GOTM 20 Statistics

What really impressed me of GOTM20 was the number of submissions for a Deity game.

We got 187 valid games, out of which 111 were victories...
And I am pleased to say that your Registrar's game was one of the 111 -his first Deity victory, actually...

I have also updated the additional plots that show us the historical evolution of some of the most interesting parameters. This is actually the fourth GOTM since Jason was introduced, and we are starting to get a feeling of how things evolve with time.

So, just to demonstrate how clever you all are, here is the history of Wins percentage...





Of course, I am not forgetting the lovely pie-chart with the chosen victory condition by us players. Here it is: with GOTM20, many of us (including myself... ) went for a Diplomatic victory, probably perceived as the "safest"...
I am expecting of course GOTM21 to be dominated by Space Race wins, because of the Tournament -which is why I tried to go for a different one...




The score and finish date diagrams.

Still remaining within GOTM20, the two plots showing the score and finish date for each victory condition.







Comments?
I'll have to check this -maybe coming next month with a new plot- but it looks to me that militaristic victories are always few, but most often selected by the skilled players that conquer the top positions.

In this game, additionally, the Cultural victories did not seem to be particularly sought after -more difficult?
My personal opinion is that it may be slightly easier to match the 'best' date with a military type of game as opposed to a cultural one (which of course does not mean that *I* will ever be able to do that... ).

Well, if you think about that, there were good chances for a medal there (provided one can pull out a win, of course...)!

Ok. Let's turn again to the historical perspective, and give a look at the TpS evolution.
I tried to give a look at the time-history of the average, max and min, but it does not reveal much. The plot below, instead, testifies to our becoming collectively more disciplined.
Or not?

More charts (you will be wondering about the different meaning people give to the word 'entertaiment', by now... ).

I am following the experiment I tried last month, in looking at the evolution of the game score and finish date.

Here they are.





Didn't I say that everybody is improving too fast?

Awesome job again Karasu!

I think everyone must be slaving away on GOTM21 for me to be the first poster after 2 days.

I especially like the finish date by victory condition. Next month, could you sort the victory conditions by their best date. You can really see the trend in lower scores by later finish dates caused by later "best" dates. That means that people are not getting to the domination limit while going for their longer wins. I guess domination has become the new GOTM's milking. If you don't do it, you won't get the high score. At least for longer victories.

I wonder if this is an incentive to adjust the Jason curve even further to make the bell curves for each victory have the same average? Not that I'm a math pro. I'm just commenting. Would that be fair? Or would it be better to have the highest scores in the category adjust where the curves should lie? Seems like the population is lower for high scores though so your data is less valid.

Hi Karasu,

Thank's for this really interesting information every month.
Comenting on the most infamous graph, the turns per reload:

I find it quite interesting, that there was no significant change in the last three games (at least if you take out the 80+ turns in gotm19).
I don't count my reloads, but I am almost sure, that I had many more sessions in gotm20, since at deity I tend to check everything at every turn.

Therefore I was expecting a lower number of turns per reloads in gotm 20 compared to the gotm's before.

Maybe the graph is also a bit misleading: Karasu did you make any statistical analysis about averages/means/percentiles etc. (whatever meaningful measure) turns per reload?

best regards,

Ronald

Very nice job Karusu.

Is it just me or is there a dirty great spike on the TPS for GOTM19?

Over 80 turns per session? Some folks must have left their computers on, because I struggle to do 20-30 without falling asleep.

The TpS is not too easy to read. It is clear that there is a tendency to leave the computer on in order to increase this parameter, and that it is especially apparent in GOTM19, for which (I checked) it appears that there were indeed 28 games with >80 TpS.

I was also expecting a lower average on GOTM20 -which didn't happen. That game has the highest st. deviation, however (which may or may not mean something).
In fact, I have been looking at the time history of max, min, avg and deviation. In particular at the last two, as the first ones do not appear to be too meaningful.

The average is very stable around 30, while the deviation is always quite large (27 to 44), so that after only four games it is really hard to see anything.
I'll keep watching, though.


@Control Freak -are you talking of a sort of normalization of the plot by means of the best date?

Originally posted by Karasu
@Control Freak -are you talking of a sort of normalization of the plot by means of the best date?

Click to expand...

No, that would requires some statistical background on my part.

I was simply saying that if you displayed the score versus victory condition graph in the order:
conquest, domination, diplomatic, spaceship,cultural

rather than:
Culture, Conquest, Domination, Space, Diplo

The histographs would have their populated "bumps" roughly in a line. That fact that the line of "bumps" being skewed, rather than straight across one median score indicated to me that the speed based component is not yet tuned to allow any average victory condition to get an average score. Domination and Conquest are still favored. (The line of bumps is skewed with the military victory average scoring higher than the non-military victories.)

This is of course purely graphical and not statistical is indicated by my "bumps" jargon.

Being purely graphical in nature, I was asking you to reorder the graphs so that everyone else would see the trend I was talking about.

If you reordered the finish date graph also you would see the trend to have earlier finishes for military and later for non-military. Hence the reason for the different best-finish dates that Jason uses.

EDIT: I leave it up to you statistically to come up with a way to quantify the trend and then use it to further tune the Jason score. For my Graphical interests, is it possible to put a line bisecting the graph that connects the average or medians of each of the victory conditions?

I think the lack of cultural wins was unsuprising.

I specifically went for a cultural win because of this. I thought I had a chance for the "fastest cultural win" medal ( I got it too ). I also wanted to see if I could get to 2x culture in a deity game without attacking an opponent.

The reasons are two fold:
1) Based on discussions, I got the distinct impression that the top players had realized that going for a cultural win was not a good way to get a top score. You have to wait longer to get it which seems to hurt ones score. I could have probably gotten a better score if I had gone for domination a few hundred years earlier.

2) Culture is hardest on deity because it is hard to build wonders and get over 2x the culture. I actually did not win when I hit 100K culture, but rather had to wait several additional turns before I passed the 2x mark. This was with me being 4x the size of the two civs that I let live and pulling in over 1600 culture a turn.

I wonder if the TpS stat gets inflated/skewed by people rushing to either finish the QSC or submitting the game.

Not that I want to create more work for anyone here, but I wonder how different the TpS is for QSC versus the remainder of the game versus the overall game.

forged

Originally posted by CruddyLeper
Very nice job Karusu.

Is it just me or is there a dirty great spike on the TPS for GOTM19?

Over 80 turns per session? Some folks must have left their computers on, because I struggle to do 20-30 without falling asleep.

Click to expand...

I for one run my computer 24x7 and only quit GOTM if my computer needs a reboot (typically less than one a month for security updates) or Civ3 has slowed down to the extent that a quit/relaunch is needed to pep it up again.

Excellent work, Karasu! I just saw this, but I felt compelled to thank you for the many hours that you spent on this project. Well done!

Thanks, JVJ!

Ehm... now I should get the GOTM21 stats out, shouldn't I?