The problem with the model portfolio is the following:
If there is a target from which a 50% return is expected, then to finance the purchase, one must not sell a stock from the portfolio that has an “Add” recommendation and a 10% expected return. This is because one must not sell/buy against one’s own recommendation.
Thanks for letting me know! It really didn’t work, I fixed it yesterday and it seems to be working now! 
Thanks for the comprehensive analysis. The past few years, and especially 2022, were indeed very challenging for the analysis team. Of course, there have been various factors making forecasting difficult (war, recession/downturn, rising interest rates, etc.). The long-term return on positive recommendations was really strong and much better than I expected. This encourages continued research into companies with a buy recommendation. However, I must admit that the past few years have somewhat dampened this enthusiasm, as companies with positive recommendations and also in the model portfolio have seen significant negative news and share price declines. @Antti_Luiro Would it be possible to get such a review of the analysis team’s accuracy from the perspective of returns, for example, annually in the future?
This was our idea now that the process for calculating the figures has been done, the update should certainly be possible annually in the future.
We only included the parameters used in this calculation in this data; recommendations are in any case our primary means of communicating the analyst’s view
The weak return of the Model Portfolio in recent years has already been discussed in several contexts, and it has been publicly visible in real-time all along, so this was now done from the perspective of the overall tracking accuracy.
Were there any specific points in the logic of this recommendation accuracy calculation that you disagreed with, or can I elaborate on something? The intention was to conduct as fair an analysis as possible, I would gladly hear comments on how this could be improved 
And this analysis can also gladly be replicated utilizing the provided recommendation data
I am asked this repeatedly. This company was founded to democratize investment information by connecting listed companies and investors. Establishing a fund would not, in my opinion, serve this purpose. It would erode the idea of the current model to act as a neutral, independent information platform. That does not mean that an excellent, sustainable, and valuable business cannot be built with this choice that deviates from the financial industry mainstream.
That is certainly a noble idea that Inderes should also aim for. I asked that question ironically, because I consider virtual returns to be virtual returns.
Investing with real money is more difficult because it involves psychological, risk management, and market mechanism challenges that merely following recommendations does not have to face.
For example, the strength of Inderes’ recommendations might be small companies followed by a single analyst. The stock rises due to a positive recommendation with low trading volume. A small portion of investors certainly benefits from this. For funds, returns should be generated in companies with higher liquidity. In these cases, achieving excess returns is much more challenging because investor information is often democratized through many banks and other investor entities that follow and analyze the company.
Yes, this is absolutely true, the calculation does not account for the impact of liquidity and is indeed a virtual return, as this strategy has not been implemented with real money. However, it still tells a lot about the accuracy of the analysts’ recommendations.
Regarding liquidity from the perspective of a small portfolio stock picker: in a portfolio considering trading costs + the Inderes effect, with a starting portfolio of 10,000 euros, the position size per stock would have been approximately 350-1000 euros for 1x weighted stocks (Add) and 700-2000 euros for 2x weighted stocks (Buy) during this 2013-2024 period.
Our coverage has grown so much from 2013-2024 that as the portfolio value increased, it would have continuously spread across a broader range of companies. With these position sizes, the liquidity of the stocks would probably have rarely limited trading.
Of course, with a larger portfolio (e.g., a 50 KEUR start), positions would sometimes be in the 10,000 euro range, which in some companies already starts to appear as a limitation, not to mention +100-1000x sized portfolios where the meaningfulness of small-cap investing in general becomes questionable due to liquidity.
Here’s the same data from our recommendations in article form 
feedback and discussion are still welcome
We also added a comparison towards the end of the article regarding the accuracy of recommendations for company clients and non-paying companies (a small group of large caps)
Non-paying clients primarily consist of larger companies, which are often followed by 5–30 analysts, and the markets can be considered more efficient due to significant investor interest. In such cases, mispricings can be assumed to occur less frequently, and generating excess returns is more difficult. For companies that pay for analysis, analyst coverage and investor interest are often lower. Consequently, mispricings can be assumed to occur more frequently, and generating excess returns is easier.
Thanks @Antti_Luiro and everyone involved in this project! I think this is an honestly conducted review of one’s own operations, and I believe and hope that companies doing such things will also find success in the future.
However, one thing kept me wondering, namely accuracy vs. benefit. If one tries to understand how accurate Inderes’ analysis is, it is naturally assumed that the analysis of each monitored company carries the same weight. However, investors generally have two options when investing in the market: index investing or stock picking. If I invest in an index, my investment is automatically allocated according to the market value of the index companies (roughly). I would love to be able to invest in an Inderes index, which would invest according to Inderes’ recommendations, but would also do so allocated by market value. The point is that if we are looking for the so-called general benefit of Inderes’ analysis, then of course it matters whether an excellently successful analysis was done for Nokia or Fifax, simply because the amount of money invested based on that analysis is in a completely different league. I naturally believe that the quality of Inderes’ analysis does not depend on who ordered it, but I am just looking for the so-called average benefit for investors who have generally followed Inderes and benefited from it. And of course, it’s not a one-to-one comparison of a company’s market value vs. how many Inderes followers have invested in them and how much, but I think it would be a good proxy and relatively easy to adapt to all this.
Summa summarum, would you be able to run these numbers one more time so that whenever the portfolio composition is updated and rebalanced, the investments in each portfolio would be allocated according to the companies’ market values on that day? So, for example, a 2x long on Nokia in this case would mean something like a 200x long. Just to make it mimic so-called index operations. It’s possible I missed something, but I believe the necessary data is already included there with the return data, etc. I also admit that it is unfair to give so much different value to the analysis of large companies vs. small ones, but my assumption is that this is an even more brutally honest way to measure the impact of the analyses.
Thanks again for this. It’s terrible to ask for something more on top of this great research, but nothing ventured, nothing gained!
Would it also be possible to consider the volatility of an index composed of the investment universe and the “Inderes recommendation portfolio”? In my opinion, this is a crucial metric when comparing strategies.
If one invests long-term in a basket with high volatility, this may require a volatility premium in returns (i.e., higher returns). Some refer to this as beta.
For me, it would be interesting to assess whether Inderes is capable of generating alpha.
I understand the point from that paying vs. non-paying return table, it also shows that we have indeed managed to generate significantly less excess return in larger companies. A typical private investor stock picker, on the other hand, has more freedom to build a portfolio outside of large companies, so there is usually no need to weight large companies according to market capitalization, and opportunities found in small companies can be utilized better than what a typical index fund allows. From the perspective of our analysis’s accuracy, the quality criteria are also the same for all companies, even though the market seems to be more efficient in larger companies. The starting point, however, is that the team must do good work regardless of the target company, and not, for example, that the 10 largest market capitalization companies determine the quality of the team’s work and consume most of the time.
This would be a very interesting addition! However, this is not easily calculable with the current code and would also require deeper data wrangling (as well as finding a bit more extra time), so we’ll put the idea in the back pocket for a better time but from the non-paying vs. paying companies comparison, you already get a sense of how the accuracy varies based on company size (the difference is clear).
We have an external research project underway related to these same figures (this will presumably be more in-depth than our simpler calculation), so we will presumably get more insight into this theme later this year through that as well. This is independent research, so its content is in the hands of the authors, but it will presumably include the themes you mentioned, and we will probably also get external calculations of similar ‘recommendation portfolio return figures’ from it, so we can see if the results are on the same playing field as our own calculation
EDIT: By the way, we made the recommendation data used in the article openly available (link at the end of the article), so feel free to do your own calculations if you have the time and inclination
Thanks for the answer. That volatility is easy enough to calculate, at least for a model portfolio, that it could be quickly calculated, for example, with pandas, even from the return curve.
I’d like to point out that the sharing link for the recommendation history data doesn’t work for me, at least, in the original article - it asks to log in with Inderes AD credentials. Could you please test the functionality of that sharing again?
Thanks, the sharing link for the article was an older version that didn’t work, fixed now!
So, was the volatility of daily returns what you were looking for? This was 0.00914 for the recommended portfolio (standard deviation) with a quick calculation.
As an Inderes shareholder, one would naturally hope that the analyses would be exceptionally good and would have beaten the markets 100-0. And I am sure that quite significant added value has indeed been produced.
However, those returns are quite theoretical unless they are somehow categorized by liquidity. Some of the stocks followed by Inderes are so illiquid that I don’t know how sensible (it might be, but hardly with the same weighting) it is to include them in the main results of the study. When a company conducts research on its own excellence, it is good to pay special attention to such matters. Of course, the Inderes effect and trading costs were taken into account here (too high for large ones, roughly too low for small ones).
Replicating that study would be quite a significant undertaking, even though the same data is available from Bloomberg. In my opinion, the study is interesting, but it is still difficult to draw direct conclusions about the materiality and sources of the excess return based on it. Although it is clear that added value has been produced.
Thank you, these insights are really valuable to us! We will iterate based on these I don’t know if any other analysis provider has released their own recommendation data or opened up their accuracy, so we really had to develop the concept ourselves 
The intention was to provide numbers that were as fair as possible and at the same time easy to understand, especially from the perspective of recommendation accuracy. But the devil is in the details - I’ll see if I can reasonably incorporate this liquidity perspective in the next round (e.g., grouping by market capitalization probably gets quite close) My hypothesis would be that accuracy/excess returns would indeed decrease as liquidity/market capitalization (and thus market efficiency) increases.
We considered this when conducting the study, which is why we wanted to make our own recommendation data public at the same time, to provide a fair opportunity to challenge the figures. We also asked a few forum members who had previously been interested in the topic if they would be willing to do their own calculations, but the project didn’t fit at this point, so we didn’t get external comparative figures for this publication.
But if no one else happens to take on the project, it looks like a public and presumably more in-depth external study on the same topic will be coming later this year
Why is the language used in these analyses like this:
So this is about a comment from Hexagon’s analysis. “Previously we considered”, “we believe”. Why would anyone care how the analyst’s own mental processes work when the company’s stock price fluctuates? It sounds exactly like talking about the opinion of an emperor or a pope, which actually has some significance. We considered the company high-quality. The company did not meet our expectations (these words suit an emperor’s statement). Ordinary mortals say that my expectations were wrong. Does the analyst imagine being superior to the company? Is it the company’s job to fulfill the analyst’s expectations?
Quite an interesting observation. I’ve also sometimes wondered why it’s often said that a company misses or exceeds forecasts, and not the other way around, as is logically the case (because forecasts should model the company, not vice versa). This is common almost everywhere (so it doesn’t specifically concern Inderes) but it shows how wonderful the human mind is at twisting and turning things. 
Indeed, some analyses have sometimes been titled that company xxx has a “moment of truth” or “something to prove”. Sometimes it makes me smile that investors and analysts probably have just as much to prove if the company goes in a different direction than they estimated.
Below is my own analysis based on Inderes recommendations and Yahoo Finance stock data. Only companies listed in Helsinki are included. Stock data has been retrieved from the day following the recommendation date.
In the first review, the investment period extends until the next recommendation. On average, Buy recommendations have yielded the best returns, and Sell recommendations the worst. The returns for Accumulate and Reduce recommendations are also logical.
It’s important to note that returns include a lot of variation. For Buy recommendations, the return can be expected to be between -10.1% and 13.9% in 68% of recommendations. Relative to the average return, a Buy recommendation provides the best return-risk ratio among the recommendations.
A new recommendation has been given approximately every 43 days, regardless of the previous recommendation. By examining only the direction of the return, it is observed that the direction matches the recommendation with approximately 53% probability.
When defining returns according to 50, 100, 200, and 300 calendar days, more variation is visible in the returns. The Accumulate recommendation has offered the best return over 200 and 300-day periods.
There is some variation in the standard deviations of returns across different time periods, but no extremely large differences.
The best return-risk ratio of 0.32 is achieved with an Accumulate recommendation and a 300-day period. The return-risk ratio is better than always investing according to the recommendation. Of course, the recommendation can practically remain the same, in which case no changes to the investment would be needed. Trading costs have not been considered in the returns, which is worth keeping in mind. Over 200 and 300-day periods, the return-risk ratios are between 0.25-0.32, so longer-term investments seem more reasonable. Trading costs would further change the situation to favor a longer investment period.
Edit: corrected a minor error in the code and slightly adjusted the calculation of returns
