RSW

Congratulations to the winner--and to all the players--of barker5a.

Now that the game is over, the full game report is publicly available, and it looks to me like I can see the remains of a fierce battle at W25. Wow! Do I want to hear the story?

David

It was mostly just a lot of pounding on each other, turn after turn, and then a miffed attempt by Zaaz and FishSpeaker to collect the unowned fleet keys. If I had it to do over again, I would have sent as much force as possible as quickly as possible, instead of just sending as much as I thought would be necessary.

What's algorithm is used to pick the target score? The manual says that scores generally fall between 6,000 and 10,000, so I was a little surprised to see the EndOfGameScore weight of 0, and the target score of 6,400.

As the comment says. This gives a nice bell-shaped distribution curve, with the center around 7500. An EndOfGameScore weight != 0 will skew the curve left or right of center (for instance, by pulling two and taking the minimum, at each pull), but even with weight == 0, there is a chance for a score considerably higher or lower than average.

The weight is a little misleading, then, when it influences only one parameter, like the end-of-game score. It means more when it influences something that has many instances, like mines or industry.

David

I thought I had a chance of doing well at W25, but eventually I realized that I was just trading fleet keys and ships for time, hoping that the game would end soon before my economy fully collapsed.

Since I was fighting against 2 others, I hoped that one would leave a turn after I retreated to allow the other to collect the empty fleet keys. So I planned to come right back that next turn to catch whoever remained...and delay someone claiming those fleet keys even further.

Wow, those Merchants sure can rake in the points.

Are the weights selected using a bell curve as well? Do I understand it right that each point of EndOfGameScore weight increases the number of samples per pull? Like this:
For each of the five pulls:
if weight == -3: sample 4, use min
if weight == -2: sample 3, use min
if weight == -1: sample 2, use min
if weight == 0: sample 1
if weight == 1: sample 2, use max
if weight == 2: sample 3, use max
if weight == 3: sample 4, use max

Yes, that's the general algorithm. The weights are chosen with an explicit bell-shaped curve; it looks like this: Each of the galaxy properties has its own draw list to pull from. For instance, to determine the number of mines for a particular newly-created world, we apply the above algorithm once, pulling from: I derived this and other lists from a statistical analysis of some sample FB turnsheets I had.

David

I just changed this algorithm to resolve the confusion: for EndOfGameScore, which is a singleton, it now pulls a sample from an unmodified bell curve, and then compares that result to the bell curve's standard deviation to compute the reported weight at the end-of-game. Kind of the tail wagging the dog in this one case, but it will be more consistent with expectation, and more sensible mathematically too.

The other properties, which are used repeatedly to draw a sample for each world of the galaxy, are still handled in the original way, where the weight is computed first and influences the shape of the bell curve.

David

I would expect scores in general be closer to 7500 now.