|
Automatic Balance Tester
I recently started studying computer science - and about a week ago I had the idea of automating the balancing process for creating a new setup in Mafia. I had several (pretty complicated) ideas for how to do it, but then I found this page: http://hamumu.com/forum/showthread.php?p=263954 It was a very simple, yet effective way of evaluating balance automatically.
Around that time I PM’d Foolishness for advice and suggestions about building this balance tester. He had several valuable suggestions and also helped a ton with the actual balancing.
The balance tester itself is programmed with Java so Java Runtime Environment ( http://java.com/en/download/index.jsp ) is required to use it. As long as you have JRE installed, it should work fine on any computer. A web implementation might have been slightly more user-friendly, but I have zero experience in using JavaScript so I decided to just program it with Java. A web implementation is definitely a possibility at some point, though.
So how does the balance tester work? The first window asks you to select the amount of players as a starting point (8-24) and generates a somewhat balanced default setup based on that number. It also asks if the game starts at day (default setting) or at night (giving a slight bonus to Mafia and Serial Killers). Then it opens the ”main” window where you can select from 23 different roles and add new roles to the setup, remove roles from the setup, remove all roles from the setup or switch to day/night start. The town vs anti-town balance value (+X for town, -X for anti-town) updates automatically as you make any changes. The text fields for player amount, day/night start and victory odds (%) for each team also update as you make any changes.
The balance and functionality of the program seem OK at this point, but I have no doubt that there’s still room for improvement. That’s why I thought it would be best to gather some feedback and suggestions before making an ”official” release of the program. Once released, it’s supposed to function as a helpful guideline especially for newer hosts (and why not for more experienced hosts as well!). It can’t completely replace an experienced player’s evaluation, but at least in most cases it should work just fine (and of course very fast)! It’s quite unusable for themed setups, though.
Here’s the download link. Any kind of feedback, suggestions and constructive criticism is appreciated.
https://www.dropbox.com/s/zb7oxzsz177h4f7/Mafia_Balance_Tester.jar (EDIT: latest version)
(Currently there’s no way to adjust how much KP the Mafia have at their disposal. You can assume that Mafia KP increases logically - in a linear fashion - as the number of Mafia players increases. So 1-2 Mafia would have 1 KP, 3-4 would have 2 KP or something along those lines. It’s a bit tricky to balance Mafia KP so it should be just assumed that Mafia KP is ”fair” considering the size of the game.)
|
Neat.
A little while back, I made a post, which listed five different setups and asked people to vote on the most balanced and least balanced.
I thought it would be interesting to see what your program thought and got the following results: (I assumed the framer would be roughly equivalent to the godfather at -7)
(Town Win %/Mafia Win %)
Setup A 60/40 +5
Setup B 38/62 -6
Setup C 47/52 -2
Setup D 42/57 -4
Setup E 39/52/7 -8
The funny part is that setup B was voted as the most balanced and setup C was voted as the least balanced. The numbers your program comes up with seems to suggest the exact opposite. That doesn't necessarily make either conclusion correct considering so much of this is opinion based and there really isn't a huge sample size to go by.
Just curious, how did you calculate the third party win percentages? They seem to be the most off from what I'd expect, but there probably isn't a very clean way to estimate using the +/- system.
I wouldn't really know where to start with evaluating the actual numbers, but they seem fairly reasonable. The blues might need to be scaled down a bit in comparison to the vanilla. For example, you'd obviously want 10 vanilla townies over 2 cops. The miller at -2 might be a bit much.
I did notice a small bug where the roleblocker/godfather/rolecop seem to disappear from the role list when adding the equivalent blue role, rather than update their value.
Nice job!
Edit: Opps wrong account
~ Kita ~
|
My biggest issue with this is the metric used. I do not think a simple +/-X is sufficient to measure a role. Further, I don't agree with the specific values assigned. For example, look at Town RB vs Town JK - equal balance, but JK is more valuable as a townie.
Before I go on, let me say that my measures are completely arbitrary as well.
Generally, when I am looking at balance, I break down each super-mechanic in the game into each role's effectiveness, conditions, and consequences. Here is an example from a setup I am aching to run, but haven't found the proper time yet.
Original Message From iGrok:Analysis of the following data at the bottom Abilities * Conditions ` Effectiveness - Negatives ^ Oneshot Kills in the game Alpha*```` Alpha*``^ Beta*```` Beta**`` Charlie****``^ One**``^ Two***`````^ Three**``--^ Four****``----^ Five*****```-^ Total: 2+7^ ************************* 25 (+14 for ^)
```````````````````````````` 28
------- 7
Avg: 2.8* (4.3*^), 3.1`, 0.8-
Alpha 9* 14` 2^
Numeric 16* 14` 7- 5^ Defenses in the game Alpha`````^ Six`````^ Seven**```- Eight**``` Total: 2+2^ **** 4 (+4 for ^)
``````````````` 15
- 1
Avg: 1* (2*^), 3.75`, 0.25-
Alpha 0* 5` 1^
Numeric 4* 11` 1- 1^
Lynch manipulation Alpha*``- Beta*``- Charlie**`` One****`- Two***-- Three*``` Four*`` Five` Six**--- Seven*```` Total: 8 *************** 15
````````````````` 17
-------- 8
Avg: 1.9*, 2.1`, 1.0-
Alpha 4* 6` 2-
Numeric 12* 11` 6-
Information Charlie*````-- One***```` Two``` Three*``` Total: 4 ***** 5
`````````````` 14
------ 6
Avg: 1.3*, 3.5`, 0.5-
Alpha 1* 4` 2-
Numeric 4* 10`
So, what does this show? Lets look at the averages. Kills are the most common ability. However, they are also the most conditional abilities there are. If you include oneshot as +**, you're looking at a ridiculous 4.3 average conditional score on kills. Defenses, on the other hand, are the most effective abilities in the game, especially when matched against their conditions. This should balance out the high number of kills. Lynch-manipulation is one of the most balanced ability types, though to be fair this one is the most difficult to score. The key here is that it has the highest -/` ratio, meaning that these abilities are tradeoffs. Combined with the anti-vote mechanic, and lynches likely will nolynch often, revealing only a bit of information regarding estimated Combat Powers. And finally, Information abilities have (as they should) the highest `/* and `/- ratios
This message is one I sent to GMarshal a while back for a game with two teams: Alpha and Numeric. The setup is designed to be extremely violent, and has a lot of odd roles and mechanics, but analyzing the components of each role gives me a much better vision of the balance of the setup.
As far as the program goes, it looks like a good start! I don't understand why the Day/Night start option window pops up first, seems like that could just be taken care of in the main window. Also, if you're looking to develop this further, being able to save/load setups from the clipboard (assuming of course that they are properly formatted) would be nice!
|
On December 01 2013 02:30 Xatalos wrote:Automatic Balance TesterI recently started studying computer science - and about a week ago I had the idea of automating the balancing process for creating a new setup in Mafia. I had several (pretty complicated) ideas for how to do it, but then I found this page: http://hamumu.com/forum/showthread.php?p=263954 It was a very simple, yet effective way of evaluating balance automatically.
I not so recently studied computer science.
I also studied some other stuff related to the scoring algorithm.
While these are comments, on the program(actually just on the web page you started with)
The program does what it does.
If people use it to do what it doesn't, it wont help. If they use it for what it does, it will help. I know that sounds cryptic and nonsense, but what it means is that the scoring model is of some merit in assessing balance, but it will have its limitations. (Time and tweaking will improve that up to some limit.)
It will if the weight numbers are right at least get you in the ball park with balance.
It will have merit as a tool.
It, your program, may have limitations.
NOT having yet looked at the app but just the page linked.
+ Show Spoiler +
I expect there will be corner cases and interaction effects.
One that comes to mind is doc that can repeatedly protect and cops generate "follow the cop"
Whereas as repeat docs without cops just/merely make the best analyst unkillable, which may be worse or better for scum depending on the particular analyst.
Whether or not the existence of both (repeat docs + cops) in the game breaks the game and how often it does, in part depends whether the roles are open or closed set-up.
Irrespective its a good start and a good idea.
Even if nothing else a tool that inexperienced players can get feel for balance with is useful. I know I spent some time playing with what the various C9 and etc setups meant for balance.
When my head vs standard wisdom, disagreed with what setups were balanced i knew I needed to think some more.
|
Thanks for the feedback! Actually I forgot to add Framer so that's probably in the next version.
That's an interesting comparison. I looked at the poll results and the following comments, and it seemed like several people had different opinions. On average, though, it seemed like B and D were considered the most balanced and A, C and E the least balanced setups.
Setups A and E are similarly unbalanced according to both sources so that's probably the case. Setup D is a bit more appreciated among the polls than in the program, but it's not that big of a difference. On the other hand, funnily enough, the winner of the "least balanced" poll (setup C) is actually the MOST balanced setup according to this calculation. And the winner of the "most balanced" poll (setup B) is actually the LEAST balanced setup......
I guess I can try explaining why there might be this big of a difference. I and Foolishness compared different all VT / all Goon setups as the "basis" of this balance tester. According to those comparisons, one Goon is roughly worth 4-5 VT's. We ended up with 5 for now. In that sense, it seems pretty logical that 9 VT's vs 3 Goons is pretty unbalanced. Although if we ended up with -4 as the value for a Goon, setup B would be pretty close to balanced.
It's a tough call. I'd certainly like more feedback on if Goon should be -4 or -5. I see you thought that setup B was fairly unbalanced in favor of Mafia, right, Kita? So do you think the current value is OK or should it be adjusted?
It's interesting that setup C was generally thought to be so unbalanced while the program considered it as the most balanced setup. I'd definitely want to see some more opinions on this! The Cop+Doctor combo is certainly powerful, but then again, this goes back to setup B. If we consider that 9 VT's vs 3 Goons is pretty unbalanced in favor of Mafia, then 6 VT's & 3 blues vs 3 Mafia (including power roles) should theoretically be pretty close to balanced. But if we consider that 9 VT's vs 3 Goons is pretty close to balanced, then 6 VT's 3 blues vs 3 Mafia would be pretty town-favored.
This all goes back to if Goon should be -4 or -5 again. If it's -4, then B would be pretty balanced and C pretty unbalanced in favor of town (just as the polls suggested). If it's -5, then B is pretty unbalanced in favor of Mafia and C is pretty close to balanced (which is against the poll results, but I'm not completely sure it's incorrect even so). More opinions please! 
The balance of the neutral roles is certainly the most untrustworthy right now. I don't think we even went through ANY setups including neutral roles with Foolishness. The values are all pretty much untested and came from "thin air", so to say. On the other hand, neutral roles are relatively rare in TL Mafia, so it's harder to have an educated opinion on them. It'd be very helpful if someone had a solid opinion on how to balance neutral roles properly.
Assassin win percentage is calculated as simply 1 / (the amount of Assassins) - since one Assassin always wins the game, no more, no less. Survivor is always 25%, which is pretty much conjured from thin air (not sure how to calculate it well?). Anti-town Survivor is (Mafia's victory odds / 3) / (the victory odds of town + Mafia + SK) since he only wins with Mafia but is less likely to win than Mafia. Serial Killer is simply (SK balance score / the amount of SK's) / (the balance scores of town + Mafia + all SK's).
Actually the blues are already scaled down compared to the original source ( http://hamumu.com/forum/showthread.php?p=263954 ). It's possible that they need to be tuned down even more, but I'm not sure about that yet. Should Cop maybe be +4 for example?
Ah yes, adding roles is working a bit weirdly in some scenarios. Currently if there are conditional values, and the condition for that role changes, it just removes the role from your choices. I can try making it to change the value instead of removing the whole role.
That's a good start but more feedback is definitely needed!
|
Oh, more posts were posted as I was typing... That was in response to Kita. I have to go now but I'll be answering ASAP.
|
AxleGreaser: oh, a fellow (former?) computer science student! I look forward to your opinion on the program.
iGrok: that's some pretty advanced methodology for balancing! I'll have to think of maybe remaking the whole balancing process with your method, although I think simplicity also has its benefits. Your method definitely takes role interactions better into account than simple +/- balancing.
Yeah, I guess it's unnecessary to ask for day/night start twice. I'll also have to consider adding some setup-saving support.
|
On December 01 2013 21:39 Xatalos wrote:AxleGreaser: oh, a fellow (former?) computer science student! I look forward to your opinion on the program. iGrok: that's some pretty advanced methodology for balancing! I'll have to think of maybe remaking the whole balancing process with your method, although I think simplicity also has its benefits. Your method definitely takes role interactions better into account than simple +/- balancing. Yeah, I guess it's unnecessary to ask for day/night start twice. I'll also have to consider adding some setup-saving support.
Setup-Saving doesn't even have to be done in-program, being able to load from the clipboard would be fine like Cockatrice does, we could have a library in this thread.
|
So iGrok, do you mean like that there would be a button along the lines of "Copy to Clipboard" that would just copy the current chosen roles list to your clipboard? That seems relatively simple to do.
I also looked at your method (not very in-depth yet) and I think it's well worth considering to break the roles into components and evaluate these components rather than specific roles. It would be easier to balance the different interactions and combos if you just consider a few components rather than 20+ roles. Currently there's some conditional balancing (for example Miller's negative effect only affects balance if there's a Cop - actually it should also trigger from Parity Cop I guess! - and Mafia Roleblocker only gains extra value from Goon if town has a blue role - but nothing to the extent of your method).
On the other hand, it gets a bit more complicated than the simple +/- system. I'll have to test this method as well to see how efficiently it can be automated. In the end, if it's more accurate (which I think it probably could be), it should indeed be used instead.
It's a bigger rework of the balancing system, though, so I think I'm first focusing on improving the current system and the overall usability of the program. But it seems likely that a component/interaction-based method would work better than the current method if there are several blue roles.
|
Ah ok, I see you meant copy FROM clipboard. Although both could be useful, I guess.
|
Yeah I meant to/from.
I think what you have is a good basic balance program. More refined methods can be subbed in later once you have the basic framework down.
|
Okay, I just implemented quite a bit of new stuff. Only iGrok's suggestion for a new method to balance the roles has yet to be implemented, but I'll probably be doing some testing on that method in the near future.
Here are the complete "patch notes":
- Changed Cop’s value to +4
(reasoning: as Kita mentioned, +5 does seem slightly excessive, but this is by no means final)
- Changed the basic value of all Mafia to -4 and adjusted Mafia power roles accordingly
(reasoning: makes Kita's poll results match quite closely with the balancing results of this program - and overall seems slightly more appropriate, but again, this is by no means final)
- Removed the unnecessary day/night start choice from the starting window
(reasoning: not much to say about this)
- Added a new role: Mafia Framer (-4/-6) (depending on if town has a Cop / Parity Cop or not)
(reasoning: a pretty common role that should be in this balance tester)
- Added a new feature: import & export a setup from clipboard
(reasoning: as iGrok suggested, this makes it possible to easily save tested setups and to test setups that have been saved for future reference)
- Fixed the bug that Kita mentioned in his feedback
(reasoning: the less bugs the better)
Here is a download link to the new version of the balance tester:
https://www.dropbox.com/s/zb7oxzsz177h4f7/Mafia_Balance_Tester.jar
Thanks for the very helpful feedback so far!
|
iGrok:
1) Is the current clipboard functionality working as you intended?
2) Any other good features that should be added?
3) Overall, how well do you think the current balance system performs? Are there balancing results or role values that you'd disagree with?
4) I see you disagreed with RB and JK having the same value. I agree that JK's power is more versatile, but doesn't it also have the potential drawbacks of saving a Mafia or roleblocking a blue role? Of course RB can also roleblock a blue role, but it seems more likely in JK's case (since JK would be more likely targeting a pro-town player). Maybe JK should still be higher though. RB +2, JK +3?
5) Could you maybe explain your balancing method a bit more? I'm not sure if I completely understand it yet.
Kita:
1) How do you like the current balance values (for example in comparison to your poll results)?
2) Especially, do you think Mafia's basic value should be -4 or -5?
3) What specifically was wrong with the neutral roles? How should they be changed?
|
|
|
Note i am pretty sure i am not helping with this, so you may be wise to ignore it.
At the very least mainly ignore it for now.
On December 01 2013 20:55 Xatalos wrote:
It's a tough call. I'd certainly like more feedback on if Goon should be -4 or -5. I see you thought that setup B was fairly unbalanced in favor of Mafia, right, Kita? So do you think the current value is OK or should it be adjusted?
Id think (know) folishness and Igrok know about balance.
and the below is motherhood stuff
Thus the existence of the above question is a confusion for me as I just see answers not questions.
perhaps its because I have the other skillz.
TBMK. (GIVEN 1 NK per night not rounded(SCUM/2) )
A basic All vanilla setup that closeish to balanced.
is
9 players (7 Town 2 scum)
13 players (10 town 3 scum)
Ref POINT 1 (see below)
The reason is, you get one mislynch for each lynch.
At this point or shortly after that TBMK breaks.
17 players (13 town 4 scum)
is I believe thought to be hard on scum at 1 kp per night.
My mathematical modelling spidy senses tell me that is for a few reasons.
Firstly elapsed days matter: Which ever scum is trying to survive to the last day has too much trouble keeping all their ducks(lies) in a row for that long. Hence 1 for 1 mislynches and lynches does not go on forever.
Secondly, as each scum flips there are more and more scum scum interactions that can look wonky.
Apart from knowing what is valid by real world experience and simply modelling/fitting that with a function I am not sure there is much chance of guessing theoretically what the effects of Firstly and Secondly are.
AT Ref POINT 1 a plauible mathematical algorithm is to decide each goon is worth -4 and the FIRST VT is worth zero as you must survive with 1. ( ) and each subsequent VT is worth 1.
(or each VT is worth 1 but game set-ups start set at -1 scum favoured)
Note How valuable a doctor is in terms of VTs is not likely at all to be a fixed number. (for larger games)
TBMK 12 player setups with Doc are kinda balanced.
BUT adding a doc to an 8 player setup will make a different amount of difference. (I think less)
The reason is the number of days on which the Doc may save a player is lower in an 8 game than a 12.
However the odds that a doc will make it to 3 man LYLO and claim (maybe eliminating the patsy...) is lower.
While your question asks for an integer value for a goon, decimal fractions are not fatal.
Neither is it implausible to just use a Lookup Table that specifies for Each sized Town How many goons it should have.
Eg Town size 12 is 3 goons but is +1 goon favoured.
Eg Town size 13 is 3 goons but is +0 goon favoured.
Eg Town size 14 is 3 goons but is -1 goon favoured.
... but it might not quite follow that pattern later as the above discussion on larger games indicates.
There are yet other even finer distinctions.
Eg Town size 13 is 3 goons but is +0 goon favoured.
This game if it goes according to PAR and ends at a 3 man lylo, at one time it had
(2 scum nks for each town mislynch & scum lynch.) x 2
thus it gets to 5 townies 1 scum.
At that point there have been 4 scum controlled towny deaths and 2 town controlled ones.
Scum has twice as much control over whether skilled and keen to play townies wind up making the next decision.
Eg Town size 12 (1 doc) is 3 goons but is probably also about +0 goon favoured.
This game if it goes according to PAR and ends at a 3 man lylo, at one time it had
(2 scum nks for each town mislynch & scum lynch.) x 2 (-1 town nk if the doc saves 1 and is thus on par)
gets to 5 townies 1 scum.
At that point there have been 3 scum controlled towny deaths and 2 town controlled ones.
Scum has less than twice as much control over whether skilled and keen to play townies wind up making the next decision.
|
On December 02 2013 20:18 iGrok wrote:Hmmm....
Heh Well, the point is to generate a default setup based on the amount of players. It's intended as kind of a starting point. If you wish to start from scratch, you can easily just click "Remove All Roles". I tried making it bigger, but I think it looks best as a small box. The JComboBox (where you select the amount of players) looks pretty stupid if it spreads to a larger area.
Okay, I'll demote RB in the next version.
That's a good idea to consider; showing the roles as "X x Role" makes the interface a bit cleaner. That might even be in the next version - we'll see how it works out.
I'll have to think about that editing option. It seems pretty hard to implement with the way my interface functions currently, so it'd require some major reworking.
Okay, good that the clipboard works. I wasn't exactly sure how Java handles clipboards of different operating systems.
By the way, could you explain your balancing system with a bit more detail? Maybe with examples from the roles I have in the program? I wouldn't want to implement the system and then realize that it's completely wrong 
|
On December 02 2013 21:03 AxleGreaser wrote:Note i am pretty sure i am not helping with this, so you may be wise to ignore it. At the very least mainly ignore it for now. Show nested quote +On December 01 2013 20:55 Xatalos wrote:
It's a tough call. I'd certainly like more feedback on if Goon should be -4 or -5. I see you thought that setup B was fairly unbalanced in favor of Mafia, right, Kita? So do you think the current value is OK or should it be adjusted?
Id think (know) folishness and Igrok know about balance. and the below is motherhood stuff Thus the existence of the above question is a confusion for me as I just see answers not questions. perhaps its because I have the other skillz. TBMK. (GIVEN 1 NK per night not rounded(SCUM/2) ) A basic All vanilla setup that closeish to balanced. is 9 players (7 Town 2 scum) 13 players (10 town 3 scum) Ref POINT 1 (see below) The reason is, you get one mislynch for each lynch. At this point or shortly after that TBMK breaks. 17 players (13 town 4 scum) is I believe thought to be hard on scum at 1 kp per night. My mathematical modelling spidy senses tell me that is for a few reasons. Firstly elapsed days matter: Which ever scum is trying to survive to the last day has too much trouble keeping all their ducks(lies) in a row for that long. Hence 1 for 1 mislynches and lynches does not go on forever. Secondly, as each scum flips there are more and more scum scum interactions that can look wonky. Apart from knowing what is valid by real world experience and simply modelling/fitting that with a function I am not sure there is much chance of guessing theoretically what the effects of Firstly and Secondly are.
AT Ref POINT 1 a plauible mathematical algorithm is to decide each goon is worth -4 and the FIRST VT is worth zero as you must survive with 1. ( ) and each subsequent VT is worth 1.(or each VT is worth 1 but game set-ups start set at -1 scum favoured) Note How valuable a doctor is in terms of VTs is not likely at all to be a fixed number. (for larger games) TBMK 12 player setups with Doc are kinda balanced. BUT adding a doc to an 8 player setup will make a different amount of difference. (I think less) The reason is the number of days on which the Doc may save a player is lower in an 8 game than a 12. However the odds that a doc will make it to 3 man LYLO and claim (maybe eliminating the patsy...) is lower. While your question asks for an integer value for a goon, decimal fractions are not fatal. Neither is it implausible to just use a Lookup Table that specifies for Each sized Town How many goons it should have. Eg Town size 12 is 3 goons but is +1 goon favoured. Eg Town size 13 is 3 goons but is +0 goon favoured. Eg Town size 14 is 3 goons but is -1 goon favoured. ... but it might not quite follow that pattern later as the above discussion on larger games indicates.
There are yet other even finer distinctions. Eg Town size 13 is 3 goons but is +0 goon favoured. This game if it goes according to PAR and ends at a 3 man lylo, at one time it had (2 scum nks for each town mislynch & scum lynch.) x 2 thus it gets to 5 townies 1 scum. At that point there have been 4 scum controlled towny deaths and 2 town controlled ones. Scum has twice as much control over whether skilled and keen to play townies wind up making the next decision. Eg Town size 12 (1 doc) is 3 goons but is probably also about +0 goon favoured. This game if it goes according to PAR and ends at a 3 man lylo, at one time it had (2 scum nks for each town mislynch & scum lynch.) x 2 (-1 town nk if the doc saves 1 and is thus on par) gets to 5 townies 1 scum. At that point there have been 3 scum controlled towny deaths and 2 town controlled ones. Scum has less than twice as much control over whether skilled and keen to play townies wind up making the next decision.
Actually, the program already basically takes into account if Mafia auto-wins (if there are more or as many Mafia as there are town, Mafia victory odds are 100% and town victory odds are 0%). The same happens if it's night start, but town just requires one more player not to auto-lose. Although currently there's no way to account for the scenario where Mafia KP is 2+ and it's night start (where Mafia might auto-win even if the program shows that town has a chance). But, in the end, it's pretty simple to see for yourself if Mafia would auto-win - and a setup where Mafia would auto-win would probably show as pretty unbalanced in favor of Mafia anyways.
That's an interesting point about how adding an even/odd amount of players for town has a different level of impact. I actually read about this on Wikipedia the other day:
"Moreover, it was shown that the parity of the initial number of players plays an important role. In particular, when the number of mafiosi is fixed and an odd player is added to the game (and ties are resolved by coin flips), the mafia-winning chance do not drop but rise by a factor of approx. \sqrt{\pi/2} (equality in the limit of the infinite number of players)."
That might be a bit complicated to include in the balance tester, but probably not anyhow impossible...
Well, it's good that all sorts of matters are considered. The current system is very simple and easy to understand (so that the user can immediately understand how making different changes affects the balance), but it's definitely not 100% accurate. Then again, I'm not sure if any system can be...
It's definitely worthwhile to improve the accuracy of the program, though, so that it can be useful to more experienced users. On the other hand, I don't really want to make it too complicated. Well, there's no hurry to get it finished, I guess
|
I just made a small update to the balance tester:
- Reduced the value of Roleblocker to +2
- Altered some victory odds calculations for neutral roles
- A couple of small bug fixes
Version #3 can be downloaded here:
https://www.dropbox.com/s/zb7oxzsz177h4f7/Mafia_Balance_Tester.jar
iGrok's suggestions for reworking the balancing method and the user interface are going to be explored a bit later on (most likely at least), but I didn't have enough time today to even begin implementing those things.
Meanwhile, I'd appreciate more suggestions! Especially suggestions that are relatively easy to implement but still can have a considerable effect
|
This has my utmost thanks coming your way. When I finally earn the ability to host, this will be very valuable to me! I appreciate your work even not having used it yet! :D
Would this be able to account for nonstandard mechanics or is that like trying to determine the load characteristics of a one of a kind experimental aircraft?
|
I highly doubt non-traditional mechanics can be balanced like this.
Hell, it's almost impossible to balance them by hand.
|
|
|
|
|
|
EPT
