Subject: Sokker-Manager.net

Maybe this helps somebody... :P

You can use google cache to search for specific players on sokker-manager.net: like this:

site:Sokker-Manager.net "player name"
(edited)

!!! sokker-manager.net is back online !!!

https://www.sokker-manager.net

Please also tell people in your country's forum that the website is back online, thank you :)

I moved the website to a new webhost and implemented additional backup-routines. The transfer-database is now backuped on a daily basis and then placed on the old server, the new server, my brothers laptop and my own laptop. After the recent incident we were lucky that the transfer-database could actually be restored. Thousands of other customers of my previous webhost lost all their data including the backups. Anyway we are very well prepared for the future: As a compensation for all the hassles, the subscription on my old webhost got extended till middle of 2023. This means:

- There are now two redundant transfer-collectors which feed two independent databases
- If my new webhost goes down for a longer time, I could easily re-route my customers to my old webhost (at least until 2023)

Important note:
=> There might be a short downtime around 3 November, because I'm about to transfer my domain to a separate host.

So let's hope for a less troublesome future and welcome back to sokker-manager.net!
(edited)

That's even better :D

Thanks!
(edited)

Thanks :) I've waited for it :)

Please also tell people in your country-forums that the website is back online, thank you :)

Btw: I wonder if the website loads faster now. To me it feels as if it does. Would you agree ??

And report to me if anything is not working correctly yet.
(edited)

praise the lord :)

Maybe faster, although I never paid attention to it :)
I will also write on our forum soon :)

Is there anyone who speaks Spanish and is posting in forums like Argentina, Chile, Spain etc. ?

Would be nice if you could inform people in these forums that sokker-manager.net is back online. Before the downtime almost half of the visitors came from these countries, but now it seems like they still don't know that they can come back ;-). (according to Google Analytics roughly 90% of today's visitors are from Europe).

@Adivip: Thank you, would be really nice if you inform them :)
.
.
(edited)

I will start to report on the Latin forum the reopening of this great page, thank you


I hope one day to see this page in Spanish
(edited)

@Chucho: There are indeed plans to create a Spanish and a Polish version of the website. Yet I rely on volunteers who will do the translations for me. Therefore I cannot give you an exact publishing-date.

Back to similar TL feature. It’s work smooth and efficient. The Swiss touch. (really like your idea of adding a deeper search input too).

I have question about the criterias…

I enter a player and get this : detect as mid def Search criteria: Age: 23 Technique: 8 - 8 Passing: 8 - 10 Playmaker: 9 - 11 Defender: 8 - 10 Pace: 6 - 10

I wonder why the search crit range differs from -2 to +2(pace) or 0 (tech) or -1 +1? Is -2+2 not a too large range and 0 too excessive and it may affect the median price too much ?
Isn’t better relevant to only compare with scores by position /age and not skill ranges ?

Glad to see it online. I used massively yesterday for scouting. My user experience has been faster, nice job! I wrote in the Italian forum for accouncing that the site came back online!

It would be too much to post the exact implementation of the algorithm. But it works more or less like this:

1.) Based on the player's skills I calculate scores for all positions.
2.) Based on the highest position-score I assign a position to the player (in your case it was DEF MID)
3.) For each position there is a separate priority list of skills
4.) I check if there are at least 6 players that are exactly identical to your player
5.) If there are less than 6 players I set the skill with the lowest priority to +/-1 skill-point
6.) I check again if there are 6 players in the given range. If not I set the skill with the second lowest priority to +/- 1 skill-point etc. etc.
7.) Please note:
- The most important skill of a position as well as the age will always equals the original value (so I do NOT set +/- 1 point there)
- I think the highest possible deviation is +/-2
- I also consider the side-skills in certain cases
- the value of "6 players" mentioned above was chosen arbitrarily ;-)

What happened in your case?
- Your player was identified as Def. Mid, so the priority-list reads as follows:
1. Technique, 2. Passing, 3. Playmaker, 4. Defender, 5. Pace
- Since, according to my criteria, there are very few players that are really comparable to your player, the range had to open up to +/-2 points for three different skills (technique will remain stable because it is the main skill). This is the absolute worst case and it will happen less and less the more the database grows.

Your suggestion is to compare position-scores instead?
- My bet is that this can easily lead to +/-2 points deviations as well. And it's not clear whether the median prices will be more accurate then. Actually, determining very accurate weights for the skills of a position will become cruical then. And when I asked different "sokker-experts" they suggested quite different skill-weights for a particular position.
- For players between age of 21 - 33 I was able to get surprisingly accurate median-predictions based on a multiple regression analysis. Yet I eventually opted against implementing multiple regressions since there were edge-cases where the predictions were completely unrealistic and it will be hard for users to detect it (unlike when they see the skill-ranges and they realize that the ranges became really large in a particular case).
- I believe many of these edge cases could be mitigated by predicting prices via neural network (which is basically an extension to a multiple regression and follows similar principles). Yet the problem with neural networks is: I need lots of training-data and it is literally impossible to explain to the users how the engine ended up with a particular price-prediction.

Isn't there a more simple solution to your problem?
- I could just set the max skill-range to +/-1. As a consequence there would be much more searches with 0 hits. Not sure if the users will like this...
- Since you are an experienced manager, why not just use the advanced-search engine where you can freely set the paramteres as you like?:
https://www.sokker-manager.net/en/Transfers/SearchPlayersByParams
- The advanced search engine might cause some more work for you, but let's not forget that in many (if not most) cases the "comparable players"-function will deliver good enough results also ;).
- Last but not least: It's my intention that users regard the resulting median-values with caution. Thats why I list all the players that were included in my calculation. Users can still decide themselves if they believe the so called "comparable players" are indeed comparable.

Btw: A while ago kryminator implemented a "comparable players function" in his sokker-plugin. After an intense discussion with users I think they ended up with setting skill-ranges to +/- 2 by default.
(edited)

@Chucho and @ringhiostarr:

Thanks for supporting my website :-).

juytt:
"...I wonder why the search crit range differs from -2 to +2(pace) or 0 (tech) or -1 +1? Is -2+2 not a too large range and 0 too excessive and it may affect the median price too much ?
Isn’t better relevant to only compare with scores by position /age and not skill ranges ? ..."
----------------------------------------------------------------------

Wow! I just read the answer from Ramonymous and I ask myself why would somebody ask me to add also my opinion here and to give answers. Yes, somebody asked me, despite Ramonymous already gave such a thorough reply.

Well, I planed the algorithm of the score and then I got involved on the setting of the limit ranges and perhaps that's why I was asked to give more information..

So first I'll describe shortly as I can what the decisions for these algorithms were based on.

For more than 10 season I cooperated with many managers and monitored their skills-criteria.
(My criteria about choosing these managers were success with NT, with U21, or with their own club. So, some of them had experienced winning world cups, some of them were champions on their highly competitive leagues and had success with Champions Cup and some of them had success with weak U21-teams. This was my criteria, their success-experience.)

Their criteria on player skills was most times very similar but there were also quite some different criteria on some positions, even if they used similar tactics.
Slight different criteria on strikers, different criteria on wingers, etc and most differences I monitored were on the position of defMid.
My intention was to create an algorithm that fits to all of the monitored coaches. That was the aim of the algorithm of the score and that's why it is as it is.


Now, about the second matter, the ranges.
It was pretty difficult to find the golden ratio on the ranges, because there was a gap on the search results between younger and older players.
Of course this is something that could be fixed with a new range-algorithm, that takes age in account. However this is something, that will not start before the big changes on sokker.
(Only some of the new changes: (1) skills from 1-20 (now from 0-18), (2) shorter seasons. This means perhaps, that player will develop slower in one season, than now)
So, I cannot imagine a further development on this sector, till the changes in sokker have manifested.

(Insider-information: Even me, most times I prefer to use the "advanced" search)


Finally, my answer about the second question ("Isn’t better relevant to only compare with scores by position /age and not skill ranges ?")

I will give an example based on score.
There are 4 strikers. One has 12-12-16, the second 16-12-14, the third 16-17-8 and the fourth 13-14-14.
Who is the cheapest, if they have the same age and who is most expensive?
The answer depends on their age. In different age classes you will get quite different results.
Ans now I will add following information: Score on strikers is based only on pace-tech-STR and these 4 strikers have the same score.

Score is relevant for (a) for letting the machine find out by itself if it is a striker, a def etc. and (b) score can also be relevant for managers of national teams (e.g. 20years old and score over 10 can be a prospect for U21). At this point I want to remind, that scores were based on skills-criteria of successful managers.
To sum up: Score is relevant to create a category of players, who fit into some criteria.
Score is definitely not relevant for comparing 2 players.

------------------------------------------
My personal EndNote:
Algorithms, ranges and relevance of criteria can be quite different and produce different results.
These algorithms, ranges and relevance of criteria were calculated or/and defined with the intention of meeting the manager's criteria and give as accurate results as possible.
Even if one cannot accept them as they are, he can choose the "advanced" search, as I do. ;)

I couldn't log in... maybe my password with weird characters or something like that. It used to happen in Sokker Asistente.