The data samples come from two different GWs. I've probably analysed stats for more than 800 GKs in each of the GWs.
The results are the same for both GWs.
This is where you want to say that my numbers are wrong, or that this just doesn't make sense. But the numbers are right. Still, I'll admit. It doesn't really make sense.
The most important attribute for not conceding goals for a GK is Acceleration.
Yes, I'll repeat that.
The most important attribute for not conceding goals for a GK is Acceleration.
I have no idea why. But then again, I just crunch the numbers and act from there. Someone else can probably tell me why Acceleration is so important for GKs.
Here is the full list from my research for attributes for GKs for not conceding goals.
If someone can tell me why Acceleration is on top, please let me know..
I can only guess it helps with coming off the line in one-on-ones... does seem a bit strange.
ReplyDeleteGood to see tendency to punch down where it should be... I always thought eccentricity had a more negative impact though.
Actually teamwork is a very interesting one too... it's the most loner position on the field?
ReplyDeleteMaybe it has something to do with how fast the GK moves sideways? Think of a ball heading towards the top corner, what is it that makes the GK take that extra step so he reaches up to catch the ball?
ReplyDeleteTeamwork is actually the odd one out, because no GK have more than 5 in this attribute. Ultimately it means that if teamwork is 5, the keeper concedes less than if teamwork is 1. I wouldn't put too much into that attribute tbh.
the thing about eccentricity that I have wondered is that yes sometimes an eccentric makes bonehead decisions. But sometimes it seems on the flip side they could get so hyped up they play blinders.
ReplyDeleteAnecdotal and biased I know but could be tested I suppose.
Andreas this is one of my favourite blogs, im sure loads visit
ReplyDeleteMaybe acceleration is used to determine how quickly a GK will react to an event in conjunction with the reflexes, even on his line ?
ReplyDeleteA GK with poor acceleration and good reflexes would still be too late to reach the ball because of a late / slow start ?
And please keep posting those things, it's easily one of the best blogs on FML.
IMO you should work with SI to point out the flaws of the ME and evaluating new releases !
Thanks for good words!
ReplyDeleteI just had a look on the list and saw that stamina was quite high up there for GKs.
I think the reason might be that stamina is one of these attributes closely corrolated to CA. So if stamina is high, GKs concede less because they are generally better. If stamina is low, they concede more.
At least that is my theory!
Yea I dunno about how useful 5 vs 1 TW could be but who knows.. ACC makes sense to me because that is how fast a GK get go from still to moving which is very important.
ReplyDeleteGreat blog m8, I check it daily!
Loving this blog m8, keep up the great work, can't believe how some of these stats pan out!
ReplyDeleteYou're missing a massively vital point, and that is the amount of chances created against the goalkeeper.
ReplyDeleteThis analysis is like me saying that the house has no advantage in Blackjack if you take out all the cards valued at 10 (it actually gives the player a small advantage, due to going first) - since that isn't a scenario that will ever actually occur. Neither will any goalkeeper be playing in any match alone against an opposition team.
Goalkeepers don't play alone. If GK A has conceded 1 in 10 games and has Acceleration of 15, whilst GK B has conceded 5 in 10 games and has Acceleration of 5, we can assume that Acceleration is more important. However, what we don't know about GK B, is that he's faced 10 penalties in 10 matches as well as an above average amount of shots on goal, whilst GK A has faced 0 penalties and a below average amount of shots on goal, thanks to being behind a 5-4-1 formation. With this information, Acceleration becomes negligible and other stats may become more prevalent.
If you could test with all penalties (saved and missed) removed, add percentage of shots saved to the data and test with all goalkeepers being put behind the same formation, you'd be on to something. However, you'd still be short as you'd need to ensure that all chances against all goalkeepers were of similar quality. A goalkeeper with a low average of goals conceded per match that faces lots of one-on-ones is liable to be better than one with a lower average that only faces shots from outside the box.
Unfortunately, I don't have the luxury of obtaining information about chances created against the GKs.
ReplyDeleteSo I pick the next best thing - goals conceded.
The thing is though. If the sample size is large enough, these random factors (theoretically) cancel each other out.
Or to use your example. If GKs A-Z each have conceded 1 in 10 games with Acc 15, and GKs AA-AZ each have conceded 5 in 10 games with Acc 5, you can still assume that Acc is influental in determining whether GKs concede goals or not, irregardless of the number of chances that have been created against them.
I respectfully disagree. The sample size is far too small to rule out a standard deviation being in play even when using "the next best thing" as the main part of your study.
ReplyDeleteIf you had data on 100,000 goalkeepers, then maybe you'd be in business, but on the small amount that actually play enough games for the sample to be classed as relavent, you could just be looking at a peak for the GKs with high acceleration or a trough for those that don't.
Yeah. You bring up some good points, but I still think there is something to my findings.
ReplyDeleteThese results are derived from two independent data sets. Known GK attributes such as reflexes and aerial ability score high, others such as kicking the ball and throwing the ball score low. The standard deviation for Acc in both data sets is relatively low compared to other atts.
Ultimately though, I crunch a bunch of numbers and write up what I think are some interesting findings in my blog. You can accept them or you can choose not to. Either way, I'm not saying I know the truth or have the solution to anything, I just do this for fun and I believe in what I do, otherwise I wouldn't post this! :)
Still, I appreciate any feedback and critisism to what ever I come up with. I don't take it too hard, I'm not even a statistician by trade! ;)
What are the sample and population sizes? If you give me those I can do a quick calculation and let you know how valid your analysis is. It's what I do for a living heh
ReplyDeletethen you probably understand more of this than me! :D - http://img339.imageshack.us/i/ishot52.png/
ReplyDeleteThe other data set was probably around the same size, so around 750 GKs for each observation.
Saw a unlocked keeper goes for 15.5K wages this morning (2nd highest for the time being in Sanchez).
ReplyDeleteThis GK has 20 Acceleration / 17 Reflexes / 17 Concentration and 16 Aerial, so maybe one follower of your blog decided to test your findings :).
In addition this GK is 21 and has reputation of 4*, so if he has a good potential in addition, it could be a steal. Will try to see hos he fares !
Ye please keep us updated on him, given these numbers, he's likely to make quite an impact.
ReplyDeleteAlso, I feel a bit sad I didn't notice this blog before, it's brilliant Andreas.