Hopefully I'm not just failing at forums as I couldn't find anything addressing this, but are we likely to see some sort of luck statistic included in our game view? I find it helpful in evaluating whether I'm using a bad strategy or if I just had some bad rolls.  Plus I take some sort of strange comfort in seeing a giant negative number when I'm frustrated with a game haha.

Thanks!

Yes, that feature will be added at some point - I agree it's reassuring to know that it was just bad luck that lead to a loss.

LOL, this is so funny. I never would have thought of such a thing, but could be quite interesting. Can't wait to see the results.

paulharrow wrote: LOL, this is so funny. I never would have thought of such a thing, but could be quite interesting. Can't wait to see the results.

WF actually had a pretty neat plug-in called luck-stats which showed how you did both offensively and defensively vs the statistical average and gave you a total score which represents how many more or (less) units you have than you should.  Here is an example:

http://rippleeffectgames.com/warfish/luckStats.htm?gid=64989043&viewer_pid=dere20070427190509

As you can see I am in the lead with the worst luck, and that all of it has come on offense where I have lost about 16 more units than I should have.  On defense, however, you'll see I'm neither up or down vs where I should be.

I know there were some other luck plug-ins out there, but this is the one I found to be the most comprehensive and useful.  It's often easier to look at the luck stats and see that I got decimated on defense rather than going through the history to figure out what just happened since my last turn lol.

THe problem with the WF stats is they were inaccurate for anything other than 6 sided dice attacking and defending. Having said that it was always good information.

Let me know when you want to start working on this tom, I cracked the modified-dice-luck-stat-nut a few months ago.

Thanks will do

Hey Tom, Where is this on the docket?

I'm curious 11's.. What is your solution? Can you describe how it works in simple terms?

I was thinking about it this morning (before finding this thread) and came up with what I thought might be an elegant if perhaps too simple a solution.

The basic premise: For every single die that is thrown, a number is generated that describes its strength as a percentage of the max possible throw. With this method, a single number can be easily generated for a player, that player’s opponent, and a combination of both (kind of an aggregate luck score).

Example:  7 sided vs. 5 sided (3 dice thrown against 2)

Attacker 6,2,2, vs. 5,3 resulting in both losing 1 army.

Attacker luck =  10/21 = .48
Defender luck =  8/10 = .8

Aggregate = Attacker luck – Defender Luck * 100 = -32

Notice that the individual luck stats max out at 100 but can never reach 0.  It follows that an aggregate can never be 100 or 0.

I don't think it really matters whether it's 3v1, 3v2, 2v2, etc., (you're either throwing 'em well or you aren't) but you could organize it this way as well.

Any of these numbers could alternately be expressed as a standard deviation type of number (maybe times 100) at the end of the game (or at any point in the game). 

I'm not good with the math, but I do note one objection. Summing the dice rolls before calculating luck loses some necessary granularity in the results. A defender throwing 6,1 on a 6-sided die is very different than throwing 4,3.

..in what way? Is it less lucky, or more lucky? I mean, it's just a stat, and it's not just about what, but when. For instance, when you look at luck stats over time, there comes a point when you're losing that it doesn't matter if your luck stats go through the roof, you're still going to lose.

Yeah, I'm not sure. I was going to expand my point until I realized that whether you would prefer 6,1 or 4,3 would be essentially totally context-dependent. I do maintain that they're not equal for the purposes of determining good or bad luck, though.

I was looking at my model and realized that the numbers for individual and aggregate luck did not correlate.  If 0 is expected then given the scenario I originally described:

Attacker luck = (0.50 - 10/21) *100 = -2

Defender luck = - (0.50 - 8/10) *100 = +30

Aggregate = - (Aluck + Dluck) = - 28

When asm mentioned the word granularity, my thoughts went immediately to resolution, which is a way of looking at luck over time..

The resolution of my example would be 1 (battle),  but you could change the resolution so that when you look at a point in time, you are looking at that roll and a number of rolls that preceded it. To keep that number relevant, you might need to multiply it by something like the resolution number itself (or perhaps a percentage of it).

asm, in that the effectiveness of a any given roll is context dependent (I agree with you on your point - for instance, if you are attacking 2v2 and you absolutely need to win the battle, throwing 6,1 is infinitely worse than throwing 3,4.) but I think it's doubtful that someone can ever come up with a method of quantifying and weighting context.

Regardless, I am perfectly willing to accept that it's not such a good idea.  I'm just floating the balloon.  At the very least it presents a simple way of dealing with different sided dice.

Unfortunately, I think the stat that M57 is proposing is meaningless. It is related to how lucky you are with high rolls versus low rolls with respect to how lucky the other player is, but not quite this. [to be better, the expected sum for two 6-sided die is 7, so this should correspond to 50% for the defender, similar for others situations..., but not sure this would be of use either.]

The luck stat I think that we are looking for is whether or not you won more attack/defense rolls than you were expected to, not whether you had unusually high or low die rolls compared to opponents (although they are related).

I have done lots of calculations to try a create a simple example of why it is meaningless and decided they are all to complicated to explain/write up, although interesting. The heart of the issue is this. For M57 stat (attacker/defender 6 sided dice), two sixes for the defender is counted as equal to 3 sixes for the attacker (aggregate = 0). These rolls are not equal, two sixes for the defender always wins, three sixes for the attacker does not.

*edit posted without reading 13, although the change doesn't effect it's usefulness in my mind.

Alpha wrote:

The luck stat I think that we are looking for is whether or not you won more attack/defense rolls than you were expected to, not whether you had unusually high or low die rolls compared to opponents (although they are related).

I don't doubt you are right Alpha.  A basic average calculation will not explicitly give information about expected wins or loses, but as you said, there is a correlation.  My idea is a simple way to answer the question, "Am I rolling 'em bad, good or as expected? ..and what about my opponent(s)?"

Personally, that's really all I care about, and I don't see how this information would be meaningless. Expressing luck in terms of expected wins would be much more complicated to calculate, more complicated to understand, and I'll bet the aggregate results wouldn't be all that different from a basic mean calculation.  Certainly, a more complex method will yield more specific information that could be pulled apart and analyzed any number of ways to Sunday, but when I started looking at all the luck stat info on Tos, it didn't take me long realize that all I cared about was one number.

"No wonder I lost. My luck stat was -12.4 over the last 200 moves" 

But I'm guessing all you stat wonks want meatier numbers you can sink your teeth into.. I'll be fine with that too.  My only request is to have the system include a few numbers that can answer those basic questions above in bold.

To make M57 stat similar to g-rating and meaningful in my mind use:

n = number of dice rolled
s = number of sides on dice

Attacker / defender luck = (sum of attacker dice - n - (.5)(n)(s-1)) / ((s-1))

This gives a symmetric distribution between -1.5 and 1.5 for the attacker and between -1 and 1 for the defender (assuming 6 sided dice). This also correct for higher/lower sided dice than the standard 6 sided and for the attacker have more dice than the defender.

Not really the stat I want, but I think this is the correct measure for what M57 is proposing.

*edit: again posted without reading M57 post above.

Alpha, Thanks for looking at my crude idea with fresh eyes and expanding on it with a clearer and more workable equation.

Really, I'm sure I'll be thrilled with whatever Tom goes with, and I'll actually be surprised if my very simple idea gains traction much less gets implemented.

For the meaningless part:

The issue that I see is that you are trying to normalize two distributions that are different so that they are the same and thus destroying information.
I may be wrong on this.

Again, I think an attacker who rolls 3 perfect dice (highest possible rolls) is more lucky than a defender who rolls 2 perfect dice.

Where I may be wrong in my reasoning is that the outcome only depends on the top two dice for both players or top one if defender only has one and thus the luck stat may be more useful to just take into account the dice that count.

Here I think there are issues as well and thus I propose my above formulas which would tell you if you had low rolls or high rolls (not whether or not you were lucky/unlucky with how they matched the opponent, which is the stat I would like.) This could also be totaled across games to tell you if you have been high or low in rolling on the site.

The issue that I see is that you are trying to normalize two distributions that are different so that they are the same and thus destroying information. 
I may be wrong on this.

Again, I think an attacker who rolls 3 perfect dice (highest possible rolls) is more lucky than a defender who rolls 2 perfect dice.

Where I may be wrong in my reasoning is that the outcome only depends on the top two dice for both players or top one if defender only has one and thus the luck stat may be more useful to just take into account the dice that count.

No, I think you are right..  Consider that over time, as more dice are thrown by the attacker, the resulting cumulative luck number will converge on 0 faster than the defender's luck number.  Even if you don't consider the defender's rolls, this would be a flaw with the system, right? But there must a way to adjust for it.

Back from vacation and will be working on this this week.

In my head, the term 'luck' in regards to this game refers to whether or not you lost more or less units than expected for a given battle/series of battles/game/games.

There are additional stats you can develop with regards to 'pure' dice related 'luck', in the sense that wow, you've rolled 4% more 6s than you should have over the course of your WG career isn't that lucky, but in my opnion it's not really what counts.

In the under/over version of luck, the actual numbers on the dice are meaningless, a 6,6,6 vs 6,6 is identically to a 1,1,1 vs 1,1. All that's needed is to tally attacker side and defender side counts:

3-2
3-1
2-2
2-1
1-2
1-1

... add the total number of battles (for whole-game stats), tally total wins and losses, and compare to the expected win/losses (for standard dice sides).

You can also calculate expected wins/losses for arbitrary dice sides, and when you throw the whole thing into a blender you come out with a single number, like you said:

"No wonder I lost. My luck stat was -12.4 over the last 200 moves"

I agree with not being able to quantify context. You can have a luck stat of +50, but if you get unlucky and fail by a single unit to eliminate the player with 5 cards towards the end-game, you were horribly unlucky.

Conversely, I've been in situations where horrible horrible luck on one turn has translated into an ultimate winning strategic position a turn or two later.

Really, the only point of the luck stats is to complain about someone else's dice or your own dice. It doesn't really help play at all since the trend is just imaginary really. Wargear doesn't care who you are, it just does the random numbers each time. Yes, I'm sure that if you tracked enough you would find a pattern as it is based off of an equation and not really random but why bother with all that.