A Thought On The Stats Tab
Moderator: Forum Moderators
A Thought On The Stats Tab
A thought occurred to me after two games (one which I won, one which I didn't) where there was some lop-sided luck in both instances that a slight expansion of the damage ratio part of the statistics tab could be useful. Currently, the tab records the daily ratio of total damage to expected damage, and compiles that into a global tab. This is good, but it doesn't tell the whole story.
What I'm thinking of is a damage ratio of chance-to-kill section of the same tab. I'll need to work out the details of how this could work (and make no mistake - I'll need help on this one), but the idea would be to create a percentile that relates specifically to the chances of killing a unit when that chance actually exists. It addresses a specific problem, which could be illustrated as follows.
Whereas the damage percentile recorder marks the total damage taken/inflicted, this has no relation to whether a unit is actually killed. E.G. three units can attack my Grunt in a village, which retaliates and hits all of them hard and perfectly, only for all of them to retreat and heal later. My damage percentile will be recorded as very high and may even balance out later, but it doesn't offer a reflection on whether or not that damage was effective.
Which isn't to say that all damage should necessarily lead to a kill - but it does mean that if I'm attacking a unit with 1 hit point, and miss ten times in a row, this communicates considerably more information than missing ten times in a turn against ten different units (battles can certainly rage everywhere with gradual effects that are lucky or unlucky, but not killing a unit that should have died on account of probability or killing one that shouldn't have results in an immediate re-distribution of one's forces, which is significant). The result would be, roughly, an indicator that demonstrates how far off a player is in terms of killing units that had a certain % chance of being killed.
Basically, what I'm imagining is that if a unit has no chance whatsoever of being killed in a battle, it simply doesn't go on record. However, if there is a CTK, then the data is recorded in roughly the same way as in the other case - if I attack a 20% CTK and kill it, then I'm up by quite a lot; but if I fail and the chances increase to 60% and I succeed on the second attempt, this will balance out the statistics. I sense that it would need to work slightly differently than the standard damage statistics, because there's two stages of calculation involved - I can grasp the general concept, but am not sure how to string it out in practice. It would also be interesting to apply it specifically to situations where a unit only has one hit left before it dies (I'm sure you get my drift when I suggest this). Anyways, that's my idea.
Thoughts?
What I'm thinking of is a damage ratio of chance-to-kill section of the same tab. I'll need to work out the details of how this could work (and make no mistake - I'll need help on this one), but the idea would be to create a percentile that relates specifically to the chances of killing a unit when that chance actually exists. It addresses a specific problem, which could be illustrated as follows.
Whereas the damage percentile recorder marks the total damage taken/inflicted, this has no relation to whether a unit is actually killed. E.G. three units can attack my Grunt in a village, which retaliates and hits all of them hard and perfectly, only for all of them to retreat and heal later. My damage percentile will be recorded as very high and may even balance out later, but it doesn't offer a reflection on whether or not that damage was effective.
Which isn't to say that all damage should necessarily lead to a kill - but it does mean that if I'm attacking a unit with 1 hit point, and miss ten times in a row, this communicates considerably more information than missing ten times in a turn against ten different units (battles can certainly rage everywhere with gradual effects that are lucky or unlucky, but not killing a unit that should have died on account of probability or killing one that shouldn't have results in an immediate re-distribution of one's forces, which is significant). The result would be, roughly, an indicator that demonstrates how far off a player is in terms of killing units that had a certain % chance of being killed.
Basically, what I'm imagining is that if a unit has no chance whatsoever of being killed in a battle, it simply doesn't go on record. However, if there is a CTK, then the data is recorded in roughly the same way as in the other case - if I attack a 20% CTK and kill it, then I'm up by quite a lot; but if I fail and the chances increase to 60% and I succeed on the second attempt, this will balance out the statistics. I sense that it would need to work slightly differently than the standard damage statistics, because there's two stages of calculation involved - I can grasp the general concept, but am not sure how to string it out in practice. It would also be interesting to apply it specifically to situations where a unit only has one hit left before it dies (I'm sure you get my drift when I suggest this). Anyways, that's my idea.
Thoughts?
-
skeptical_troll
- Posts: 500
- Joined: August 31st, 2015, 11:06 pm
Re: A Thought On The Stats Tab
Although I agree that the overall damage statistic is somewhat limited (I basically use it just to know how much I'm entitled to feel victim of a universal conspiracy 
) I'm not sure if what you suggest could be very meaningful. I suppose you want to convey with a number the 'luck in crucial moments', but why is it just the killing blow the important one? You always need many units to take an enemy down, so also the luck of those softening him matters a lot. Plus, injuring a units to few HP also forces it out of the battle and makes it useless for few turns, and kills done at the end of the battle, when the tide has turned on one side or the other, are quite irrelevant. Anyway, you can still try to implement it and see how it goes.
If you want a suggestion on how to do it: you can simply use the rate 'successful kills/expected kills' as it is done for the damage, but the expected kills are a bit more complicated to calculate. Say you had N fights so far were you could kill, then it is like tossing a coin N times and calculating the expectation value for the number of 'heads', except that the probability is different for every toss. I don't know how strong your math is, this sounds hard to compute analytically, if possible at all, but you can always Monte Carlo the hell out of it pretty quickly (basically randomly generate a lot of possible outcomes for those fights), I believe.
EDIT: on second thought, maybe the expected number of kills is simply the sum of all individual probabilities, only the full distribution is hard to compute..
) I'm not sure if what you suggest could be very meaningful. I suppose you want to convey with a number the 'luck in crucial moments', but why is it just the killing blow the important one? You always need many units to take an enemy down, so also the luck of those softening him matters a lot. Plus, injuring a units to few HP also forces it out of the battle and makes it useless for few turns, and kills done at the end of the battle, when the tide has turned on one side or the other, are quite irrelevant. Anyway, you can still try to implement it and see how it goes. If you want a suggestion on how to do it: you can simply use the rate 'successful kills/expected kills' as it is done for the damage, but the expected kills are a bit more complicated to calculate. Say you had N fights so far were you could kill, then it is like tossing a coin N times and calculating the expectation value for the number of 'heads', except that the probability is different for every toss. I don't know how strong your math is, this sounds hard to compute analytically, if possible at all, but you can always Monte Carlo the hell out of it pretty quickly (basically randomly generate a lot of possible outcomes for those fights), I believe.
EDIT: on second thought, maybe the expected number of kills is simply the sum of all individual probabilities, only the full distribution is hard to compute..
Re: A Thought On The Stats Tab
I don't see this being very useful, because there are too many things that affect how influential the damage/kill was. Besides the points skeptical_troll mentioned, another important factor is the value of the unit which was killed - which depends not only on the unit type, but how useful the unit is in a particular matchup or map, its position (e.g. whether it is holding a critical point of the formation), how close it is to leveling, time of day....
Another approach might be to only take the stats during the crucial turns (turns where a lot of combat takes place), since most Wesnoth battles are mainly decided by the outcome of one or two large battles which only last a few turns.
Another approach might be to only take the stats during the crucial turns (turns where a lot of combat takes place), since most Wesnoth battles are mainly decided by the outcome of one or two large battles which only last a few turns.
Screenshot playthroughs: Let's Play Dead Water, Let's Play Invasion from the Unknown and Let's Play After the Storm