Guild Wars 2

It recently occurred to me that if you cannot accurately predict pro football games even with all the stats they have…

…how can we create an Expected Win prediction that is accurate enough to fairly award ladder points?

How accurate are the BEST football predictions? Consistently.

Isn’t it something like 60% accurate at predicting which team will win on any given Sunday?

most games for me are like Green Bay(me) vs Highschool freshman B team (enemy)…

Laserbolt.6731:

It recently occurred to me that if you cannot accurately predict pro football games even with all the stats they have…

Sure you can. Those odds, lines, and spreads you see all the time…those are predictions, and if they were not set accurately enough in the first place, then the people hosting bets would lose money.

That said, the purpose of these odds is not actually to predict a winner/point difference. It is to ensure an even volume of bets occurs on either both sides so the house is guaranteed to win on the vigorish. This probably demands at least as much precision, if not more, than simply predicting the margin.

The fact that you generally can’t beat the house in the long run with your calculations means that the house has already taken into account all the stats that you might consider, and then some.

How accurate are the BEST football predictions? Consistently.

Isn’t it something like 60% accurate at predicting which team will win on any given Sunday?

That says a lot more about the relative parity of teams than about the inaccuracy of the predictions.

It’s not that hard to predict that a team has a 60% chance of winning. But that also means there’s a 40% chance of losing, so you shouldn’t call the system broken when that 40% happens.

The problem with GW2’s matchmaker right now is two-fold

• They don’t really have stats to go off of since the last patch threw a giant wrench into things.
• It’s pretty much inactive at the moment since Justin took off for the weekend.

Dave,

I meant there is no computer calculator that accurately predicts pro football games at the accuracy GW2 needs for FAIRLY awarding Ladder points from it’s grid.

I used fantasy football because it somewhat parallels the GW2 situation where you try to predict games using individual player stats that were gathered with them all playing on different teams.

Pro football is difficult to predict because teams are relatively equal to each other.

In one of those games the computer could easily say that each team has about 50% (or even the much more lenient 40-60% that GW2 would use) chance of winning.

You’ll find that looking at a combination of odds/predictions as well as the confidence of those predictions will return exactly the information you want, and a very equivalent parallel to what the matchmaker will aim to do.

tl;dr—when a computer says it’s 50% confident about a pick, it’s saying something close to “I’m fairly sure these teams are evenly matched”, rather than “I’m unable to make any conclusions of value”.

PS: there’s no argument from me that the current parameters may not be optimally precise yet. I suspect they aren’t. But the gist of your post seems to argue that they could never be precise enough, which is not the case.

Dave,

I’m seriously wondering if can they really predict, with 95% confidence level, that Team A has exactly a 20-39% chance of winning… based only on the MMRs of the players (which were calculated in matches full of confounding factors like 4×5s, afkers, teams on Teamspeak, groups playing with each other against them and knowing each other;s typical roles!)

I guess we’ll find out when they start adjusting the numbers of the predictor on the matches that already happened. My best guess is that they aren’t going to be able to make it predict the previous games all that accurately. I HOPE they can, but I have my doubts. Too many random factors in the creation and adjustments of MMRs.

Laserbolt.6731:

I’m seriously wondering if can they really predict, with 95% confidence level, that Team A has exactly a 20-39% chance of winning… based only on the MMRs of the players (which were calculated in matches full of confounding factors like 4×5s, afkers, teams on Teamspeak, groups playing with each other against them and knowing each other;s typical roles!)

Whether MMR in GW2 can stabilize and become a good representation of actual ability remains to be seen. On the other hand, 95% confidence is also something not necessary here, as the cost of being a bit off is also not that high (and expects to average out over the long run so long as the inaccuracies are not non-random).

But in theory and in general, this concept can work very reasonably. All you need is a variable that corrects and stabilizes itself over time and a reasonable expectation that this variable is a good metric for determining ranking/skill/ability/experience/etc.