Luck?

[hypercross]

So hello, this is another complaint about luck from a new player. If people are mad at these routine rants, well, sorry about that.

I just had a game where I had bad luck. Sent out two footpads to crucial forest tiles, got them killed in one turn, by two units, respectively. They should be dodging at 70%, but they got shot in a spree. On turn 10 I checked my statistics and it reads Damage Inflicted -16%, Damage Taken +92%.

Well, I know, that's not typical. Typically I don't get that unlucky. Typically I get +10%, or -15%, or +16%, it happens all the time. People will say, I am just being extremely unlucky today, it's not my fault or the game's fault, it's just possibility.

Unfortunately, I am not enjoying this Damage Taken +92% in the first 10 turns. Why ? Because after the first 10 turns probability is still even. It will go on to be "fair" after turn 10. The most likely scenario is, I will expect to do close to +0% damage and get close to +0% damage after turn 10, and unless I am really strategically successful (which is unlikely because I've already lost advantageous tiles), or I get lucky enough, I will never catch up my opponent.

Now you will tell me, I should go play 100 more games, after which I will find I am lucky in 50 of them.
But, nah, I don't have that time, you see. I played one game, it didn't work out, it's already my day. I will most likely drop Wesnoth to get myself happier, because it's just a game and I am no professional player.

So, This is the problem I am proposing here:
If a player is unusually unlucky in some part of the game, is amendment to the probability system unnecessary because its his duty to deal with his temporary unluckiness?
If a player had 1 unusually unlucky game out of 1000, is amendment to the probability system unnecessary because 0.1% is statistically ignorable ?
If there is 1 player out of 1000 played with mostly bad luck, is amendment to the probability system unnecessary because 0.1% is statistically ignorable ?

Just think about how probability works in Wesnoth. It's a string of independently random integers from the range 0-99. As a result we get a binomial distribution. As a result you can expect every situation to EXIST, just in smaller numbers.

As a result, there's almost certainly a player who played with sheer unluckiness.
Is he statistically ignorable ?

Think about slavery. Is it a fair thing, if you have the same chance of being born as a slave, and being born as a slave master ?
Obviously not, because no one should be a slave, not even if that's a result of a completely balanced and independent random variable.
Now, should a player be treated unfairly, even if the unfairness is a result of a completely balanced and independent random variable?

I think there's some geeky ideology in here. Sheer independent randomness is elegant, simple, and beautiful. For a game it's a very adorable and exciting feature. But, uh, unfortunately I am a player, not a developer, it's comparatively harder for me to appreciate that prettiness. I am only concerned whether I am treated fairly or not.

In short, RARE cases in Wesnoth is getting me really annoyed. For my 9-3 70% attack I don't want to see it miss 3 times in a row, nor do I want to see it happen on my opponent's attack. There should be a lower bound, whatever bound will work, because the current lower bound is ZERO, or below zero considering retaliation. No I don't like the no-randomness mod either, because just as the Luck Rationale paper says, luck is a play factor in Wesnoth. I am completely with randomness in Wesnoth, but it shouldn't be sheer binomial distribution, that's the only thing I am against.

I actually made some edits to the source code to achieve a kind of compromised randomness, and end up very happy with it. It gives me around +- 1% damage deviation every single game (where vanilla most likely gives around 10%).

It works like the following :

Multiply the chance of hit by times of attacks. For my favorite a 9-3 70% attack, it gives me 3x70% = 210%.
The integer part of the result is the number of time where an attack always hits. So 210% means I always hit 2 times.
The decimal part of the result is the probability where I can hit one more time. So 210% means in 10% of the cases I hit 3 times.
Now that we know how many times we will hit, we generate permutations of hits and misses, and distribute chances across them.
So for 210% we know in 90% cases we hit 2 out of 3, and there are 3 possible permutations. Each permutation will have a chance of 90%/3 = 30%.

Examples:

9-3 70% attack result distribution:

30% : hit, hit, miss.
30% : hit, miss, hit.
30% : miss, hit, hit.
10% : hit, hit hit.

It always hit 2 or 3 times.

12-2 50% attack distribution:

50% : hit, miss.
50% : miss, hit.

always hit once.

17-1 40% attack distribution:

40%: hit.
60%: miss.

exactly like vanilla.

Unfortunately the code is lost, and it's written a year ago. I was writing directly to the c++ source, I don't even have a c++ compiling environment now. If it's possible for someone to give me a hint on how to use the Lua modding engine it would be much appreciated.

[Chris NS]

If we're using the slavery analogy, the obvious response is that slavery is unfair, but that was the reality of the situation. It would be ludicrous to demand all references to slavery get removed in a computer game about the American Civil War because we disapprove of slavery.

Similarly, you may think it's unfair if a weaker force defeats a stronger force in a battle, but that's the way it is. Things don't always go to plan. The question is how you respond to it. A Wesnoth player whose whole strategy collapses on the first bit of bad luck isn't a good player. A player who is prepared for unexpected turns of bad or good luck and adapts strategy accordingly is much more likely to win.

[hypercross]

It would be ludicrous to demand all references to slavery get removed in a computer game about the American Civil War because we disapprove of slavery.

Not exactly the point I made. I meant the game is not fair, not the game means to describe unfairness. Or does it?

I am not standing on the "Become Better Players" team. I play for fun, I don't really care if I am good or bad, I just want to enjoy the thinking. Yes a good player handles bad luck, but he may not enjoy it.

[8680]

Have you seen threads 35757 and 38987? Perhaps you could adapt their work.

[alluton]

I tried to search for thread by Dave about the philosophy behind luck. Perhaps someone else can find it.

[lipk]

I tried to search for thread by Dave about the philosophy behind luck. Perhaps someone else can find it.

Try looking in the "Important Links" sticky

[Captain_Wrathbow]

Well it would be nice if someone just actually provided the link.

@hypercross: also see this and this - not exactly what you're particularly asking for, but related; you might find it helpful.

[Dave]

Bad play or bad luck? Why did you lose? Many people seem to think this is an easy thing to distinguish, but in fact it's insidiously hard. Distinguishing the two is the most important skill in becoming a good Wesnoth player.

Removing luck or 'tweaking' it would most likely remove this tantalizing aspect of Wesnoth.

That said, Wesnoth is Open Source and easily moddable. Make your own mod that tweaks the way it works, try it out, share it with others, put it on our content server! Maybe you'll make something more fun that more people like and we'll end up moving over to your creation.

David

[Crow_T]

Where in the source code is the battle code?

[AI]

src/actions/attack.cpp

[Crow_T]

Hmmm, I think I'm way too stupid to do anything with that...

[hypercross]

src/actions/attack.cpp

yeh that's where I was editing before, but now I don't have a cpp environment ( I'm on a new windows pc in China with super slow network, downloading everything from scratch is really painful ). I saw Dugi's no randomness mod but that's too much - randomness is not bad, it's just too wild.

Well it would be nice if someone just actually provided the link.

@hypercross: also see this and this - not exactly what you're particularly asking for, but related; you might find it helpful.

Thanks, I've read that rationale like a year ago, still I am feeling more like "geeky ideology" in there. Well, probably for the designers every challenge in the game could be enjoyable. Not for me though.

Wish there's some documentation on Dugi's code.

Bad play or bad luck? Why did you lose?

Thanks Dave.
Well my statistics read Damage Taken +92%. I don't feel particularly bad in that game, I managed to get back later and statistics went back to about +20%. I am just particularly unhappy that I had an unwanted bad start. Just because I can win under higher difficulty, doesn't mean I always want to and expect to.

I am not really expecting to convince anyone, it's already a commonly accepted design thing. Best outcome of this discussion is I make a mod and let the mod speak.

and I need to do some research on that.

[Sapient]

Here's another one of the more thoughtful randomness in MP discussions: On Randomness and Game Longevity. There have been so many of these threads now I know it's easy to lose track of the sheer number of them. But I thought that one was one of the few actually worth reading.

[hypercross]

Here's another one of the more thoughtful randomness in MP discussions: On Randomness and Game Longevity. There have been so many of these threads now I know it's easy to lose track of the sheer number of them. But I thought that one was one of the few actually worth reading.

That parry and accuracy rating system is overly complicated imo. Adding two values to every single unit is crazy. Those terrain defense values are hard to memorize already.

Only complaint I have is not the "lack of control", but "lack of predictability". Dave is saying predictability equals control. Well, if it were true then Go must be a really easy and boring game for him. I wish I could be that clever.

For most of the games out there if a player plays twice as good as his opponent, you know he almost always wins.
In Wesnoth, however, playing twice as good means he wins twice as many games as his opponent. Played 300 games, lost 100 of them probably.

And that's a feature of a gambling game. In Texas Hold'em if you are good you can win a constant salary from rookies, but that doesn't make the game fair.

[The Black Sword]

For most of the games out there if a player plays twice as good as his opponent, you know he almost always wins. n Wesnoth, however, playing twice as good means he wins twice as many games as his opponent. Played 300 games, lost 100 of them probably.

It's hard to quantify exactly what you mean by 'plays twice as good', winning twice as many games seems as a good a definition for that as any tbh. However I suspect you're over-estimating the luck factor in wesnoth in these statements.

The win percentage of the top ladder players is around 80%:
http://wesnoth.gamingladder.info/ladder.php.

If you subtracted out all games against players with rating above 1800, it'd probably be a lot higher.