GOTM 31 Results and Congratulations

[grahamiam]

great job to the winners and the staff for getting results out so quickly no wonder they are ready for COTM & GOTM!

 

[ainwood]

The person most responsible for getting the results out so early is AlanH. He has spent a lot of time and effort to automate the system. Others on the staff helped him, most notably ainwood and space oddity, but the bulk of the work was done by AlanH.

Many rounds of applause are in order

Actually I didn't need to do anything this month. This automation makes life a breeze!

 

[alamo]

Here's another graph - kind of complicated, but it has some interesting features.

The curve probably reflects the baseline score used in the Jason calculation, while the points above the curve are the exceptional players. People below the curve have the distinction of being like me!

The milkers stand apart at the 2500AD line, but only DaveMcW manages to do better than the 570AD domination (impressive, Dynamic!).

The time played does seem to be an upper bound on actual playing time, because of people leaving the game on idle. The large circles do seem to fall on curve for the most part.

The cultural victories tend to be above the curve. I guess that is the bonus for concentrating on building.

I'm not sure what to make of the trailing conquest victories at the bottom. I guess those despots should have built an improvement or two.

 

[AlanH]

I'm not sure what to make of the trailing conquest victories at the bottom. I guess those despots should have built an improvement or two.

Conquest defeats?

I think the time played should be expressed as time per turn, not elapsed time, as that would eliminate the effect of the sheer number of turns needed to reach 2050 AD.

 

[Dianthus]

Hey, cool graph Alamo! I'm not sure the time played is telling us all that much, but the curve is interesting. I can see my conquest win at 1270AD. I've yet to decide if I'm close to average or close to exceptional .

 

[alamo]

Good point - there were several conquest and ss defeats. I guess I should have indicated the losses.

Time per turn would be a good thing to have. Anyone got a formula for date-to-turns conversion?

@Dianthus: Looks above average to me - well done!

 

[AlanH]

here

 

[Offa]

Alamo

That graph is very cool.

 

[SirPleb]

Very cool graph Alamo!

There's some crude but correct javascript which does date to turns and v.v. in the attachment to the first post in this thread

 

[a space oddity]

Great graph, Alamo, very interesting! I've been thinking about ways to display this kind of info and this one the job wonderfully.

I also like the graph, since being owner one of the four 100k blobs it's easy to identify my game. And it's not so bad after all.

 

[Aeson]

I like that graph Alamo.

My interpretation of it:

The curve probably reflects the baseline score used in the Jason calculation, while the points above the curve are the exceptional players. People below the curve have the distinction of being like me!

When viewing the curve(s), it's important to look at the same victory condition. The 'best' date should be near the top of the curve for each condition, which creates a different curve for each.

Most noticeable is the milking 'curve', as it can only be achieved at 2050AD, so runs straight up that line. Conquest (1150AD), Domination (1000AD), and Diplomatic (950AD) were very close this month, and follow just about the same curve. Spaceship (1300AD) and Cultural100k (1575AD) curves would terminate towards the center of the graph and so the games that look 'above' the main curve (because so many were Conquest, Domination, and Diplomatic) are actually pretty close to being on their respective condition's curve still.

Your graph suggests that the Diplomatic date was a bit too early, and the Conquest/Domination dates a bit too late, which is why the Diplomatic scores tail off a bit more strongly. Though the disassociation between Diplomatic date and how well played the game was is also a factor

The milkers stand apart at the 2500AD line, but only DaveMcW manages to do better than the 570AD domination (impressive, Dynamic!).

The y position on the graph is how 'well played' the game is.

The form of the curves for each victory condition is more indicative of how the scoring system is designed.

Whether an individual game is left or right of the condition curve indicates the factors the game was focused on. If to the left of the condition curve, it denotes focusing on best date to the detriment of 'milking' (population/territory). To the right indicates the reverse. The milking games (which can have their own victory condition) are the extreme of the second case.

 

[Dianthus]

@Aeson, am I right in thinking that ideally the best Jason score for any date and any victory condition should be the same? If so I guess the curve is really due to the fact that finish date and "how well played" aren't independent of each other as everyone (almost, with the exception of the Histographic condition) is trying to finish as soon as possible. I.e. the better players will finish earlier.

 

[a space oddity]

...Cultural100k (1575AD) curves would terminate towards the center of the graph and so the games that look 'above' the main curve (because so many were Conquest, Domination, and Diplomatic) are actually pretty close to being on their respective condition's curve still.

Hey, stop spoiling my party.

Whether an individual game is left or right of the condition curve indicates the factors the game was focused on. If to the left of the condition curve, it denotes focusing on best date to the detriment of 'milking' (population/territory). To the right indicates the reverse. The milking games (which can have their own victory condition) are the extreme of the second case.

I guess I should try milking my games... now the only thing I need is time. Anyone now where I can find it?

It does show (again) how well the Jason scoring system works. We owe you Aeson.

 

[alamo]

Ok, good stuff. Thanks, everyone.

I'll try a per-turn graph as soon as everone leaves me alone. They expect me to do some work for my pay - sheesh!

 

[Aeson]

Whether an individual game is left or right of the condition curve indicates the factors the game was focused on. If to the left of the condition curve, it denotes focusing on best date to the detriment of 'milking' (population/territory). To the right indicates the reverse. The milking games (which can have their own victory condition) are the extreme of the second case.

This can help a player identify why they might not be scoring as well as their skill shown by date suggests. The scoring system is designed to try to keep games to the right of the victory curve about even (it consistantly held up around +/- 5%), but will falter at trying to reward games to the left of the condition curve where population/territory tradeoffs have been made to obtain the victory condition sooner.

This is an intentional design 'flaw' because the diminishing returns of the tradeoff result in an exponential curve that would quickly ruin the scoring system. When dealing with the horizontal lines (which are pretty close to indicative of the 'flat' portion of an exponential curve) to the right of the condition curve, +/- 5% isn't that big of a deal and is in line with what the exponential date bonus curve would look like. To the left of the curve, the lines don't represent how the exponential date bonus curve would look like. That is intentional, so that the same +/- 5% (think of this as horizontal variation) doesn't result in the very large discrepancies that would occur otherwise.

 

[Aeson]

@Aeson, am I right in thinking that ideally the best Jason score for any date and any victory condition should be the same?

To the right of the best date for that victory condition it should be. +/- 5%

If so I guess the curve is really due to the fact that finish date and "how well played" aren't independent of each other as everyone (almost, with the exception of the Histographic condition) is trying to finish as soon as possible. I.e. the better players will finish earlier.

Yes, although there is a linear regression on the date bonus from the best date to 2050 which also will 'slant' the curve (sorta like \ ). As the players skill goes up, their score should also go up, and they will beat (be left of) the scoring system's slant \ more and more while maintaining the same y coordinate (which is basically indicative of population/territory they would have at 2050AD).

 

[alamo]

Here is a graph with turns instead of dates. Obviously the curve was just the timescale, not the fabled Jason curve. Other improvements are size by minutes per turn, with values more than 20min censored.

As a nod to the Jason calculation I added a victory progression line at the top that joins the median victory dates. Players who scored their victory within (or before) that color band did better than 50% of the players.

I'm sure we could make a better graph using the real Jason score properties, but someone will have to help me bring those formulas down from the stratosphere. Someone must be doing some work to generate the appropriate parameters for each GOTM.

 

[smackster]

What is most striking (to me) is how much higher the diplomatic win scores are over the space ship wins. In fact should the Jason score win type dates be adjusted to even the scoring out between victory types. There is a clear score advantage conquest->domination->diplomacy->space, not respecting the other victory types that have too few wins to show a trend, and disregarding certain people who could beat the rest of us with any victory condition.

BTW, alamo, thanks for the charts, fascinating.

 

[AlanH]

@Smackster: Check the Jason calculator. The difference between diplo and space dates is roughly equal to the number of research projects between Fission and Robotics, multiplied by 4 turns each. There is actually a slight discrepency that I raised with Aeson, and plan to adjust as we move forward, but it's only a matter of a few turns, and probably not enough to make the difference you are noting. I'd say space, on average, is attracting fewer of the higher scoring players.

PS. The Jason parameters - best dates and max score - are available in the online Calculator. The formula that uses them is published on the gotm site.

 

[smackster]

Well its tough to get 4 turn research at that point in the game, and you have to build the parts too.

Maybe you are right however, and space is often default for games that don't quite go so well.