Hugh wrote: 

We stayed on topic until Captain Anti-context insisted we had to mean what he wanted aggressiveness to mean, and then decided that the reason for keeping track of durability was to reward runner-ups (granted he was spurned on by the humor of another poster - context and humor were both lost on him).

-H

uhhh...wait. was this a off-handed slap?  

Hey risky, lend me some crap to fling! 

weathertop wrote:

uhhh...wait. was this a off-handed slap?  

Of course it was!  It wasn't aimed at you, though.  

Mongrel wrote: Negative aggression makes sense though, for being both aggressive and silly. So I think -6/7 is a fine assignment to that scenario.

Disagree. The stat doesn't measure good ideas, it measures aggression. It's not a value judgment. The majority of people near the top of that stat may well be bad players.

asm wrote:

Mongrel wrote: Negative aggression makes sense though, for being both aggressive and silly. So I think -6/7 is a fine assignment to that scenario.

Disagree. The stat doesn't measure good ideas, it measures aggression. It's not a value judgment. The majority of people near the top of that stat may well be bad players.

I on the other hand agree with mongrel. it will be a separate ranking with nothing to do with if a  player is good or not just who is the most aggressive

what i think would be a good idea especially if you want to keep a Ranking board fairly accessible for new player is have a weekly ranking board, best players of the past week or most points earned... something like that. 

KrocK wrote:

I on the other hand agree with mongrel. it will be a separate ranking with nothing to do with if a  player is good or not just who is the most aggressive

This means you agree with me.

asm wrote:

KrocK wrote:

I on the other hand agree with mongrel. it will be a separate ranking with nothing to do with if a  player is good or not just who is the most aggressive

This means you agree with me.

well then I'm just confused. i thought you were saying that a ranking based on aggression would have no merit due to the fact that shitty players may be on top of the ranking.

*note to self: don't smoke and type ... you get confused to eazy... oh look a dog with a puffy tail*

No, I was sort of saying the opposite. The aggression stat would be cool on its own and have no bearing on "good" or "bad" play.

My contention was that taking one person out in an 8-player game and then getting killed because of it should advance your aggression rating, not depress it. Precisely because the aggression stat doesn't distinguish between "smart agression" and "stupid aggression." Mongrel was saying that in that scenario, the aggression stat going down would be justified because it was a bad move.

Unfortunately, aggression is influenced by many factors (elimination bonus, cards, dice, number of players remaining). I might go for an elimination that has higher incentive (more cards, ability of others to chop me down, etc.).

So an unbiased aggression stat should detect the amount of attacking relative to player size, with an emphasis on units used for elimination.

That said, if we want a "pure aggression" stat rather than "quality aggression" stat, how about this?

Let n= the number of players remaining.
calculate, on a turn by turn basis

log_2(n-1)*(# of units spent by player X eliminating someone during the turn) +
1*(# of units spent by player X attacking during the turn)
---------------------------divided by------------------------------------
(# of units for X at start of turn, including X's bonus)

add up these numbers up and divide by the total number of turns taken by player X. Then average over all games.

This measures what % of your forces you were willing to commit to taking someone out, boosted by one's proclivity for attacking early and often - two aggressive traits to be sure. The log_2 factor is incorporated to dampen the boost from eliminations.

To calculate this stat, I think one backwards pass through the history would work.

Other things done right.
1.) Keeps "aggressive play" and "quality play" as uncorrelated as possible, and I think addresses what asm wants.
2.) Handles two player games nicely (no bump for eliminating)
3.) Not awarding much for cleaning up someone else's failed elimination.
4.) Not sensitive to the steady unit growth of the board, only the relative sizes of eliminators and eliminatee's troops.
5.) Denominator deals with "return to place from attack" games
6.) Accounts for dice mods. In the angels and demons board, with 6v8 dice, one needs far more units to eliminate say 40 units than in typical 6v6. But this is reflected by X spending more units to eliminate.
7.) accounts for aggressive, but failed, eliminations.
8.) accounts for players who just like attacking.

Hard to do an example with this one, but I think the numbers make sense. Also, I edited the hell out of this note, which was the result of talking to myself for about 3 hours, so sorry if it's dense.

I had a "things not done right" list but continued tweaking the formula until I could move those items to the other one.

Trying to post the formula again

(log_2(n-1)*(# of units spent by player X eliminating someone during the turn)+ 1*(# of units spent by player X attacking during the turn))
----------------------------------------------------------------------------------
(# of units for X at start of turn, including X's bonus)

Well that's a little better.

Quite excellent Mongrel. In particular, I like the term log_2(n-1) * (units spent eliminating) in the ratio. It does do well with "obvious" takeouts.

The other term is interesting. If there wasn't a lot of variation in takeout success rates (ie if everyone had about the same kills per attempts ratio), then we could dispense with the (# of units spent attacking) term.

As a hybrid (takeout + regular aggression) stat, one issue is that it would vary quite a bit from board to board as to how much contribution you get from each term. As it stands now, I suspect the units spent term would dominate on most maps. With proper weighting constants in front of each term, this problem would disappear.

Mongrel wrote: Trying to post the formula again

(log_2(n-1)*(# of units spent by player X eliminating someone during the turn)+ 1*(# of units spent by player X attacking during the turn))
----------------------------------------------------------------------------------
(# of units for X at start of turn, including X's bonus)

Let E= units spent eliminating during the turn

      A= units spent attacking during the turn

      U = # of units player X has at the beginning of the turn, including bonus (and any additional armies placed from elimination bonuses)

      n= number of players remaining

Then it should look like

log_2(n-1)*E + A

----------------------

           U

then average over all turns.

I don't like the 2nd term in the numerator that much either, but leaving it out makes #7 and #8 a problem.

I think we can do a mock scenario. Suppose a 6-player game goes 10 rounds before I get eliminated, and each turn, I attack with 25% of my standing army( after bonuses). In round 4, there are 5 players left, and I eliminate the 5th man standing (using the 25% of my forces), and In round 10, I eliminate the 3rd man (same thing, 25%). So A = .25U every turn, E = 0 for all rounds except rounds 4 and 10, when E=.25U

.25,.25,.25,.75,.25,.25,.25,.25,.25,.50

aggression coeff = average =.325

If instead I failed on both eliminations, the aggression coefficient is .25

I definitely agree, that the stat would be very sensitive to board conditions, especially the number of players. In classic risk, with escalating cards sets, I might be handed a gift of 5 cards from a failed elimination, and sweep through the map with the card cascade. Not aggressive, though I did a ton of attacking and eliminating (and would register a high aggression). Something should be done about this.

Furthermore, as you said, there are too many "boring" rounds that dampen the special rounds where an elimination happens. We could average the aggression for the highest 3 rounds and the lowest 3 rounds, or do something like that.

I think we can take care of the board sensitivity problem. Once we agree on some measure of "universal aggression", we could normalize, on a board by board basis. Average the aggression rankings from all players for pairs (board X, number of seats),  then divide your aggression rank by that.

So that your overall aggression would be similar to the G rating

(#games played on (board X,m seats))*(normalized aggression for (board X, m seats))+ ...

-----------------------------------------------------------------------------------------------------------------------------

                                                      (total # of games)

It's OK that this formula is involved, a refined aggression formula should require a close look at the attacks each turn. If someone decides to attack heavily into me for a round, that shouldn't be recorded as me not being aggressive. It's all about what you do with your units during your turn.

I think the round-by-round formula can be improved, but I have to make dinner.

Just went through the thread, and practically all points of consideration were discussed before my rambling. Credit where it's due.