Site scoring system (Feb 2nd 2013)

[bitwise]

I play a lot of tactical Gunboat games. Very fun, and not as time-consuming as normal Diplomacy (though I make sure to always be in at least one of those!) But, I don't see a way to see rankings using a filter of Gumboat. Wouldn't it be easy to rank people based on Gunboat-only games?

ah - yes this is a tricky one. I have had a think about it on a number of occasions, but I keep running into the complexity of how to do it without opening pandora's box.

Once again, this is nothing to do with the new system, which is exactly the same as the old system. But here is the issue that makes it tricky.
At the moment, you can show game results and scoring (will be rating in the new system), and you can select by 'variant'. So you can have All, Classic, Fleet Rome, Chaos, Build Anywhere, AoE. In the new system you can also see the TrueDip rating. Now could we add Gunboat? Yes, of course - but of course gunboat is an option that can apply to All, Classic, Fleet Rome, .... So do we show Gunboat for each variant? Or just a total value? And then what about FoW and Stuff which can ALSO apply to all the variants. And then what about the different map variants, which apply to all variants and also to PPO, Gunboat, Stuff, FoW, .....

Essentially, the way to do it would be to have something akin to what we have when you do a game search, where you select the exact combination of values you want. BUT this now means we have to keep a score record for each variant/map/game type combination, which starts to mount up exponentially.

Not a very satisfactory answer, and something that does trouble me, but I honestly am not sure it is worth the investment to sort it. And since it is not a feature in the current system and does not seem to have been desperately missed...

Right, sorry it's in the new system thread. You could - if you did it- track only the very most popular combinations. As for me, I'd vote for just the Gunboat flag - on or off.

+1. This is a very different type of game, but one that's very popular. No need for lists with FOW, Stuff etc. Don't know how popular PPO is here, but it might deserve it's own list as well. Public Press has been around pretty long type in the hobby.

[BlueBorjigin]

1) The long list of characteristics in the OP of this thread explains HOW the rating system will affect you as you play your games. Just as the Chess elo rating system does. It does not tell you exactly how it is worked out, but as I said in the chess example I am of the opinion hardly any chess players could tell you anyway. But the rating system works just fine, and I believe ours will too.

So there is little to be gained and more to lose by publishing the algorithms. The only gain you have come up with, which in my mind is pretty tenuous, is that if a maths geek were to study the games and results they might be able to infer the algorithm and somehow get an advantage from that. I would be fascinated to see that in action - particularly since in order to work out the results from each game in rating terms the person would have to take at least daily or intraday snapshots of the ratings table to see what change each player had for each game. This is something individual players can see, but seeing someone else's information is much harder.

2) The final reason for not publishing the algorithms is that I guarantee that whatever algorithm we published, there would then be a never-ending round of criticism from people who don't like it, or want a different one, or whatever. And I will also guarantee that there would never be any consensus. If you want some evidence of this, take a quick scan of any one of the multitude of scoring system debates on suggestions - warning though, they can easily be 20 pages long, so you have a lot of reading .

3) Does this help? I am not trying to be difficult - just trying to do what I personally feel is best for the site. I could easily be wrong, and have been many times before. But I do not know any other way to progress something when you have no sense of consensus. The alternative is nothing ever happens .

1) You make a strong argument.
2) And this is a winning case. Okay, I'm satisfied with your plans.
3) Lol, Dipsy, I think you've earned our trust by now XD

Would you be able to respond to this:

Also, if draws can lower your score, can solos lower your score?

Thanks

[asudevil]

Posting so I don't lose it.

And I think he has said (maybe in the mod pre-briefing) Solos can NEVER lower your score...after all your score reflects based on how you were expected to do...they can't expect you to do MORE than solo...

[super_dipsy]

A solo win can never lower your score.

[BlueBorjigin]

Awesome guys, thanks

[sinnybee]

The idea is that your rating will stabilize around your current playing strength, gradually increasing as you become a better player.

Or decresing if someone puts less effort into their games .

...I like the shield and charging idea!

[diploguy]

Any consideration for awarding a higher rating increase for solos from countries that are at the bottom of the solo stats? (i.e. a solo playing Italy versus playing Russia)?

Also, any consideration that in a game that features 0 NMRs/CDs there will be a boost in rating points for all players? A way of acknowledging the spectacular gamesmanship as well as the harder game played?

[asudevil]

Any consideration for awarding a higher rating increase for solos from countries that are at the bottom of the solo stats? (i.e. a solo playing Italy versus playing Russia)?

Also, any consideration that in a game that features 0 NMRs/CDs there will be a boost in rating points for all players? A way of acknowledging the spectacular gamesmanship as well as the harder game played?

To the first no...because you will get every country approximately the same amount of times.

To the second...not a bad idea, but no...because then a spiteful NMR late game from someone losing to prevent a perfect game for the winner would just cause problems.

[sinnybee]

4. What’s a good rating?
• Ratings in the 1500-2000 region can be considered to be very respectable. Ratings of over 2000 are likely to put you clearly in the top level of players on the site

I'd guess that your descriptions will likely better apply to 1400-1800 and over 1800 for the site.

When I played Euchre online, there were a very small amount of players in the 1900-2050 range, but it was nearly impossible to get enough players in the 1750-1850 range to play against to be able to rise higher.
I believe that as with Euchre, at this site we don't quite have as many players / games being played / spread in skill as there is in chess to be able to support a large enough pyramid to allow enough players to climb to a high rating.

[super_dipsy]

I'd guess that your descriptions will likely better apply to 1400-1800 and over 1800 for the site.

I honestly don't know how things will turn out, but in the modeling and tuning I paid quite a lot of attention to the history of the site. I realize you can't claim this is an accurate comparison because as has been said before, if you knew the scoring rules in place you might have played differently and chosen your games differently. But it is the best historic data to hand and there is a LOT of it (the old 'never mind the quality, feel the width' argument ). After all, we are almost at 25,000 completed ranked games on the site now.

If I just pretend the old data run through the new system has some value, then if we just stick to players who have played 5 games or more (so discounting people who come in, try a game or two and don't like it and leave, which tend to overwhelm the 1000 point area of the curve ) then if you were >2000 you would be in the top 0.5% of the site. If you achieved 1500-2000 you would be in the top 6% of the site. So I am reasonably comfortable on that data that

• Ratings in the 1500-2000 region can be considered to be very respectable. Ratings of over 2000 are likely to put you clearly in the top level of players on the site

has some merit. I would say top half a percent is clearly in the top level of players, and top 6% is not at odds with a rating that is very respectable. The reason I chose the numbers I did was I was mentally think something along the lines of top=0.5% and very respectable=5%, but the numbers were nice and round at the 6% mark. However, your statement is no less true, since the wording of 'top' and 'very respectable' is so imprecise.

So in the groupings you chose, the old data shows >1800 = 1.5% and 1400-1800 = 8.3%. I would be just as comfortable calling the first group the top players, and the second the very respectable ones. However as I said, the old data is by no means an indication of what will happen. We wait and see - exciting isn't it