Changing the curve

[Aeson]

The new curve still gives highest scores around that date for most victory conditions. The old curve was just a little bit outrageous in how much discrepancy there was.

Please, no GOTM19 spoilers in this thread until after submissions are closed though.

 

[Ronald]

I was looking into the file with bamspeedies and Bremps scores for every turn of the game. This actually leaves me quite uncomfortable with the scoring system and the new curve:

The scores are almost the same after they reached domination limit, no matter which victory condition (+- 100 points with the new curve, +- >1000 points with the old one).
So I conclude, that the curve represents a well miked game for every turn. So far so good, but where comes the victory condition in place? There should be a difference between a domination victory in 1700 AD and a 20k cultural victory at that date. Looking at the provided excel sheet it hardly does.

It even suggests, that you get a higher score being slower. Increasing your population outweighs being later for your goal.

Not to mention the victory conditions hurt most: dipomatic and space ship. Getting the fastest time (best game?) prevents you from getting to domination limit, therefore your Jason score is lower.

I still think, that the Jason score is much better than the civ score, but there are still the flaws I mentioned above. I hope someone can come up with ideas to solve them as well

Ronald

 

[Aeson]

Please edit out the GOTM19 spoiler info Ronald.

I think the reason why the best dates aren't showing up as much is because the leeway degrades in a linear fashion. It should drop off pretty steeply after the best date most likely.

It is possible to keep a fast tech rate and still hit the Domination limit at a decent date though. It will depend map to map which victory condition and timeframe are 'best', and mainly I'm trying to limit the variation there. The awards are for fastest finish, and that should keep interest in playing extremely fast victories even if on some maps they lag behind hybrid or fully milked games. The main thing is to try to give players the widest range of playstyles possible that they can still be competitive, and still have an accurate measure of the skill (and/or luck) that was a part of their game.

While very fast 'turtle tech and diplomacy' type games are very impressive in their own right, games which show similar levels of skill in many areas: expansion, conquest, tech rate, empire management, and/or diplomacy will and should score as well (or better). Civ is a 4x game after all.

 

[Ribannah]

Originally posted by Aeson
Please edit out the GOTM19 spoiler info Ronald.
While very fast 'turtle tech and diplomacy' type games are very impressive in their own right, games which show similar levels of skill in many areas: expansion, conquest, tech rate, empire management, and/or diplomacy will and should score as well (or better). Civ is a 4x game after all.

All those skills are also needed in a speed game. Plus, in addition to knowing how to do all those things, in a speed game you also have to know at what pace you can do it and when you should stop. No mistakes there or it really hurts.
So IMHO some of the aspects of the game are in fact only used to the fullest if you do play for speed, and speed games (of any victory type) should be contending for medals just as any other well-played approach.

 

[cracker]

Ribannah,

Speed games of each type will continue to have their own medals.

The point you continue to needle and pick at is that you think speed should be the exclusive arbiter of success and this opinion may be deliberately self serving in that it fails to ignore the need for a measure of game skill mechanics that are less dependent on raw luck, exploits, and potential temptations to just plain cheat.

I am personally very, very comforatble with systems that do not focus purely on speed and the measure of the best played game in an overall sense of the game. The foundation skill being rewarded here is "building a civilization" and not just destroying all your enemies.

As long as the requirement for demonstrating thoughtful and commanding game skills within the progress of the game is present, it provides us with a strong and balanced community that is not so cut throat as to provide significant reinforcement of many negative behaviors that I am glad we do not have to deal with very much.

 

[Ribannah]

* Sigh. *

As usual, you deliberately misinterpret my intentions.

And you know very well that I have not given in to any 'temptation' in all of my speed games; the same goes for about everyone else.

Why can't you say something constructive, instead of always hammering other people's views?

 

[Aeson]

All those skills are also needed in a speed game. Plus, in addition to knowing how to do all those things, in a speed game you also have to know at what pace you can do it and when you should stop. No mistakes there or it really hurts.

My statement was refering to 'turtle' type games where the player ignores expansion and runs only a minimal empire. It cuts the scope of gameplay down, even if the tech rate is maximized, other parts of the game are ignored almost completely. So when comparing a very fast, small empire game to a very fast, huge empire, the latter will score more.

It should be noted that conquest games which only focus on killing the AI will also score lower than conquest games which are in the same neighborhood of speed, but also build while conquering.

So IMHO some of the aspects of the game are in fact only used to the fullest if you do play for speed, and speed games (of any victory type) should be contending for medals just as any other well-played approach.

I never said they shouldn't. They will contend for medals in many cases, and obviously have the awards. They won't be the only way to play for a medal though.

---------------------

I agree with cracker that pure speed games (especially conquest) are heavily influenced by luck, regardless of skill level. The longer a game plays out, the less luck becomes a factor.

We aren't trying to punish good luck, or reward the bad luck, just to give those who do run into bad luck and still play well options to compete. The player who doesn't get the Settler from the hut, or doesn't get a Leader until their 20th elite victory... who's AI doesn't build the Pyramids for them, ect. should still have a playstyle available to them where they can end up scoring well. Hybrid and fully milked games give a player longer to catch up (of course skill and/or luck is needed to do so) and minimize the impact of luck on their early game.

The goal of the scoring system is to balance early, hybrid, and milked games which are played with comparable levels of skill.

 

[Ribannah]

Originally posted by Aeson
My statement was refering to 'turtle' type games where the player ignores expansion and runs only a minimal empire.

Ah, I see. Playing like that will normally not yield a really fast finish though except maybe for 20K culture.
I still like the idea of a special award for OCC.

I agree with cracker that pure speed games (especially conquest) are heavily influenced by luck, regardless of skill level. The longer a game plays out, the less luck becomes a factor.

More influenced, yes, because in a longer game good and bad luck will tend to average out. But depending on luck as Cracker says would be plain silly and only lead to far worse results on average. It will be a long wait before finally everything comes together in a monthly event!
Instead it requires a lot of skill to minimize the chance that bad luck completely spoils your speed game, while still keeping an eye out for really profiting from any good luck that comes your way.

We aren't trying to punish good luck, or reward the bad luck, just to give those who do run into bad luck and still play well options to compete. The player who doesn't get the Settler from the hut, or doesn't get a Leader until their 20th elite victory... who's AI doesn't build the Pyramids for them, ect. should still have a playstyle available to them where they can end up scoring well.

Hmm, the 20th elite victory already - I'd almost sign for that.
But I agree, of course. And not just from bad luck, one should also have a chance to recover from a mistake. As long as we don't go back to before Jason when that would guarantee you passing someone else who really fared/played better, only because your game lasts longer.

The goal of the scoring system is to balance early, hybrid, and milked games which are played with comparable levels of skill.

Exactly.

 

[Ronald]

Hi Aeson,

I was playing around with the calculator for gotm19. The more data I compare, the more questions I have. Since all would contain gotm19 spoilers, where can I post these questions and observations?

Ronald

 

[Aeson]

You can PM me about it, or post them in the appropriate GOTM spoilers thread.

 

[Ronald]

HI Aeson,

I posted my observations/question in the gotm19 spoiler3 thread (game submitted).
Maybe you can answer the questions there or open an appropriate thread.

best regards,

Ronald

 

[Aeson]

Quoted from Ronald in the Gotm19 End Game Spoiler thread
Your have not adressed and explained the fact, that every other victory than 20k and 100k cultural score less and the others more with the new curve at 1600 AD, so I still express my doubts about the validity of the new Jason system

I did:

"As for cultural games scoring 'lower', yes, they score lower than they would have with the old curve. That is because they scored much higher than they should have with the old curve. In fact, any victory condition with a best date around 1600AD would (the range was actually ~1200AD-1800AD for the 'worst' discrepancy). The old curve was the furthest off at that date, and the best date portion of the calculations was using that discrepancy and amplifying it. So basically, any victory condition at that date would have scored more (relative to skill/luck) than it would at say, 1000AD (or any date outside that 'off' range). And any victory condition with a best date around that date would have scored more (relative to skill/luck) than at 1000AD too. In GOTM18, a cultural game around the best date was off in both regards, both the curve and the best dates were giving it too much of a boost."

 

[Ronald]

Originally posted by Aeson


I did:

"As for cultural games scoring 'lower', yes, they score lower than they would have with the old curve. That is because they scored much higher than they should have with the old curve. In fact, any victory condition with a best date around 1600AD would (the range was actually ~1200AD-1800AD for the 'worst' discrepancy). The old curve was the furthest off at that date, and the best date portion of the calculations was using that discrepancy and amplifying it. So basically, any victory condition at that date would have scored more (relative to skill/luck) than it would at say, 1000AD (or any date outside that 'off' range). And any victory condition with a best date around that date would have scored more (relative to skill/luck) than at 1000AD too. In GOTM18, a cultural game around the best date was off in both regards, both the curve and the best dates were giving it too much of a boost."

OK, just one more question: If the old curve was too high around 1600AD, then I should have been too high for all victory conditions (it's the same curve). So why score all other victory conditions higher with the new curve at this date and only cultural victories lower?

 

[Yndy]

Following Ronald's experiment (in the Gotm19 End Game Spoiler) I realised that scores are close for any victory type while, marginally (10%) favouring victories before the target date. So it would not make much of a difference what victory condition do you choose after the latest date, the score would be similar. Was that the aim Aeson?

I have a feeling that this discussion would continue and needs to be moved in a separate thread.

Edit: Just made some trials to see how would my Jason score modify should I have delayed domination, and milked the game for up to 30 turns. The result was that the Jason score would have remained the same for 45 ingame points per turn which seems reasonable to me. But I have to disagree with you Jason, about the score increase, 50 ingame points per turn are very doable in the 30 turns after you discover Steam power. Plus, if you can relax the war budget and build some marketplaces in the newly conquered lands...
My preliminary conclusion is that under the current Jason the ideal is to get to your victory condition asap and then wait for some turns to make some easy ingame points.

 

[Aeson]

OK, just one more question: If the old curve was too high around 1600AD, then I should have been too high for all victory conditions (it's the same curve).

The curve was/is modified by the best dates, and so isn't the same curve actually. It was too high for all victory conditions that had best dates around 1600AD. GOTM18 those were the Cultural victories.

 

[Aeson]

Following Ronald's experiment I realised that scores are close for any victory type while, marginally (10%) favouring victories before the target date. So it would not make much of a difference what victory condition do you choose after the latest date, the score would be similar. Was that the aim Aeson?

For games that are quick to the domination limit, yes. The reason is that a game already 'won' shouldn't vary too much even if the player delays winning. This has a side effect of giving more leeway to games that miss the best dates, but is offset by the fact that usually those games are going to have lower base scores to begin with. If they don't have a lower base score, then they'd be at the domination limit early like the other games.

EDIT: Spoilers removed after the post was moved

 

[Renata]

Perhaps the issue is the discontinuity in civilization power around the time of steam power and hospitals? (That is, power seems to develop fairly .. linearly isn't the right word, but I'll use it anyway ... until steam power, at which point it almost instantaneously undergoes a huge acceleration before eventually levelling off again. It seems like if a game was to finish just prior to this point, it would indeed provide a significant in-game score boost to delay winning the game until *after* the post-steam-power/hospitals optimization, and that no reasonable scoring penalty based on victory date could possibly keep up with it.

That said, I have no idea how the issue could be gotten around, because steam power/hospitals are not going to arrive at the same time in every game, and the discontinuity can therefore not be timed accurately.

Renata

 

[LKendter]

I don't think you can get a perfect curve no matter what is done.

What I do want to see is wins that are not 2050 milk games. I simply don't have the time to compete and hit the top slots if I have to play until 2050.

 

[Ronald]

Hi Aeson,

Sorry for bothering you once more. So far I understand the following: The old curve was too high during the time between 1200 and 1800 and therefore you lowered it.
Since 100k victory has the best date around that time it got a higher boost in score (for playerturn = bestturn, datescore = curve)
So far so good

In your formula: Finalscore = (Datescore + Basescore)/Maxscore)*10000
The only change between the old curve and the new curve is in Datescore.

{raw data that may or may not impact this discussion has been left in the Gotm19 spoiler discussion}

I understand, that with a lower value for curve, I must get a lower value for datescore and the difference is highest at or even before bestturn.
What I still do not understand is, that how it is possible that a game with all the same data: Same year, same victory condition, same civ points can score higher with the new curve than with the old?

Ronald

 

[Aeson]

To put it as simply as possible, in the old curve some games were being hurt, and some were being helped. The new curve tries to bring things closer to the middle. So the games being hurt previously should score more, and the games being helped previously score less. (comparing games of different victory types in relation to each other)

The goal is that over the course of a consistantly played game, the player can choose any victory condition available to them at any date, and score at as close as possible to the efficiency that they would at any other available date and victory condition.