We Need New Honor Calculations!

[The Black Ship]

Richie in the News thread asked about new ideas for honor calculations in Multi-play. Sorta a "you don't like our way, why not tell us what you want!" challenge.

I remember Obake proposing using the Battle log, ala custom battles, as a starting point. Not sure how the Newbie versus Expert problem would be resolved though.

I also remember reading about the concept of a Honor Pool system, and a player would obtain his percentage of the dynamic pool.

I'm sure there's more! So post away!

[ShadowKill]

Damn I have nothing that comes to mind right now I kinda like the honour system right now other then people staying in the top spot and never playing. well as people propose things i will be here to give my opinion on what they have to say

------------------
Fear not the sword to your front, but the blade at your back.

[Slychaos]

Whilst I am thinking of a better method of calculating (formulae aren't my strong subject) I thought, "isn't a great deal of the problem with the honour system a problem because it can drop when you win or lose against an opponent with very low honour?"
Surely this would to at least some extent become moot when we can see the honour level of those we are fighting?
More as I think it up. But maybe using a league system as they do in Football, like +3 for a win, 0 for a loss!!


------------------
SlyChaos
HOST OF CHAOS

[ShaiHulud]

Just began multi-play and must say it's disheartening to me. In game one my opponent was demolished and waiting for the coup-de-grace when , inexplicably, I dropped. He got that win. Second game he won fairly, 3rd he dropped (no decision),4th I won. Against another I won next game. Against my last opponent I split two games with him. Final total? 1 win, 3 losses!
I expect to now hear that this is common, but it's a saddening initiation.
BTW, I have cable modem. Is there known to be any advantages to this?

------------------
Wind fells blossoms, rain
fells steel,yet bamboo bends and drinks

[Methabaron]

Ok, here you go my initial thoughts on the subject.

The battle score formula "per se" is not too bad. I would recommend anyone interested in knowing more about the current formula visit the thread
http://www.totalwar.org/ubb/Forum1/HTML/000059.html

here at the dojo. As Peptis suggests in that same thread there are other formulas like the chess one which I personally like better too because when you win you can never lose points and when you loose you can never earn points:

Quote delta = K/(1 + 10^(diff/50))
explanation:
*K would generally be a number around 5. It just scales the number of points given to or taken away from players.
*diff is winner's honour - loser's honour
*delta is the amount of honour added to the winner's total and taken away from the loser's.
In practice:
For an example take 2 players whose honour differ by 15. If the higher ranked player wins his honour will increase by 1.67 (and loser will lose 1.67), however if the lower ranked player wins their honour will increase by 3.33 (loser's will decrease by 3.33).[/QUOTE]


But as I said, the current formula is not too bad. The problem resides rather in rank control. At the moment there is none. The major points that should be adressed are:

1- Players (specially those in the top 100) should loose honour for not playing. How much and how often is open to discussion. I would say that deducting 10 points if the players has been inactive in a week. And do so every week after until he is not anymore in the top 100

2- ESC control. Tough issue this one. I already recommended that any ESCaper or team where one person escapes should be given an automatic loss. I know that many games have connection drops (adressed in my next point)and some people use ESC to restart but at least real losers will not be able to escape their destiny. This will keep the ranks more fair.

3- Droppers. The game should recognize a drop and cancel the game without winner/loser if it is a 1v1 or a competitive one. The number of drops should be recorded in the player history and if possible shown along with the honour in the lobby. Accidental dropers will be able to prove they are honourable players just by behaving so regardless of what their current drop mark is (they can play friendlies to prove it first oif necessary and we all know who the honourable players are after a little while). Intentional droppers (i.e. unpluggers) will think twice before unplugging now that people can see their marks.

Now, if you want to continue with the lately released official Clan League I would recommend:

1- All of the above, plus:

2- Not to allow 1 man clans in it. Let's say that a limit for a clan should be at least 3 persons for exemple. The system should not recognize a clan unless it has 3 persons in it.

3- Have the system to recognize a clan battle when 2 clans are playing against each other in 1v1, 2v2 or more if each side only has players from a given official clan. Only these battles will count towards the Clan League; the system should have the option to allow the players if it is a clan league game or a normal (non clan)ladder game.

I think Magyarkhan worked harder than most of us to find out about the flaws of rank control. I'd love to read his advice.

Have fun,

Metha

------------------
"...Violence is the last resort of the incompetent..."

[This message has been edited by Methabaron (edited 11-25-2000).]

[Link Shumeisan]

I also like the Peptis formula and you have been quicker to quote it Methabaron
One other advantage of this formula is the balance between honour gained by the winner and honour lost by the looser.
A true ELO rating system can also be used but would give much work to the developpers and could necessitate the implementation of rematches to avoid map difficulty biased scores

For the other points:
1&2. I agree totally with this 2.
3. It would be good, but I think some players could simply disconnect to avoid a D.
I would add a
4. If a "Peptsi formula" can't be implemented, why not simply give a score equal to the current honour of the player +1 if the score of the game is less than the honour of the player who won. This method has some side effects but it would allow to play every player. I don't like to say "No I can't play you your honour is to low, do you want a friendly ?"
5. One other point on the honour system is that the score doesn't take into account the difficulty of the map. You have the same score when you win as attacker or as defender on Shinano. I would like to see a difficulty factor given to a map. As it was stated in a previous topic, this difficulty factor can be calculated using the data stored in the database :
eg just to show the tendency:
p= Constant*number of V as defender on the map/ number of V as attacker
Constant determinated to have "fair" scores

(it should also take into account the Honour of the Att and the Honour of the Def in each battle but I don't know what is recorded in the Database)
Defender win :
Winner Score=Honour of the looser+50*(1+1/p)
(Min=Winner's honour+1)
Looser Score=Honour of the attacker-50*(1+1/p)
Attacker win :
Winner Score=Honour of the looser+50*(1+p)
(Min=Winner's honour+1)
Looser Score=Honour of the attacker-50*(1+p)

With the capped things*1+p (or 1+1/p) used today.

Just for information there is a topic on EA-forum where all as been discussed : http://pub24.ezboard.com/fshoguntotalwarfrm1.showMessageRange?topicID=628.topi c&start=61&stop=79


For the Clan league I totally agree with Methabaron suggestions.


[This message has been edited by Link Shumeisan (edited 11-24-2000).]

[Puzz3D]

I don't think taking away honor from inactive players is really fair. An alternative would be to move players who haven't played in a while to an inactive list. This time period could be as little a 1 week, but maybe something a little longer, like a month, would be more reasonable.

The honor scale is too compressed. It effectively runs from 50 to 150. In chess, the scale runs from 0 to 3000 with the mean at 1500. A difference of 50 points represents a substantial difference in playing strength. That's equivalent to 1.7 points in the Shogun system.

In chess you have to win 3 games from an equally rated opponent to get 48 points. In Shogun, you get 1.7 for a single win over someone with equal honor. That's equivalent to 50 points in the chess system. Because only the last 30 games are averaged, each game of Shogun is heavily weighted. A player's position on the ladder is going to jump around quite a bit, but their actual playing strength can't posssibly be fluctuating like that.

Once again, in chess, the first 30 games are use to establish a baseline rating based on the ratings of your opponents (I don't remember the exact formula). After that, every game played contributes to your rating as quoted by Methabaron. If you beat a grandmaster 3 years ago, those 32 points gained are in there contributing to your overall rating forever. Your rating tends to be a lot more stable in a system like this, and yet, if your playing strength does change, your rating will gradually shift to reflect this. The rating can become inaccurate if you don't play very often.

Shaihulud: I spoke to you right after that first game the other night. Sorry to hear you were tagged with a loss for that. Your opponent in that game has a high rating of 133 (18 wins/1 loss). You only lost about 1/2 point off your honor for that. I just checked, and the database has you with 1 win and 2 loses for the night. So, that's a little consolation.

[ShaiHulud]

Puzz...Guess to the well-read 'GodWorm' is obvious..hehe.. Now I have to figure out what your nomme de guerre is!
And, yes, it's somewhat mollifying to hear that result. Some flaw in the system must have ignored the other games that should have registered. I lost but one point for the two losses.

------------------
Wind fells blossoms, rain
fells steel,yet bamboo bends and drinks

[Methabaron]

Bump !!

We definitely need some good input in this one guys.

Metha

------------------
"...Violence is the last resort of the incompetent..."

[The Black Ship]

Perhaps honor can be adjusted by dropping games from the honor calculation on a monthly basis. This would stop the "I'm in the top 10" campers.
Your overall record would remain, but your honor would change on a month to month basis. Heck the new player will like this system since in his first month of play he'll undoubtably get spanked a few times, while good players will remain good players I'm sure.

[Link Shumeisan]

I agree Blackship
If you have a Top of the 100 players par month/per week/year it would be a great improve and would not necessary necessitate to remove honour to the players who dont play.

I have to repeat all the stats that I'd like to see in the Shogun topic, but I am tired now and will do that tomorrow

[Magyar Khan]

a good ranking system is promised to me by EA/CA when SubudaiLives cheated the ranking. Since then nothing happend.

[Puzz3D]

BlackShip: I think the monthy honor calculation is a very good idea. It would be much better than what is currently posted. A player with less than 30 games in the last month could have the missing games filled in with the standard 100 point values. If people feel that 1 month is to short a time period, then just make it something longer. As stated in another thread, you could have several rankings for different time periods.

[Methabaron]

Geez Magy,

We all are in awe before your wisdom ... we all know that you spare us words but... can you please elaborate a bit more and tell us what exactly you would like to see in an improved rank control and honour system?


I know that you are the one that has tried harder to find any potential flaws and you proved it with subudailives.

Metha

------------------
"...Violence is the last resort of the incompetent..."

[This message has been edited by Methabaron (edited 11-27-2000).]

[Methabaron]

Blackship,

I am not sure of understanding your proposal. You want to make only the games played in the last month count towards the average that calculates the honour?

The current honour is the average of the last 30 battles. You mean you want to drop part of these 30 battles for the average?.

Examples please.

Thanks,

Metha

------------------
"...Violence is the last resort of the incompetent..."

[Link Shumeisan]

Here are my thoughts about Top 100 players of the Month/Week/year/Global :

1.Get rid of the average of the scores of the last 30 games
2.Global Rank (GR):
Average of the scores of all the battles of a player using the following formulae.
ndsb : number of days since the beginning of the ranking system recording has started
nb : number of battles of a player

If nb

[Methabaron]

Shumeisan,

Correct me if I am wrong but you seem to suggest an average of the TOTAL number of battle scores obtained to date (using 100 points for "missing" battles if there are less battles than days) for the GLOBAL rank.

The problem with that is that averages on the long run over a big number of battles tend to be quite... fix. No matter what you do, the more you play the more is going to cost you modify your average !

Example: Assume I have played 250 games so far (in less than 250 days for the sake of argument), and that all those scores total a number of 27500. Now my Global honour according to your formula is 27500/250 = 110

Now, if I play another game, no matter how good or bad my score, it is going to be almost of no effect to my score. Assuming I win and my score is 180 which is very good and much higher than my average. My global score will be 27680/251 = 110.2. In other words i would not see any significant change for such an impressive victory.

In a way, the higher the number of battles you average the worst the system works in updating your level if you improve or worsen.

Currently we use 30 which is not too bad. But when you reach over 60 or 70 values averaged then the resulting average start to be very hard to modify.

So I would recommend further discussions on the chess system because it will give you increase in ranking right away after a win and it will do so in a fair manner. The negative side of the chess system is that it may make high rank players not want to play with low rank players because of the higher risk involved for them. But in the other hand, we already have this same problem now and it has always been in all ladders (except the one at Case's http:\www.igl.net where a higher rank player loosing with a lower rank one will never loose his rank, only the lower rank one will improve) so I dont think it will be a real problem, specially when all players can see their ranks after next patch and can choose opponents.

Metha

------------------
"...Violence is the last resort of the incompetent..."

[This message has been edited by Methabaron (edited 11-27-2000).]

[This message has been edited by Methabaron (edited 11-27-2000).]

[Link Shumeisan]

Metha,
You are right for the Global rank and you will have the same thing using a ELO ranking system except that the scale is wider. To see more variation you can just multiply the current rank by eg 20. So with your exemple you will have :
before : 2200
after : 2204
You can see a thiny variation.
But I think that it is a good thing that a single victory couldn't give you a too high jump in ranks. If you really want to jump in ranks you will have to have a lot of good results.
In another hand, your good score will have a bigger influence on your Month rank and even bigger in your week rank.
eg Month (30 days) rank :
40 battles and 110 score average (4400)
one 180 score battle : (4400+180)/41 =111 or 112 (depending on the rounding rule)

eg Week rank : 9 battles and 110 score avg (990)
one 180 score battle : (990+180)/10 =117

I think that if we keep a similar score calculation as the current one the advantage of the calculation that I described is that you can see variation in Month and Week ranks but you have to keep make good score to be high ranked in the global rank.

Don't forget that with the current ranking system you have your biggest jump in rank when you put out of your average your defeats and that your great V will go out of the average one day if you keep battling (not for rank campers):

eg current system :
30 battles with 110 avg your 1st battle was a victory with a 150 score.
one 180 score battle -> 111
(30*110-150+180)/30
hehe NOT MUCH DIFFERENCE

30 battles with 110 avg your 1st battle was a defeat with a 50 score.
one 180 score battle -> 114
(30*110-50+180)

Now imagine your 1st was an exceptionnal 200 V
one 180 score battle -> 109 honour
(30*110-200+180)
you have loose one point for your great V !!!


which can lead to more Escapers than in a system where you keep all the battles results for the average especially if you see the drop/esc stats in the tables.

[This message has been edited by Link Shumeisan (edited 11-28-2000).]

[Avenging Warrior]

Theoretically, that shouldn't be a problem. With a fair ranking system (such as the chess system), if a player is ranked lower than you, it means he's not as good as you, right? So you have better chances of defeating him. And if you do lose to a player who isn't as good as you, then you deserve to lose more honor, and he deserves to gain more honor for beating a tougher opponent. So it all looks pretty well-balanced to me. Of course, the key here is a fair ranking system, so a lower ranking really DOES mean that you're worse than the higher ranked players...

------------------
Life has no meaning. Do not concern yourself with it.

[TosaInu]

Hello,

Extra honour points should be gained for the map you play.
I.e. Attacker easy 2
Attacker medium 4
Attacker hard 6

Defender easy 1
Defender medium 2
Defender hard 3

On totomi, an attacker will get 2 extra points, defender 3.
In a bridgebattle, the attacker will get 6 and the defender 1.

I don't mean whole honour points, I just picked numbers to illustrate. But there should be a bonus. This is not impossible to implement as all necassary information is already recorded in the logfiles.

Yes the drop thing is very annoying: I have a new loss and the battle didn't even start.
The ranking system did record that it didn't start, but I still got a lost.

[The Black Ship]

Too bad about the loss...seem to be hearing a lot of the same gripes on this front

I like the province rating scheme, some maps don't deserve full credit for camping defenders- while a victory on one of these maps surely deserves a honor boost!

[Slyspy]

My biggest gripe is the honour loss suffered when beating someone of lower honour. My own score is under 100, and I am fed up with higher ranked players not wanting to play me because they will lose their hard won honour no matter whether they win or lose the battle. Its even worse when your allies show a reluctance to fight alongside you in 2v2! I can see why it is important to stop newbie-hunting for easy points, but why not simply leave the points as they were rather than take them away? Notch up another win for the veteran, but leave his score untouched. Stop the negative honour bonus, because it damges the community by encouraging a two-teir system in the rankings.

------------------
"Put 'em in blue coats, put 'em in red coats, the bastards will run all the same!"

[ShaiHulud]

My suggestion is this..create a tiered honor system. Old pros with many battles would belong to the one, newbies with under a given number of battle would belong to the other tier. No honor would be lost by the higher tier players, just a new honor roll for those in the other, new, tier. Achieving the higher tier would be done by playing 'X' number of games competitively. Movement in either tier would, as before, be determined by the Honor system presently used.
Perhaps it doesn't seem like much but a pro would lose little in standing among HIS peers by a loss to a low honor player and the wrecking of the honor system by cheats would be unlikely because of a high number of games needed. This would give time for cheats to be broadly known instead of as now when 40 games buys the top slots. Would a cheat be able to win 200 games and not be spotted? I doubt it. Folks would stop playing against him.
The problem of cheats would still exist in the 2nd tier of new players but they have a lot to learn about the game and honor and can, with discipline and play, leave the cheats behind.

------------------
Wind fells blossoms, rain
fells steel,yet bamboo bends and drinks

[Vanya]

Hmmm... seems that if you make your new battle score equal to or greater than the score of the 31st previous battle (new one included), then you can prevent loss of honor to opponents with lesser honor. An example then would be:

Battle n-30: 155
Battle n-29: 140
...
Battle n: 180
Your honor=130

NEW Battle: Opponent is honor 20, so your score is 20+50

[Rob]

Yes it does. This a totally sensible idea and I don't know why nobody thought of it sooner

------------------
AkaiRob Hojo
Honour to Clan Akai Ken

[Slyspy]

Whatever happens do not implement a two tier system in the rankings. Such a distinction between vet and newbie can only decrease accessibility and thus smother the growth of the online STW community. I repeat, why not simply remove the possibility of losing honour for winning? Quick, uncomplicated and everyone understands!

------------------
"Put 'em in blue coats, put 'em in red coats, the bastards will run all the same!"

[Methabaron]

Hi there,

Quote Originally posted by Vanya:
Hmmm... seems that if you make your new battle score equal to or greater than the score of the 31st previous battle (new one included), then you can prevent loss of honor to opponents with lesser honor.[/QUOTE]

That is pretty much correct.

Quote [B]An example then would be:

Battle n-30: 155
Battle n-29: 140
...
Battle n: 180
Your honor=130

NEW Battle: Opponent is honor 20, so your score is 20+50

[Methabaron]

Admin please feel free to delete this.
Metha

[This message has been edited by Methabaron (edited 12-05-2000).]