Spring RTS Engine

Split from BA 7.87 thread http://springrts.com/phpbb/viewtopic.php?f=44&t=31283.

klapmongool wrote:The awards don't reflect what happened in the game. The player(s) deciding the game often don't get an award while players who barely played do. I think that makes em broken. I know it is hard to measure properly though..

Yes, its hard to measure proeprly. As far as I know all the awards *do* always get awarded to the player who deserved to win that particular award (possibly the 'efficiency' award is a bit unreliable, but inherently). There is no 'who decided the game' award!

Silentwings wrote:

The awards don't reflect what happened in the game. The player(s) deciding the game often don't get an award while players who barely played do. I think that makes em broken. I know it is hard to measure properly though..

Yes, its hard to measure proeprly. As far as I know all the awards *do* always get awarded to the player who deserved to win that particular award (possibly the 'efficiency' award is a bit unreliable, but inherently). There is no 'who decided the game' award!

That is my point exactly: the awards don't show which teammembers made the win. I might think about better rewards this week. Or not :)

klapmongool wrote:That is my point exactly: the awards don't show which teammembers made the win. I might think about better rewards this week. Or not :)

This has never been clear in ANY GAME. Even something simple like a MOBA.

smoth wrote:

klapmongool wrote:That is my point exactly: the awards don't show which teammembers made the win. I might think about better rewards this week. Or not :)

This has never been clear in ANY GAME. Even something simple like a MOBA.

So no awards at all?

Actually it is pretty clear for an experienced player which players did best in a BA game. It is just pretty hard to pick the right set of variables to reflect this.

Think of "awards" as "achievements", and you'll fast realize they don't always denote a good player.

klapmongool wrote:So no awards at all?

just because you don't feel honored by them doesn't mean they should be taken out.

klapmongool wrote:Actually it is pretty clear for an experienced player which players did best in a BA game. It is just pretty hard to pick the right set of variables to reflect this.

Then quantify it, give details or stop complaining.

smoth wrote:

klapmongool wrote:So no awards at all?

just because you don't feel honored by them doesn't mean they should be taken out.

The problem is that people feel honored wrongly. Stimulating strategies/tactics that are actually detrimental.

klapmongool wrote:Actually it is pretty clear for an experienced player which players did best in a BA game. It is just pretty hard to pick the right set of variables to reflect this.

Then quantify it, give details or stop complaining.

Wow. You are missing the point. I am not going to explain this to you considering our previous encounters (After which I thought we agreed not to bother each other again).

klapmongool wrote:Wow. You are missing the point. I am not going to explain this to you considering our previous encounters (After which I thought we agreed not to bother each other again).

Just saying. rather that say it should be taken out (because it doesn't really effect gameplay) you should be suggesting how to improve it.

gajop wrote:Think of "awards" as "achievements" ... they don't always denote a good player.

That is how I've always thought of it too. I think there's very little hope for algorithmically deciding who played 'best' and I won't try to code it myself, but I don't want that to stop anyone else from trying!

I got strange errors:
[f=0077789] Spectator Dora_the_Explorer left the game: normal quit
[f=0077790] Error: LuaRules::RunCallIn: error = 2, GameFrame, [string "LuaRules/Gadgets/unit_prevent_sliding_hax.l..."]:34: attempt to perform arithmetic on local 'x' (a nil value)
stack traceback:
[string "LuaRules/Gadgets/unit_prevent_sliding_hax.l..."]:34: in function 'GameFrame'
[string "LuaRules/gadgets.lua"]:949: in function <[string "LuaRules/gadgets.lua"]:947>
(tail call): ?
[f=0077805] Error: LuaRules::RunCallIn: error = 2, GameFrame, [string "LuaRules/Gadgets/unit_prevent_sliding_hax.l..."]:34: attempt to perform arithmetic on local 'x' (a nil value)
stack traceback:
[string "LuaRules/Gadgets/unit_prevent_sliding_hax.l..."]:34: in function 'GameFrame'
[string "LuaRules/gadgets.lua"]:949: in function <[string "LuaRules/gadgets.lua"]:947>
(tail call): ?


full infolog is here: http://pastebin.com/5CFPXsZR

Fixed in svn, thanks.

Note: you've posted in the wrong thread

It might be me but I usually also fail to see how the awards at the end are related to the game I just played. Without analyzing the code for me they appear to be bugged. But maybe I just dont understand them. Either way I would vote to remove them, sorry.

PS.
I dont like the graphics on the awards screen. It looks imo too retro (for some reason it reminds me at Amiga games). Again sorry...

which awards don't make sense?

Can you link to something you like better as far as visuals?

The awards are:

1) Eco Killing
Given to the player who kills the most enemy eco-production (each kill scores mcost*60 + ecost of unit killed)

2) Fight Killing
Given to the player who kills the most enemy fighting stuff (scored same as above)

3) Efficiency Award
Given to the player with the highest damage dealt / eco used ratio (with some extra conditions for having done > x damage etc)

4) Taking Damage
Given to the player who took the most damage.

5) Eco Award
Given to the player who produced the most eco (e + m*60)

6) Sleep Award
Given to the player who went the longest time without doing any damage.

I don't see how there is any ambiguity there, although (3) is a bit simplistic imo. I don't know of any bugs in code and its always looked to me as though the awards are given correctly.

So far this thread has contained no suggestions for improvement and not even a clear description of the 'problem', so atm nothing is getting done here.

I would add the following "being useful" award (even some might argue it is not)

Count your ranking in the damage dealt you did in your team, and the rank only.

something like this:

score = integration( 1/rank_in_damage_dealth(t) *dt, t= 0 , t= end_of_game )

And normalize it with end_of_game

rank_in_damage_dealth(t) in [1, number of player in the team] , and this whatever players are dead or not. That has the advantage players who made huge damage but die still count.

I will tweak the end_of_game value so that it can be stop much before end of game (for example clip to the time the total damage that has been dealt is 99% of total), to avoid pips purposely waiting to increase stats.

99 is an example and up to you.

Silentwings wrote: So far this thread has contained no suggestions for improvement and not even a clear description of the 'problem', so atm nothing is getting done here.

Okay, sorry, maybe I was unclear (English isnt my first language). I will try to put straight what I think:

Problem:
Awards are given to the wrong player (at least sometimes).

Improvement:
Give it to the correct player.

I just had a game on Charlie. According to the end game graphs I had like 5 times the damage of the 2nd most damage dealing player (my teammates shared me metal and build power). Either way, the award for "Destroying enemy units and defenses" was given to someone else.
It might be correct but this is a good example of an award that felt wrong. Not sure though.

This is the game:
http://replays.springrts.com/replay/18f ... 51cdf4421/

Of course its up to you if you want to look into it or call it bullshit. I am just reporting it being a nice player.

Thanks, there was indeed a bug and was easy to find thanks to the replay; now fixed in svn. It affected at most one player per game so was often not visible.

I re-ran the replay with the bug fixed, and the awards for that game should have been:

Nice, thanks!

albator wrote:I would add the following "being useful" award (even some might argue it is not)

Count your ranking in the damage dealt you did in your team, and the rank only.

something like this:

score = integration( 1/rank_in_damage_dealth(t) *dt, t= 0 , t= end_of_game )

And normalize it with end_of_game

rank_in_damage_dealth(t) in [1, number of player in the team] , and this whatever players are dead or not. That has the advantage players who made huge damage but die still count.

I will tweak the end_of_game value so that it can be stop much before end of game (for example clip to the time the total damage that has been dealt is 99% of total), to avoid pips purposely waiting to increase stats.

99 is an example and up to you.

Can we get this ?

If yes and but dont want to implement it, can you point out where it is done in the code (the place the other awards are coded) and I will do it.

No - I would like to keep the awards as simple-to-understand statistics, not guessed formulas to try and approximate some individual idea of 'played best'.

But you can easily write code to calculate your formula in a widget for yourself - once the game is over you can get all the statistics that go into the endgame graphs using Spring.GetTeamStatsHistory (see wiki, in LuaSyncedRead), already broken down into bitesize intervals ready for you to approximate the integral with. Of course if it became a popular thing that many people used I would re-consider including it.