Txurce
Deity
Aeson, I endorse your having as much fun as possible, and look forward to reviewing your QSC.
Changing the curve
- Thread starter Aeson
- Start date
JonathanValjean
Porschephile
Originally posted by Aeson
The number crunching is actually something I enjoy doing as long as it's on my own terms, I'd hate it as a job...I have a very strange concept of 'fun' it seems.
It appears that your warped sense of 'fun' is our great gain! No, really, I'm glad that you enjoy it.
- Joined
- Nov 16, 2001
- Messages
- 2,686
After some more thought I think this is the curve that's going to be used. It's only good for 550BC (turn 98) on, but there shouldn't be too many games finished before then. If there ever is a game finished that early, it would just get curve=maxscore.
curve=maxscore - ((maxscore * 11/20) * sin((playerturns - 270) / 180) + (maxscore * 9/20))
Here is a graphical comparison of the new curve, old curve, and some of the GOTM18 games.
curve=maxscore - ((maxscore * 11/20) * sin((playerturns - 270) / 180) + (maxscore * 9/20))
Here is a graphical comparison of the new curve, old curve, and some of the GOTM18 games.
Ribannah
Fighter Druid
Proud to see that my game fits the new curve so well while it lasted. 
Aeson, could you give some info on how much difference it makes in Jason score when comparing speed games, fully milked games and buttermilk games?
I'm thinking that a cumulated probability density funtion with sufficient parameters (Weibull?) can probably be found by estimation that fits our games even better from 4000 BC onward.
As it stands, the race for a pre-550 BC victory in the GOTM is on!
Aeson, could you give some info on how much difference it makes in Jason score when comparing speed games, fully milked games and buttermilk games?
I'm thinking that a cumulated probability density funtion with sufficient parameters (Weibull?) can probably be found by estimation that fits our games even better from 4000 BC onward.
As it stands, the race for a pre-550 BC victory in the GOTM is on!
Does this mean that Bamspeedy, Bremp and SirPleb 's games would came out a little bit better than mine? Since they did a wonder job in their games and they were all better at this game than me (on the average, they were all beating me by about 1000 points at 1020 AD; but thanks to the old curve, I got the gold), I don't mind if they are taking over the spotlight.
About the curve, am I missing something or is it that the new curve will DECREASE the scores for the period 550BC to circa1000AD? Not that I care but it actually means that early victories will not add that much score as they used to. If you ask me I would take the upper side of both curves.
In the past I did not believe that the new curve would worth the time required to make it. But since you did, the better the system, the better the game (like most of the players have wrote).
In the past I did not believe that the new curve would worth the time required to make it. But since you did, the better the system, the better the game (like most of the players have wrote).
Ribannah
Fighter Druid
The formula as a whole should take into account that any element of speed will automatically mean that you drop below the curves of the milkers.
But I think that can be accomplished by making the target dates per victory type a little more conservative. I'd prefer that over a curve that is sharper than that of the real games, as the curve we had until know.
But I think that can be accomplished by making the target dates per victory type a little more conservative. I'd prefer that over a curve that is sharper than that of the real games, as the curve we had until know.
- Joined
- Nov 16, 2001
- Messages
- 2,686
The games I picked I picked because I felt they were all played at about the same level (very close to maxed out).
Fast Victories:
Going for 20k takes a rather big bite out of base score because you give up your best city (or one of them) very early on as far as contributing to territory aquisition. 100k will slow things down as well, because all those cultural improvements cut into the number of units to go beat down the AI. Conquest and Domination both need to sacrifice population to conquer the AI's territory quickly. Spacerace and Diplomatic both need to focus on tech rate, and so a good portion of the economy isn't available to the military for expansion.
All of them have their ingame score tradeoffs which will keep them slightly lower than the curve pure milking attempts can reach. So just because your line wasn't up there with Bamspeedy's doesn't mean the games weren't comparably well played. I think the way the curve modifies things needs to be changed to account for date. Something like a 10% 'leeway' very early on, fading to 0% at 2050AD.
There are some changes that need to be made with MaxScore, as it's consistantly beatable by about 5% on large landmass (in relation to overall landmass) starts, and about 5% too high on small landmass (same relation) starts. It causes too much of a swing between what type of game will score best from map to map. To fix this, we could use a modifier like:
NewMaxScore = OldMaxScore * (0.1 * ((StartingLandmass / OverallLandmass) - 0.5))
The problem with that is to get a utility to do the counting. An easier (and less accurate) way is just to account for it in the crossing level modifiers.
Fast Victories:
Going for 20k takes a rather big bite out of base score because you give up your best city (or one of them) very early on as far as contributing to territory aquisition. 100k will slow things down as well, because all those cultural improvements cut into the number of units to go beat down the AI. Conquest and Domination both need to sacrifice population to conquer the AI's territory quickly. Spacerace and Diplomatic both need to focus on tech rate, and so a good portion of the economy isn't available to the military for expansion.
All of them have their ingame score tradeoffs which will keep them slightly lower than the curve pure milking attempts can reach. So just because your line wasn't up there with Bamspeedy's doesn't mean the games weren't comparably well played. I think the way the curve modifies things needs to be changed to account for date. Something like a 10% 'leeway' very early on, fading to 0% at 2050AD.
There are some changes that need to be made with MaxScore, as it's consistantly beatable by about 5% on large landmass (in relation to overall landmass) starts, and about 5% too high on small landmass (same relation) starts. It causes too much of a swing between what type of game will score best from map to map. To fix this, we could use a modifier like:
NewMaxScore = OldMaxScore * (0.1 * ((StartingLandmass / OverallLandmass) - 0.5))
The problem with that is to get a utility to do the counting. An easier (and less accurate) way is just to account for it in the crossing level modifiers.
- Joined
- Nov 16, 2001
- Messages
- 2,686
Compared to the old curve it would increase scores from 4000BC to about 500AD. The old curve was unrealisticaly high during that period. Most games don't finish in that timeframe, so it won't affect much in that way.
All other fast finish scores would be lower because the old curve was far too low during that timeframe. There was a hedge (not balanced very well) against this built into the system that now would need to be removed. Otherwise the fastest finish type games are getting hit twice.
My idea is that instead of comparing fastest finish games to the actual curve, they are compared instead to a % of it based on the date. So a very early victory would be rated against Curve*0.9 and a midgame victory would be rated against Curve*0.95... The best dates would be factored in to keep things even from map to map.
If we keep the old curve, everyone would have to finish at about 14-1800AD to stay competitive, regardless of victory condition.
Ribannah
Fighter Druid
Aeson, that sounds good, better than my idea above of shifting the target dates. It should stay hard to beat a target date.
That said, please keep both target dates and maxscore just within reach - it feels so good to beat them!
- Joined
- Nov 16, 2001
- Messages
- 2,686
The 'leeway' would looks something like:
(BestTurn/PlayerTurn) * ((540-PlayerTurn)/540) * .15)
The first part is to show the 'intensity' of the early victory push. The earlier PlayerTurn is compared to BestTurn, the higher the number it gives. At BestTurn, it would be 1. The second part is the % of the game remaining * 15%. That means at turn 1, it would be 15%, at turn 180 it would be 10%, at turn 360 it would be 5% and at turn 540 it would be 0%.
So if you hit the best date at turn 180, you get a 10% leeway on score. If instead you miss the best date by several turns, you get less of a leeway, partly because you missed the best date (thus weren't focusing all your resources into the fastest finish), and partly because you've now had longer to 'catch up' with the curve.
(BestTurn/PlayerTurn) * ((540-PlayerTurn)/540) * .15)
The first part is to show the 'intensity' of the early victory push. The earlier PlayerTurn is compared to BestTurn, the higher the number it gives. At BestTurn, it would be 1. The second part is the % of the game remaining * 15%. That means at turn 1, it would be 15%, at turn 180 it would be 10%, at turn 360 it would be 5% and at turn 540 it would be 0%.
So if you hit the best date at turn 180, you get a 10% leeway on score. If instead you miss the best date by several turns, you get less of a leeway, partly because you missed the best date (thus weren't focusing all your resources into the fastest finish), and partly because you've now had longer to 'catch up' with the curve.
cracker
Gil Favor's Sidekick
I am really pleased with the progress made in this scoring effort and will be comfortable with a scoring system that continues to emphasize a balanced approach to speed and strong internal game mechanics of building a Civilization. This later element is what sets the game apart from First Person shooter and basic Hack and Slash games.
Fast finish games are rewarded by those medals. The key issue in balancing overall placement or ranking of the finished games is to make sure that the tradeoffs between the accumulated recognition of absolute power is given strength against pure speed. Absolute power as reflected by long term average of territory and population does provide an indicator of game skill. Within a range of finish dates that are close together by just a few turns, the more powerful game probably does reflect the stronger overall performance than the game that is one or two turns faster but with substantially lower accumulation of Civilization power.
This will be a long term classic argument that many people have chosen sides on. My perception is that pure speed games reflect a high skill attainment level that can be truly remarkable, but that hidden among the star athletes we can find a number of allowable and legitimate techniques that provide a very strong dependence on the luck of the draw in leading edge events. A lucky early conquest with a single archer or a miraculous sequence of 2 or 3 early great leaders can really impact speed in ways that not every game in the comparison pool can replicate. If we could control the random events so that they would appear random outside the game when comparing many diifent games on different maps but then allow for "random" event sequences that where more similar in a single game then this would be less of an issue.
As it is, I am very comfortable with a system that holds the ranking of the games close and then pushes the balanced well played game that reflects speed and power forward into the limelight. Speed in and of itself does not necessarily reflect power particularly in the bloody victory categories. An army of one is clearly infinitely more powerful than ten headless corpses but is that event set reflective of a more powerful game than a glorious civilization that also subdued its neighbors with the sword but 5, 10, or 15 turns later.
We have in the past seen examples of horrific pursuit of raw speed at expense of all personal morals, ethics, and basic game fundamentals and it took several months to clear out this detritis.
I also think that it is comforting for all players to be able to see the impact of choices of balancing speed and power as a reflection of performance in the game. A player in the mid range of the results can gain great reinforcement from know that they are at or near the different performance curves even if they are in 30th or 45th or 60th place etc. The performance curves also give the high end players an ongoing set of goals to shot at that extend well beyond just another head on the wall.
In the future, I see a process where players will be able to confidently list what they think are there best/favorite games that they have played and to recount key events and results from those games with a focus that includes a balanced mixture or tactitical and strategic genius events interwoven with stories of cool random stuff AND things that were just plain ole fun to do. I would personally find it fascinating to have SirPleb or Moonsinger (or almost any player) tell me what his/her list of his/her three favorite Civ3 Games would be and why and I would consider us lucky if a GOTM game would be honored to be on the list.
Fast finish games are rewarded by those medals. The key issue in balancing overall placement or ranking of the finished games is to make sure that the tradeoffs between the accumulated recognition of absolute power is given strength against pure speed. Absolute power as reflected by long term average of territory and population does provide an indicator of game skill. Within a range of finish dates that are close together by just a few turns, the more powerful game probably does reflect the stronger overall performance than the game that is one or two turns faster but with substantially lower accumulation of Civilization power.
This will be a long term classic argument that many people have chosen sides on. My perception is that pure speed games reflect a high skill attainment level that can be truly remarkable, but that hidden among the star athletes we can find a number of allowable and legitimate techniques that provide a very strong dependence on the luck of the draw in leading edge events. A lucky early conquest with a single archer or a miraculous sequence of 2 or 3 early great leaders can really impact speed in ways that not every game in the comparison pool can replicate. If we could control the random events so that they would appear random outside the game when comparing many diifent games on different maps but then allow for "random" event sequences that where more similar in a single game then this would be less of an issue.
As it is, I am very comfortable with a system that holds the ranking of the games close and then pushes the balanced well played game that reflects speed and power forward into the limelight. Speed in and of itself does not necessarily reflect power particularly in the bloody victory categories. An army of one is clearly infinitely more powerful than ten headless corpses but is that event set reflective of a more powerful game than a glorious civilization that also subdued its neighbors with the sword but 5, 10, or 15 turns later.
We have in the past seen examples of horrific pursuit of raw speed at expense of all personal morals, ethics, and basic game fundamentals and it took several months to clear out this detritis.
I also think that it is comforting for all players to be able to see the impact of choices of balancing speed and power as a reflection of performance in the game. A player in the mid range of the results can gain great reinforcement from know that they are at or near the different performance curves even if they are in 30th or 45th or 60th place etc. The performance curves also give the high end players an ongoing set of goals to shot at that extend well beyond just another head on the wall.
In the future, I see a process where players will be able to confidently list what they think are there best/favorite games that they have played and to recount key events and results from those games with a focus that includes a balanced mixture or tactitical and strategic genius events interwoven with stories of cool random stuff AND things that were just plain ole fun to do. I would personally find it fascinating to have SirPleb or Moonsinger (or almost any player) tell me what his/her list of his/her three favorite Civ3 Games would be and why and I would consider us lucky if a GOTM game would be honored to be on the list.
- Joined
- Nov 16, 2001
- Messages
- 2,686
Just to show how the new curve affects things:
EDIT: just noticed a major error... will repost in a bit
This is using the 'leeway' defined as (1 + ((BestTurn/PlayerTurn) * 0.15)). The initial ((540-PlayerTurn)/540) inclusion didn't add much of anything, so was taken out. So:
DateBonus = (1 + ((BestTurn/PlayerTurn) * 0.15)) * Curve
Curve = (MaxScore-((MaxScore*11/20)*SIN((Playerturn-270)/180) + (MaxScore*9/20)))
The curve is the one shown up above in green, just upside down as it's supposed to represent the score being passed up. For any victory condition before 2050AD, the score being passed up calculation is amplified. At the best date for the victory condition, it's 115%, meaning that to hit the best date, 15% of the score available at that date to a pure milk run is going to have been passed up in favor of speed to victory.
EDIT: just noticed a major error... will repost in a bit
This is using the 'leeway' defined as (1 + ((BestTurn/PlayerTurn) * 0.15)). The initial ((540-PlayerTurn)/540) inclusion didn't add much of anything, so was taken out. So:
DateBonus = (1 + ((BestTurn/PlayerTurn) * 0.15)) * Curve
Curve = (MaxScore-((MaxScore*11/20)*SIN((Playerturn-270)/180) + (MaxScore*9/20)))
The curve is the one shown up above in green, just upside down as it's supposed to represent the score being passed up. For any victory condition before 2050AD, the score being passed up calculation is amplified. At the best date for the victory condition, it's 115%, meaning that to hit the best date, 15% of the score available at that date to a pure milk run is going to have been passed up in favor of speed to victory.
- Joined
- Nov 16, 2001
- Messages
- 2,686
Bamspeedy
CheeseBob
Aeson, do you perhaps have a calculater we can use that is using this new scoring curve for inputting dates/scores from GOTM18?
I was looking at what some of our Jason scores would have been, had we finished at 1020 A.D. instead of milking, and what our jason score was at 2050 A.D. doing the full milk, and seeing what, if any benefits there was to milking. I would like to do the same for the new curve for comparison purposes.
I was looking at what some of our Jason scores would have been, had we finished at 1020 A.D. instead of milking, and what our jason score was at 2050 A.D. doing the full milk, and seeing what, if any benefits there was to milking. I would like to do the same for the new curve for comparison purposes.
Well Ribannah, congratulations! You finally got your wish. Remember the story about the Rabbit and the Turtle I mentioned in another thread... with the new curve, once the rabbit take off, it would be impossible for any turtle to catch up. In another word, if we mess up our opening moves, there is very little chance we can catch up.
You finally got your wish. Remember the story about the Rabbit and the Turtle I mentioned in another thread... with the new curve, once the rabbit take off, it would be impossible for any turtle to catch up. In another word, if we mess up our opening moves, there is very little chance we can catch up.
Ribannah
Fighter Druid
It's deja vu for me - the second time that I would have won gold had the formula been changed earlier! 
Of course the 15% is still kinda arbitrary, based on the idea that the top x players are about equally good. But since the speed players do their thing each month, and the milkers too, there is no way of knowing. Perhaps SirPleb would put us all to shame if he went for a Space Race!
Maybe we should all play a non-typical game next month to get a good feeling of how tough that really is ....
About messing up: that you can do in every part of the game.
Of course the 15% is still kinda arbitrary, based on the idea that the top x players are about equally good. But since the speed players do their thing each month, and the milkers too, there is no way of knowing. Perhaps SirPleb would put us all to shame if he went for a Space Race!
Maybe we should all play a non-typical game next month to get a good feeling of how tough that really is ....
About messing up: that you can do in every part of the game.
Originally posted by Ribannah
It's deja vu for me - the second time that I would have won gold had the formula been changed earlier!
Yes, it's deja vu for me too - this is the second time that I won gold because the formula hadn't been changed earlier.
For the record, I did change my style to adopt the Jason Scoring System. I normally don't build hospital until I discover Ecology; however, in the GOTM18, I rushed the hospital ASAP and I didn't even bother with Ecology or the mass transit. In away, I put all my cows on steroids so they could produce as much milk as they possibly could in a very short burst of time. Of course, the cost for that is inconceivable. If I disband my culture city and milk it all the way to 2050 AD. I would most likely end up with at least 500 points less than my domination score at 1758 AD.
SirPleb
Shaken, not stirred.
It looks like I'll be going for one of the eary victory conditions for sure this month, won't have time to go for the late ones. And it will make a fun change
