A proposal for a slight modification of Jason scoring

[klarius]

Over the last few months I saw quite a few complaints about Jason scoring favouring early domination/conquest.
But I haven't seen any constructive proposals.

I analyzed the formula a bit and have a proposal for a slight improvement without altering the basics.
Currently part of the formula is (taken from the calculator):


jasonbonus = (1 + ((bestturns/turns) * 0.15)) * 0.55 * (1-Math.sin((turns/180) - 1.5));


I propose to change this to:

jasonbonus = 0.12 * (bestturns/turns) + 0.55 * (1-Math.sin((turns/180) - 1.5));


The reason behind this is that the later victory conditions currently get a much lower Jason early victory bonus for scoring at the Jason best date than the earlier.
This is so, because the early victory part is multiplied by the milking curve, which gives much more score at earlier dates.
This is especially true if the domination best date is very low (as in all these low difficulty games lately).

As an example GOTM 35:

Domination at 400AD gave an early victory bonus of 1317 Jason.
Culture 20k at 1700 only 555 Jason.

My proposal would give 1200 bonus for both (and for space and diplo also, when at the Jason best date).

This still would not mean that you can get a high score at late dates without some severe dairy farming, but may help to limit the frustration of people doing a good late victory.

 

[Ronald]

My proposal would give 1200 bonus for both (and for space and diplo also, when at the Jason best date).

This still would not mean that you can get a high score at late dates without some severe dairy farming, but may help to limit the frustration of people doing a good late victory.

Hi Klarius,

I do like idea: the bonus is independent of the actual year and it is the same for every victory condition given that it is at the same compared to optimal time.
Ronald

 

[Aeson]

Domination at 400AD gave an early victory bonus of 1317 Jason.
Culture 20k at 1700 only 555 Jason.

My proposal would give 1200 bonus for both (and for space and diplo also, when at the Jason best date).

This would require everyone to finish with the victory condition that had the latest best date to score well.

The reason for the discrepancy in how much the bonus gives is because the territory/population score should be higher the later the date. Most of the bonus is based off the curve's "expected milking score" that is being passed up at any given date. If you finish earlier, you are passing up more territory/population score than if you finish later. By taking this out, there would be no reason to finish before the latest victory best date, because up to that point in the game you would be increasing territory/population, and then still be able to get the same bonus at 1700AD as you could get in 400AD.

 

[alamo]

Here's a graph showing the curves. There's a shift and a consistent gap in the new one.

My problem with the scoring system is how it gets so complicated trying to adjust a flawed original scoring system. Why don't we just create our own model? We know the factors, and we can get enough people to play games to give us the necessary data.

 

[klarius]

This would require everyone to finish with the victory condition that had the latest best date to score well.

To get the best possible score you still have to milk to 2050, because the best milkers can still do better than the milking curve plus the bonus for every victory condition.
If these people want to have a few 100 extra Jason by timing a 20k to 2050, so be it.

But there would also be other possibilities to decay the bonus (maybe again a 540-turn factor like in former times).
I just wanted to keep it simple.

I checked the last few GOTM result pages and a little bit more bonus for 20k and 100k would for sure not have been unbalancing, quite the opposite I would say.

And I'm quite sure that even with the modification, I could still not reach the score I can get by domination/conquest, when I try to milk a 20k or even up to 2050.

The formula as is currently definitely gives a double benefit to very early victories. There is the higher bonus due to the milking curve and it's increasing steeper when below the best date.

 

[akots]

These changes would greatly encourage milking whereas the whole purpose of Jason score is to discourage milking and encourage early finish. Instead, the best dates for 100K or 20K victories can be moved forward a bit to compensate and improve best turns/turns ratio.

 

[klarius]

These changes would greatly encourage milking whereas the whole purpose of Jason score is to discourage milking and encourage early finish. Instead, the best dates for 100K or 20K victories can be moved forward a bit to compensate and improve best turns/turns ratio.

You greatly overestimate the effect of Jason best dates.
Best turns would have to put the best date outside of 2050 to have any noticeable effect for 20k.
Also the effect on milking is really not big, but if everybody is so concerned about it here the newest incarnation of modification:

jasonbonus = 0.12 * ((540/turns-1) / (540/bestturns-1)) + 0.55 * (1-Math.sin((turns/180) - 1.5));

 

[Denniz]

The math here is beyond me so I will have to trust the exports.

My question is this: Does the theoretical milking curve compensate all victory conditions equally? I have been wondering if some of the strategies employed to obtain early conquest/domination victories would be sub-optimal for a long term milk.

The example that comes to mind is that happy people=2 points. Short term, they may not matter but long term they are a bigger percentage of total score. But long term, you have to work harder to get maximum pop and maximum happiness. Quick victories get the assumption that they did a good job. Is that something the math is getting correct or are some positions better than others if they were to milk through 2050?

 

[ainwood]

AFAIK, all victory conditions favour getting to the domination limit ASAP.

 

[bed_head7]

Yes, but it is easier to get fastest finish and an award with domination because there is only one thing that needs to be done, and that is reach the domination limit and then pass it. Conquest is similar, because often along the path lots of territory will come into the players hand as a result of conquest. But with every other victory type, there must be a balance struck between reaching domination and going for fastest finish. Meaning that winning an award and medaling in the same game is much easier in conquest victories and domination especially, because achieving a large population and territory is the main part of those victory types.

Culture, diplo and space all could win a medal - if someone milked them after reaching domination limit - but still probably wouldn't win the award. The only possible exception I see is a really well played 100k game, but even in COTM4 SirPleb's fell short in score of the top domination games. The extreme opposite is 20k games, where to get the best date the needs of the whole rest of the empire have to be ignored at time. Like not researching sanitation until the Modern Ages because the 20k city desperately needs a wonder to build.

 

[Denniz]

AFAIK, all victory conditions favour getting to the domination limit ASAP.

That is assumed in my question. The question is about maximizing pop growth and happiness after that as that is the only way grow per turn score after the dom limit. The theoretical vs. the potential of the position. How does it account for variations in pop and happiness improvements at the point of victory?

I assume that territory points are very linery, but pop takes longer to peak and maximum happiness requires improvements not necessary for earlier victory types.

 

[solenoozerec]

I think there is a flaw in the game itself. If you play at the level which is hard for you, you cannot reach domination or kill everybody, so you go for other types like Diplo or SS.
If you play on easy level you can win by any type and the fastest and easiest types are, of course, domination and conquest. To get to other types, something extra is needed. While you do this extra you are loosing your final score.
I do not know under what situation you should naturally go for cultural victories. The only reason to use them is a desire to use them.

What I am trying to say is that I think that victory types are not balanced in the game itself and it is very difficult to compensate for this unbalance with a scoring system.
For this reason I like the idea of awards, because they stimulate people to play not for score, but for something else. This is a very nice feature of GOTM.
If there were many more players, then it probably would be reasonable to make separate competitions for every victory type.

 

[Aeson]

To get the best possible score you still have to milk to 2050, because the best milkers can still do better than the milking curve plus the bonus for every victory condition.

This isn't true. On some maps milking may be "the best way". On others it will be a very early finish. On others a mid-game finish. It depends on where the estimate is furthest behind what is actually possible to achieve. That can happen anywhere on the curve and is map dependant.

 

[Shigella]

I weighed in on this issue after posting my GOTM35 final spoiler, so I'll just post the link HERE

I previously believed that I had posted a reasonably decent spaceship launch date for GOTM35. Now that the spoiler best dates for GOTM35 are available, however, I see that I missed the best date by over 40 turns! In reviewing the best dates for recent GOTMs, it appears that the best dates for space and diplomatic victories are accomplished rarely. Conversely, the best dates for domination and conquest victories are routinely trashed by the best players.

I recall Aeson indicating in another thread that Jason scores are relatively insensitive to the "best date," and I am convinced this is true.

I also remain convinced that one can get a high Jason score with any victory condition. The most obvious example to prove that any victory condition can earn a high score would be to push to domination early, milk out to 2050, and then take any victory condition on the last turn. Under this scenario, all 7 victory conditions score the same number of points (but you would only be competitive for the cow).

AFAIK, all victory conditions favour getting to the domination limit ASAP.

I agree with this except for diplomatic and space victories. Early victories in these games require a more substantial and early investment in infrastructure (rather than focusing most of your efforts on military), and you "should" (make that "might" in some cases) benefit from working with other civs to conduct research. I think this is especially true for the higher difficulty levels. On the lower levels (regent and below) I agree that you might as well whack your opponents early, as they just aren't going to contribute to the tech pace after the ancient age.

 

[akots]

Actually, the Jason scoring system does its best to equalize various victory conditions because the game scoring system does not compensate for that at all. To get a gold medal with 20K, you apparently have to expand to domination limit and then milk until 20K date is reached. It has been possible in PTW but not possible in C3C. It is just a feature of C3C (MGL cannot rush wonders). Similar is 100K.

Actually, I personally find these 20K and 100K victories boring and not interesting. They stand on the same step for me as milking. It is very tedious task (mostly accumulating cash in 100K) and I don't see why it should be well rewarded especially 20K.

We then might end up with something SGOTM-like with another coefficient for 5CC and the third for AW game. Why cannot a player win a medal if he's been playing AW? Because he scores lower than non-AW. Same reasoning is applicable for cultural victories. they just score lower and GOTM is played for score.

 

[akots]

... In reviewing the best dates for recent GOTMs, it appears that the best dates for space and diplomatic victories are accomplished rarely. Conversely, the best dates for domination and conquest victories are routinely trashed by the best players. ...

This leaves two possibilities open:

1) Best player play mostly for Domination because they find other victory types boring or don't have enough time of don't think they can get a very good score.

2) The best dates are wrong.

IMO, number 1) is more close to the actual state of the art but this is a personal opinion.

 

[ainwood]

I agree with this except for diplomatic and space victories. Early victories in these games require a more substantial and early investment in infrastructure (rather than focusing most of your efforts on military), and you "should" (make that "might" in some cases) benefit from working with other civs to conduct research. I think this is especially true for the higher difficulty levels. On the lower levels (regent and below) I agree that you might as well whack your opponents early, as they just aren't going to contribute to the tech pace after the ancient age.

Sorry - I meant that the score favour getting to the domination limit ASAP. Of course, to actually achieve the victory condition requires the investment in infrastructure, so you need to strike the balance.

 

[solenoozerec]

Sorry - I meant that the score favour getting to the domination limit ASAP. Of course, to actually achieve the victory condition requires the investment in infrastructure, so you need to strike the balance.

Therefore it brings me to a conclusions:
1. Getting the fastest SS victory is an art.
2. Getting the highest score using SS victory instead of domination is stupid.
Am i wrong?

 

[ainwood]

Firstly, Aeson is the architect of this system, and I am certainly no expert, so hopefully he can correct any errors or simplifications I have made.

The point I was trying to make (and I should have been more explicit about) is that the Jason score is purely a substitute for the Firaxis end-game bonus for an early finish. The Firaxis end-game bonus us arbitrarily based on the difficulty and the number of years (Years, NOT turns!) between the finish date and 2050. It decreases to zero at 2050 AD.

When the first GOTMs were beng played, it was quickly recognised that there were only two really viable ways to get the highest scores - win the game very early (conquest or domination), or "milk" the game to 2050 AD, noting that the scoring increase you could get by milking from circa 1600 - 1700 through to 2050 AD would normally out-weigh the firaxis end-game bonus.

As a result, the Jason system was conceived. The aim was to provide a end-game scoring bonus for an early finish that was equal regardless of the chosen victory condition. It does this by trying to estimate the score that you could have achieved by milking the game to 2050 AD, without the need to do it. It looks at the amount of food and tiles available on the map, and provides a "curve" that approximates the score that someone could achieve in a well-milked game.

My remarks about getting to the domination limit ASAP were in reference to the fact that the Jason system is an early-finish bonus replacement. Your base score is still important, and this is definitely improved by getting to the dom limit quickly, although improving your tiles and getting the food / population up is also important.

 

[Aeson]

2. Getting the highest score using SS victory instead of domination is stupid.
Am i wrong?

You definitely can get a high score with a SS game. This was a thread dedicated to replaying SirPleb's GOTM 19 start which he ended with Conquest, and instead get a SS victory. SirPleb's actual submission was a 390AD Conquest that won the Gold medal and Conquest award with 11787. My SS from his 30AD save was a 1265AD launch that scored 12229 that would have won the Gold medal and Spaceship award. It would have been the second highest Jason score in any game to ADel's GOTM26 12290... and it could have been highest if the details at the end had been looked after.

Many of the Domination/Conquest victories you see winning medals could have a higher Jason score by playing out to Diplo/Spaceship with only small modification to the end of the Domination/Conquest phase, and in most cases have a shot to compete for the fastest award. If someone would just start doing it consistantly and winning all the Gold medals we could have the opposite discussion...

Cultural 100k and 20k games would take a different approach from rather early on, but still can compete for top scores. Moonsinger won a Gold medal and 20k award in GOTM18. Plus you can even have a shot at the Cow award some months by going this route.

In general though, the difference between a Gold medal game and a high scoring but no medal game for a top player will be "random" events like a Settler from a hut, a proper guess as to the best approach to the map, ect. Not the victory condition itself.