Obviously if you are predicting just win loss or draw (as in Fussball
Toto, Quiñela, Football pools etc). it is quite simple. Even if you
added a dimension of over/under for goal difference that just gives five options instead of three.

But if you want to reward exact predictions of score as opposed to just
geting the result right, you run up against the issue that for a game
that ends 5-5, one could argue that a prediction of 5-4 or 6-5 is
actually better than 0-0. Jesper's scoring system does try to address
this in part by giving points for predicting correct goal difference and
goals scored by away and home team.

But what about doing it differently. Suppose you allocate 10 points per
game. Getting the score bang on gets you 10 points (is 10 too many ?). Anything other that exact score results in deductions.

Minus 3 for getting the result wrong. (is this too few ?)
Subtract points for the difference between predicted home team goals and
actual goals scored.
Subtract points for the difference between away team goals and actual
goals scored.

So in my example, 5-5 draw, someone who predicted 0-0 would get zero
points. Some who predicted 4-4 would get 8 points, 3-3 six points, 2-2
four, and 1-1 two. Someone who predicted 5-4 or 6-5 would get six
points. Since predictions of very high numbers of goals are likely to be
rare, Zero points scores should not be that common.

Any commments or suggestions along those lines ? Proposed improvements ?

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, August 15, 2022 at 6:53:18 PM UTC+1, MH wrote:

Obviously if you are predicting just win loss or draw (as in Fussball
Toto, Quiñela, Football pools etc). it is quite simple. Even if you
added a dimension of over/under for goal difference that just gives five options instead of three.

But if you want to reward exact predictions of score as opposed to just geting the result right, you run up against the issue that for a game
that ends 5-5, one could argue that a prediction of 5-4 or 6-5 is
actually better than 0-0. Jesper's scoring system does try to address
this in part by giving points for predicting correct goal difference and goals scored by away and home team.

But what about doing it differently. Suppose you allocate 10 points per game. Getting the score bang on gets you 10 points (is 10 too many ?). Anything other that exact score results in deductions.

Minus 3 for getting the result wrong. (is this too few ?)
Subtract points for the difference between predicted home team goals and actual goals scored.
Subtract points for the difference between away team goals and actual
goals scored.

So in my example, 5-5 draw, someone who predicted 0-0 would get zero
points. Some who predicted 4-4 would get 8 points, 3-3 six points, 2-2
four, and 1-1 two. Someone who predicted 5-4 or 6-5 would get six
points. Since predictions of very high numbers of goals are likely to be rare, Zero points scores should not be that common.

Any commments or suggestions along those lines ? Proposed improvements ?

Someone who's correctly predicted that the game would be a draw should get some points regardless of what the score was, I think.

I've actually invented a scoring system for predictions. I've used it occasionally on rss before. So now you've posted this message, I might as well officially introduce it:

Exact score: 100 points
Correct winning margin but 1 goal out (eg you predict 2-2, it ends 3-3): 90 Correct winning margin but 2 goals out (eg you predict 3-2, it ends 1-0): 80 Correct winning margin but 3 goals out (eg you predict 4-4, it ends 1-1): 70 Correct winning margin but 4 goals out (eg you predict 7-0, it ends 11-4): 60 Correct winning margin but more than 4 goals out (eg you predict 2-0, it ends 7-5): 50
Correct winner, but 1 goal out (eg you predict 0-1, it ends 0-2): 85
Correct winner, but 2 goals out (eg you predict 2-3, it ends 2-5): 70
Correct winner, but 3 goals out (eg you predict 1-4, it ends 0-6): 60
Correct winner, but 4 goals out (eg you predict 5-2, it ends 7-0): 50
Correct winner, but more than 4 goals out (eg you predict 8-0, it ends 4-1): 35 No correct winner, but 1 goal out (eg you predict 3-2, it ends 3-3): 30

So for your 5-5 draw example, a 0-0 prediction would get you 50 points, a prediction of 5-4 or 6-5 would get you 30 points.

I haven't thought much about what to do about matches that go to extra time and/or penalties.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

MH escreveu:

Obviously if you are predicting just win loss or draw (as in Fussball
Toto, Quiñela, Football pools etc). it is quite simple. Even if you
added a dimension of over/under for goal difference that just gives five options instead of three.

But if you want to reward exact predictions of score as opposed to just geting the result right, you run up against the issue that for a game
that ends 5-5, one could argue that a prediction of 5-4 or 6-5 is
actually better than 0-0. Jesper's scoring system does try to address
this in part by giving points for predicting correct goal difference and goals scored by away and home team.

But what about doing it differently. Suppose you allocate 10 points per game. Getting the score bang on gets you 10 points (is 10 too many ?). Anything other that exact score results in deductions.

Minus 3 for getting the result wrong. (is this too few ?)
Subtract points for the difference between predicted home team goals and actual goals scored.
Subtract points for the difference between away team goals and actual
goals scored.

So in my example, 5-5 draw, someone who predicted 0-0 would get zero
points. Some who predicted 4-4 would get 8 points, 3-3 six points, 2-2
four, and 1-1 two. Someone who predicted 5-4 or 6-5 would get six
points. Since predictions of very high numbers of goals are likely to be rare, Zero points scores should not be that common.

Any commments or suggestions along those lines ? Proposed improvements ?

Sounds like an interesting idea. I would probably award a minimum of points, though, for getting the result right. In your example above, even if 5-5 is quite "far" from 0-0, it is still a draw nonetheless, so it probably deserves a little award for that. Perhaps 1 or 2 points?

For what it's worth, here's what we do at our offline pool:

10 points for nailing exact score;
7 points for the winner AND the amount of goals of one of the teams;
5 points for getting the correct result (ie, right winner or draw), but
not the amount of goals of any of the teams;
2 points for nailing the amount of goals of one of the teams (but not the result)

So, say that in the upcoming World Cup you predict Canada 2-1 Belgium.

If Canada wins 2-1, you get 10 points;
If Canada wins, say, 2-0 or 3-1, you get 7 points;
If Canada wins, say, 3-0, you get 5 points;
If it ends in a 1-1 or 2-2 draw, or Belgium wins (0-1, 2-3), you get
2 points.

Any other draw or win by Belgium in which they don't score only one
goal or Canada two, you get zero points.

If you predict draws, things are simpler. Say, you call USA 1-1 Wales,
you get 10 points for a 1-1, 5 for any other draw. If either team wins,
you get 2 points if any of them scores exactly 1 goal.

We don't account for goal difference, though. In your system, if I got
things right, for the Canadian example we'd have something like this:

10 points if Canada wins 2-1;

Other Canadian wins:
9 points for a 2-0 or 3-1;
8 points for a 1-0, 4-1, 3-2 or 3-0;
7 points for a 4-2 or 4-0;
6 points for a 4-3, 5-2, 6-1 or 5-0;

Draws:
6 points for 1-1 or 2-2;
4 points for 0-0 or 3-3;
2 points for 4-4;

Belgian wins:
5 points for 0-1, 1-2 or 2-3;
4 points for 0-2, 1-3 or 2-4;
3 points for 0-3, 3-4, 1-4 or 2-5;
2 points for 0-4, 1-5 or 3-5;

This is interesting. I was thinking about tweaking the 3-point deduction
for deviation from the exact score, but it seems like it's not really necessary? I'll try to retro-fit your scoring system to my Conmebol WCQ predictions and see what I get.


--
Lléo

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)