole population.

If we would not "calculate the average percentage" then we could not take into account the amout of best or worst analysts each month and as markets tend to vary a lot and some months seem to be either very hard or very easy to forecast, then we would miss a lot of information.

So in this example as the proportions of different populations were: 38%, 32%, 11% and 19%, then they would get percentages correspondingly as: 81%, 46%, 24.5% and 9.5%

This percentage describes that they are "on average better than" 81%/46%/24.5% or 9.5% of other analysts.

Months calculated from current date

When we talk about months, we do not refer to calendar months like August, September etc. but the month can be e.g. 4.10.2004 - 4.11.2004. So months are always calculated from the current date.

Points are calculated again every night and thus they always cover different time period: if today is 14.11.2004, then the points are calculated from time periods like: 13.10.2004 - 13.11.2004 (1st month back); 13.9.2004 - 13.10.2004 (2nd month back) etc. And then tomorrow the points are calculated from: 14.10.2004 - 14.11.2004; 14.9.2004 - 14.10.2004 etc.

Further improvements to recommendation accuracy points in the future

We have plans to further develop the recommendation accuracy points with the following things:

•   There will be also longer observation periods than just one month: the share prices tend to fluctuate quite much and thereby it might be that a share that has an increasing trend, might however fall in one individual month. Thus e.g. 3 or even 6 months periods might tell us better about how succesful the recommendation has been than just 1 month periods.
•   There will be more categories than currenty: instead of just +1 and -1, we should give also +2 and -2. E.g. currently hold and reduce -recommendations will both go to -1 category if the share price increases dramatically. After the new categories an analyst would get -2 points if he/she has "Reduce" -recommendation and -1 points if he/she has "Hold" -recommendation and the share price increases dramatically.

See also info about
admin points
customer points
all points

Back to freelance analysis intro
Proceed to freelance analysis start

Copyright © 2004 Valuatum Oy. All rights reserved. Disclaimer.