Posted: Sun Feb 10, 2019 4:58 am Post subject: Results by Seat and Selected Nation
Hi all, a while back I solicited your game results here on Mabiweb and at Boardgamegeek.com to build a database for analysis. Recently we broached 100 results for Dynasties games.
I've started looking at the data. The results live here, if you want to see it yourself. The column headers inherit the organization of the google form that produce the row-entries, so it may be cantankerous to use if you haven't seen the form.
Before looking at this, realize three things. 1) We have no idea of the quality of the data. Who knows who submitted most of these game results? Even I stopped contributing results after a couple months. Two thirds of the games are reported to be from Mabiweb, but I have audited 0 of the results to ensure they're accurate. A third of the games are from "real life plays" so might be the results of a very skilled or very bad group of players.
2) This is all games (2ers, 3ers, 4ers, 5ers, and 6ers) lumped together. The strength of nations varies (I predict lightly) by table size. As well, I expect initial seat order has more impact in bigger games than smaller, because the book awards are greater and the relative power of choosing your nation before other players grows the later in the turn order you start. None of those effects are visible here, but surely at play.
3) These are still small sample sizes, and you're prone to observing the evidence that confirms your expectation without heeding its huge limitations. I recently wrote a post about how Japan is one of the strongest nations and, lo!, it has the most wins. However, it also has the most plays. How many wins should it have? I dunno.
If nations are randomly selected to be present in a game as on Mabiweb, with a large sample size we expect the yellow bars to all be the same height. If the nations are perfectly balanced, we expect the blue bars to match the yellow bars. However, there are some standout variations like Korea. Venice, Greece, and Mali all win lower than random chance might expect. Perhaps they fair well overall, but we can't see that yet since this is only 1st place finishes. Stay tuned for some more analysis by table size.
Last edited by LordChambers on Thu Oct 17, 2019 5:00 pm; edited 1 time in total
Interesting. I actually like Greece for a nation, the starting power is pretty decent to start with; even though I like going Sparta as soon as I can. Venice I can see being so low, they don't start out with a good set of buildings and the base power is terrible. Mali is another one I can see why it might be lower. The power to kill off advisors for points and books seems good but is very gold and action intensive so is harder to pull off than it would appear (plus there have to be advisors to buy and then kill).
Posted: Thu Oct 17, 2019 4:58 pm Post subject: Re: Results by Seat and Selected Nation
Stay tuned for some more analysis by table size.
You can safely tune out at this point. I don't have as much curiosity as I used to, and the data set probably should verify the MabiWeb results and exclude the real-life plays, since right now it's like comparing apples to dragonfruit.
Eh, I'm a liar. I discarded the real-life plays and base game results. Most of the results left were for 4ers. Here are the results.
I can't post images without getting a SQL error, so to see the graph you'll have to click here
Out of 28 games played on MaBiWeb with 4 players; where a first place finish earns the Nation 6 points, a second place finish earns 3 points, a third place finish earns 1 point, and a fourth place finish earns 0; Portugal has the highest average score of 4.2. America has the lowest average score, with 0.8.
These sample sizes are very low. Despite random selection, some nations are relatively well represented in the dataset (n=28), such as China in 13 games. Most nations are represented by only 6.5 games.
No major surprises to me. Every nation appears pretty solid, including Korea. There might be a slight bias for military-advantaged nations, but this is too little data to draw a conclusion.