A proposal for a slight modification of Jason scoring

[AlanH]

I think it might be it might be useful to show difficulty and do the averages by difficulty. It also might be interesting to add a column to each VC with the number of players.

The table is quite wide already. How does it fit on your browser?

The counts by VC for each game are already available in graphical form on the Statistics page, and the difficulties are below the X-axis.

 

[civ_steve]

Very cool data!! Thanks, AlanH!

I had Space and Diplo as being worse off (good), 100k and 20k as being slightly off (diverging), and Conquest / Domination being best off (a little switched with regard to 20K and Conquest.

Diplo and Space are clearly too aggressive. In fact, the best Space performance against Jason (-28 turns), Bremp massively abused the Rank Corruption Bug.

100K is probably not bad; a good effort can be somewhat predicted by land type available.

20K estimate is questionable for Classic GOTM, since you depend on Leader Farming for the later Wonders.

Domination is somewhat predictable; Conquest a bit less. Take the poor performance in Gomt30 . China occupied a one-tile island with a Luxury on it, and would not let it go for nothing!! Conquest depended on 1.) occupying said one-tile island first, or 2.) researching Amphib Warfare. This type of on-off condition totally skews the results vs Jason Best date.

I was surprised at the considerable variability in the fastest finish vs Jason Best date. IMO, this is more reason to compare to fastest finish only; Best date is too open to discussion and argument and there will always be someone saying the date was determined improperly. There is no arguing over who submits the fastest of a given victory type! In the Olympics, if the top 3 runners in a race trip over each other, letting the current 4th place person win, that person still gets a gold medal, silver for #2 and bronze for #3.

 

[Denniz]

The table is quite wide already. How does it fit on your browser?

I have a very wide laptop so it fits great. (That's why some of my images have had to be reduced in the past. ) I am not sure how it would look with 6 more columns, though. Maybe a small font size.

The counts by VC for each game are already available in graphical form on the Statistics page, and the difficulties are below the X-axis.

A little copy/paste and ... (I wish I was doing this with regular excel instead of MS Works!)


I don't know what the count of players are but they might provide a better average. I have attached a CSV file that shows my work.

 

[akots]

Can AlanH please actually explain how the data of games played for Jasons score can be adjusted to fit into another system? I'm just puzzled at the outcome. It is not that it could have been impossible to achieve a certain victory condition earlier but by monitoring the best dates, the results of the best players are taken. And at the time the players played these games, they were struggling to keep the balance of getting the fastest finish and the best scores as well, all at once. They did not know that the scoring system is supposed to be modified according to that. Not that there is something wrong with the changes, it is just seems a peculiar way of getting "some data". But the data are not suited for the purpose for which they are being used IMO.

This actually explains the surprising variability of the data noted by civ_steve.

 

[solenoozerec]

In the Olympics, if the top 3 runners in a race trip over each other, letting the current 4th place person win, that person still gets a gold medal, silver for #2 and bronze for #3.

Exactly. GOTM is about a competition, not about a skill measurement.

I don't know what the count of players are but they might provide a better average. I have attached a CSV file that shows my work.

Except for conquest it looks like best dates are more frequently beaten at higher levels.

Can AlanH please actually explain how the data of games played for Jasons score can be adjusted to fit into another system? I'm just puzzled at the outcome. It is not that it could have been impossible to achieve a certain victory condition earlier but by monitoring the best dates, the results of the best players are taken. And at the time the players played these games, they were struggling to keep the balance of getting the fastest finish and the best scores as well, all at once. They did not know that the scoring system is supposed to be modified according to that. Not that there is something wrong with the changes, it is just seems a peculiar way of getting "some data". But the data are not suited for the purpose for which they are being used IMO.

This actually explains the surprising variability of the data noted by civ_steve.

Best dates were created for Jason score calculation.
What AlanH did is that he compared predicted dates to actual dates of the fastest victories or awards. People who goes for awards rarely care about score. This kind of comparison makes sense even without any new scoring system.

But I think that if this new system will be implemented in addition to the existing one, it should be considered as unofficial for those games played before. Because indeed many players played for different score and were unaware of what is coming.

 

[akots]

... People who goes for awards rarely care about score. ...

I beg to disagree with this. And if at least the Noble Eptathlon winners including SirPleb, Ronald, and DaveMcW will indeed confess that this is true, so that they completely disregarded score while going for awards, then I will certainly change my opinion on the matters.

 

[A'AbarachAmadan]

I'm relatively new to CivFanatics. At first I hated the scoring system. I didn't think it favored quick victory enough. Then I realized you give up alot to go for a 20K cultural victory which isn't factored in - still didn't like it. Then realized that there are awards for fastest victory of a type, which is cool as most players seem to go for domination and I might be able to get an award by trying something else.

Final analysis: Jason scoring combined with special awards is the superspearman!! Great job coming up with the current system.

Don't change nothin'! (Yes, I realize this is a double negative and could mean the opposite, but think colloquial, not logical.)

 

[civ_steve]

(edit: to avoid some possible confusion, when I mention decay rate in this post, I'm talking about score within a given game from highest score to lowest score; not about the decay rate for scores from older games listed on the Rankings Tables.)

I think the original formula used was:

x=100x(540 - player_turns)/(540 - fastest_turns)

This formula decays to 0 at turn 540 (2050 AD), but the decay rate varied depending on the fastest_turns for each VC; e.g. assume a 350 AD Domination win vs a 1300 AD Space Race (turns 150 and 250, if I counted correctly ), players submitting 50 turns after each of these dates would score 87% and 82% respecively. I expressed some concern that the slower decay rate would draw people to the early Victory Types, reducing the potential # of players attempting later Victory Type games.

Current Formula from AlanH's webpage is

x=100x(540 - (player_turns - fastest_turns)/540

This enforces a constant decay rate without regard for VC or fastest_turns date; however, the minimum score for someone submitting a 2050 AD game would be 28% and 46% respectively for the two cases above. This seems somewhat unsuitable; the accomplishment of the fastest submittal for each VC is diluted, and to guarantee close to a 50% score for just submitting a poor later victory performance will reduce the utility of such a table.

To accomplish both a constant decay rate regardless of VC, reward Fastest Finishes, provide some measure of performance for slower submittals and put in a floor for lowest scoring winning game, I've come up with this awkward two line formula:

x=5+(45*(540 - player_turns)/(440 - fastest_turns))
if ((player_turns - fastest_turns) < 100) x=100-((player_turns - fastest_turns)/2)

The 2nd line decays scores evenly, regardless of VC, from 100% to 50% over the 100 turns after fastest_turns (in our earlier example, 350 AD Domination win is 100%, so a 1300 AD Domination win would be 50%; and 1300 AD Space Race win would be 100%, so a 1770 AD Space Race win would be 50%). Wins submitted after fastest_turns+100 would score by the 1st line; this line puts a floor of 5% in for lowest possible score, and decays evenly from 50% at fastest_turns+100 to 5% at 2050AD. This reinstates the problem with differing decay rates depending on how early the fastest_turns is for each VC, but the diffence is reduced (wins submitted at fastest_turns+150 for the two examples would score 42% and 38% respectively) and pushed out into the lower 50% of possible score.

For a fastest_finish within 100 turns of 2050, the subsequent scores (if any) would be artificially high. I have hope that implementation of a rankings system focused on fastest_finishes for all VCs would prevent that from happening (by increasing participation in all VCs), but its really no worse than giving an award to a 2050 AD milked game that decided to finish with Conquest rather than Histographic (which I think happened at least once.)

The faster decay rate over the first 100 turns after fastest_turns should help accentuate the performance of those who submit at or close to the fastest_turns for a given game. Obviously any of the parameters can be tweaked, and by all means, I would be happy to see a less awkward formula that accomplishes similar scoring method, if someone comes up with one.

 

[mad-bax]

I'm no mathematician but if the object is to get a reasonably constand decay rate but provide more score separation at the top end of the scores, then a reasonable simple formula might take the form

100/(1+X((PT - FT)/540))

X can be any constant or formula which plays with the decay rate, but if it's just set to 1 for example, then 1 turn difference gives a score difference of 0.2%, 100 turn difference gives a score difference of 15.625% (0.156%/turn average), and 200 turn difference gives a score difference of 27% (0.135%/turn average).

If X is bigger (5 say) then 1 turn difference gives a score difference of 1% and 200 turns gives a score difference of 65% (0.33%/turn average).

Just tuppence worth of hot air.

 

[AlanH]

Can AlanH please actually explain how the data of games played for Jasons score can be adjusted to fit into another system?

I'm not suggesting they can. All I'm doing is to try to show the dates achieved currently, and how they relate to Jason dates. I'm providing statistical reports, I don't mind what you use them for. You are free to ignore them if you think they are meaningless.

 

[AlanH]

But I think that if this new system will be implemented in addition to the existing one, it should be considered as unofficial for those games played before. Because indeed many players played for different score and were unaware of what is coming.

The "official" recognition for excellent games is and will continue to be delivered in the form of Jason score-based medals and fastest finish awards.

The other stuff - rankings, statistics etc are just a different view of existing data. Anyone could have created the page I've produced using the published information. As I have access to that information in a form that makes it easy for me to manipulate I've done so. "Official" and "unoffical" don't really come into it. Either it's a useful and/or interesting view of the game results we have now or will have in future, or it isn't. If it's neither then we'll improve or discard it.

beg to disagree with this. And if at least the Noble Eptathlon winners including SirPleb, Ronald, and DaveMcW will indeed confess that this is true, so that they completely disregarded score while going for awards, then I will certainly change my opinion on the matters.

If we are saying that the fastest finish awards have so far been won as a side-effect of playing for score then it should follow that there are easy pickings to be made by anyone who targets those awards. Why aren't players doing so?

 

[akots]

... If we are saying that the fastest finish awards have so far been won as a side-effect of playing for score then it should follow that there are easy pickings to be made by anyone who targets those awards. Why aren't players doing so?

It is not what was implied when the issue was discussed. Playing for fastest finish does mean that there is a compromise between getting a decent score and at the same time accomplishing the fastest finish date without sacrificing everything including happiness/expansion to get that fastest finish date. IMO this is very clear.

As an extreme example the GOTM 30 played by Kuningas comes to my mind, where he revolted every 20 turns or so to Despotism and pop-rushed culture in corrupt cities. Also in the very same game there has been a very dramatic competition between MarCutt and Gozpel for the cow award, where the difference was only 103 points. It is indeed a photofinish.

Also, the only one GOTM where I went for Space victory (GOTM 28) ended up in a very tight space race result, where there were 5 player launching within 3 turns from the fastest finish. However, their Jason's score was completely different which basically shows that a player which sacrifices a lot, can get that fastest finish, whereas if still expansion and happiness are not neglected substantially, it is hard to compete for the date. It could have been not the best game I played, but it was a rather frustrating outcome.

In general, it is always better to compete with 100 players and have place number 20 of 100 than compete with unknown number of players (assume 5) and get 5th place out of five (which might as well let you into the 20 top if it were 100.

 

[AlanH]

In general, it is always better to compete with 100 players and have place number 20 of 100 than compete with unknown number of players (assume 5) and get 5th place out of five (which might as well let you into the 20 top if it were 100.

I think I may be misunderstanding you. If the five who play for 100K are playing also for score then I am suggesting that, you, or other good players, have a good chance of beating them and winning an award. You should be able to beat their date if you play a no-compromise 100K game for the fastest date and don't try to maximise your score. Why don't people take that approach? If the Global Rankings such a strong motivation, then with two games to play in a month you could play one for speed and one for score to win speed awards and still maintain your GR position.

 

[solenoozerec]

I think I may be misunderstanding you. If the five who play for 100K are playing also for score then I am suggesting that, you, or other good players, have a good chance of beating them and winning an award. You should be able to beat their date if you play a no-compromise 100K game for the fastest date and don't try to maximise your score. Why don't people take that approach? If the Global Rankings such a strong motivation, then with two games to play in a month you could play one for speed and one for score to win speed awards and still maintain your GR position.

September 22nd:

One reason is that in COTM I play for Awards and in GOTM for Medals.