Gaining Part 5: Superstitions & Chance (03/03/2014 23:44)

"The root of all superstition is that men observe when a thing hits, but not when it misses.” 

Just a quick post now, after the long one earlier today about how lineup age might affect player gains. I'm going to wrap up my look at player gaining by addressing the remaining two mantras on that list I posted when I started this blog. Not because they're especially convincing on their own, but because they're illustrative.

We've all been in the situation where we're playing a few friendlies and just not getting anything. We reload ml-reports and that big fat 0 is staring accusingly at us. Fine, it happens every now and again, but if you just played three friendlies in a row and got nothing for any of them?! Or you got two goalkeeping stats and nothing else from the last five games? Surely you must be doing something wrong, right? Time to change the formation, throw some different players in there quick, before you use up all your friendlies for nothing! As several guides I've read have suggested, perhaps you should even take a break and come back to the friendlies in fifteen minutes, or an hour.

Well, we can check this pretty straightforwardly. If this is true - that the way your team is setup is stopping you gain, or you just get onto a hot streak and start piling up the attributes - then the gains I earned on a match should predict the gains I earned on the next - because I keep the same formation and just rotate players out one by one as they lose fitness, and I often play the same sides consecutively because I'm playing sets. So, I checked the correlation between each match I played in my database and the one that came straight after. The sum effect of all those similar factors? Playing often against the same teams, with pretty much the same formation and players? Nothing. The R-squared is 0.0024. We can predict less than one quarter of one percent of the gains in that next match - the other 99.8% is effectively just luck. So just because you get no gains this match, doesn't mean you'll get nothing in the next one. Or that you'll get loads! You've got pretty much the same chance as you always do.

To understand why sometimes it seems that can't be the case, it's worth very quickly considering how statistics work. Gaining attributes is a random event with two properties: We know roughly how often we should gain an attribute, on average (e.g. we might expect a player to gain an attribute 10-15% of the time he plays a friendly) but we can't predict exactly when it's going to happen. When you have several players on your team this means you could get no gains in a match, or one, or two or several. Such a process is well described by a Poisson distribution. If you know the average rate at which events occur - e.g. 1.4 gains per friendly - then you can easily calculate the probability of getting zero gains, or of getting one, or 2... etc.

So even if you average 1.4 attributes per game, that doesn't mean you're always getting 1s and 2s. You should get +0 on about a quarter of your games, +1 on around a third, +2 on another quarter and really good gains of 3 or more on the rest.

Now, even though your average gaining rate is pretty good, and netting you close to 300 attributes a year from friendlies, you'll still be getting the dreaded +0 for fifty games a season! And you'll be getting two blanks in a row about 12 times on average. And three blanks a row should happen 3 or four times a season too. In fact, you're more likely than not to get four friendlies in a row with no gains at all at some point during the season. And as awful as it seems at the time, that doesn't mean you're actually gaining badly or doing anything wrong - you'll just get a +5 a couple of times, or a bunch of +4s to balance things out. This is just the way chance works, even when there's no difference between the games you get 0 for, and the ones you get +5 for. The problem is, you'll remember that run of zeros because it seems so improbable and meaningful - and you'll forget all the other times a zero was followed by some gains, because that doesn't seems so surprising. That's what we all do, it's just the way our brains are wired - to spot patterns. We're not so good at figuring out how unexpected and meaningful that pattern really is though, which is why a basic understanding of statistics - not a bunch of equations, just an appreciation for how often totally random events can seem meaningful - is really helpful to have, whatever your job. 

So don't put too much stock in how much you gain on a handful of games. Get nothing from that friendly set with team X, but +6 against team Y? Get no gains on your prize midfielder for ten matches? It's very tempting to conclude that playing against team Y is better for your players, or that you should avoid playing against team X or similar sides, or that you should swap your youth's position with that old player who gained +3 - but you just can't tell that from so little data. Those are the sorts of things you'll see from time to time, even when those two teams are identical, and your youth is actually a great gainer. Instead, you should look at hundreds of games, or hundreds of players - that will start to give you an idea of what really affects your gaining rate, and how much. Or if you don't have time to do that, just check my results!

Oh, and the pumpkin thing isn't true either.

 

- Belizio

 

 

Click Here to return to the Wonderstar Analytics home page