Subject: Sokker helper - for player price estimation and prediction.

impossible to change price in € with new release

great job :)

Strange, it works when I try it.

Great tool, man :) I think it will helps many users with good deals, also a player, team estimation is easy and usefull function.

I would like if this tool would remember that i use euro. Maybe with cookies or something.

Ok, added a task in my list for this, thank you!

Thanks :)

I just deployed now, it should work fine

superdivine skills are not calculated in your program as I see:>

also when I enter skills for att player in skills panel and then put adittional numbers to his keeper (which is not important of course:P) his estimated value slowy drops . So I understand that with adittional gk skills player is started to be estimated more for keeper, even if he has 3xdivine skills for att;)
(edited)

As far as I understand the way that service works there is no clear answer to Your question. The service is neural network and prices are estimated through data mining. Data mining is probably complicated process and probably consists of subprocesses like classification (like classify this player as GK) or anomaly detection (the case You inputted is an anomaly) and it is possible that in this case player after exceeding some value in keeper skill is classified as goalkeeper, but it also may be additional rules for that and I am not sure if they may be fully controlled by ciury.

I suspect that he used some tool, framework or programing language for all that data analysis and some things cannot be controlled by him. The network must learn. The more data it has (also detailed data like 3xdivine attackers with high keeper skill) the more accurate prices will be.

However, I may be mistaken, I wrote this answer because I have nothing else (at least nothing interesting enough) to do :P

Thank you kryminator for explaining the process. It's pretty close like you desribed it, only that there are no special rules or magic numbers. It's like a black box which has inputs for each parameter, and it must give a single output (or a function). Using a neural network you can give both inputs and output so that when you call the function only with inputs, it will give you an output which is close to the previous output. Over-simplified, with one input and one outupt, it's like a simple function F(x) = y. You basically give a lot of points (x,y) to the program, and what the program does is it aproximates a continuous function for all these points, and when you try to calculate f(x) for every x you can think of, it will aproximate with the value of y for that function. Now, depending on how scattered are the points which are used for learning, and how you train the black box, the output can differ. You can have a more smooth function, but in these case the "edge" cases will not be very well aproximated, or you can have a "pointy" function, and in this case based on bad values (far away points) the aproximation could differ. Of course, if all the training points are correct, and not scattered, everything should be fine. But they are pretty much scattered, because prices differ very much especially for young players or cheap players. Young players prices are very oscillating because of human emotions, bid wars, etc... and cheap players are often bought by beginners which overestimate the prices, or by users who have two teams and are doing illegal transactions. For example for a cheap player the estimation could be 20 000 eur. But this might mean that this kind of player either sells for 40 000 eur because of what I described earlier, or doesn't sell at all. So cheap players have very oscillating prices. Very young players - the same. You can check similar transfers to see how much the prices are oscillating for the same player.

"So I understand that with adittional gk skills player is started to be estimated more for keeper"
I guess something like this happens, because in reality there are very few players with high keeper values and also other high levels, so instead of the player getting closer to a real transfer, it gets closer to a keeper transfer probably, which was done with less money.

There is no hidden formula, or special rules.

The same probably can be noticed with stamina for some particular cases.

The only conclusion which I would make from price dropping when a skill increases is that that skill is not important for that particular set of skills (not really that the player is less valuable).

Grat job ciury! Nice tool ;)

What do you think about setting page to default currency € ? It's more common than lei ;) Currently when you link your player with predictor,everyone who opens it, gets prediction's in lei and i think many of those user's don't know what is lei currency rate and they dont bother to check it.
(edited)

LEI is the most important currency in the world, everybody shoud have heard of it and it means LIONS :)

Thanks, good idea :) I'll do that with the next deploy.

link on first page /predictor is now wrong since you updated page :)
suggestion: when i change currency i have to refresh the page for it to happen. can you make it auto refresh? or do what was suggested and give cookie of favourite currency (but i have a cookie killer :( )

whatever tweaks you did to calculations for players is strange. one of my best mid traineess was valued at 6M. now he is only 2.7 :D
all my players have gone strange ways. most of 19-20year olds are accurate now though
i know it is based off previous transfers so it will take more data to get an idea. keep it up!
awesome work and great site layout too.

nom.
Total value: 87 357 058 €
Total players: 20
Avarage value: 4 367 853 €

The default currency is now eur, and changing the currency also refreshes the pagenow.