Recommend
6 
 Thumb up
 Hide
4 Posts

Leaving Earth» Forums » Variants

Subject: solo variant with an AI rss

Your Tags: Add tags
Popular Tags: [View All]
Rus
United States
Mountain View
California
flag msg tools
Avatar
mbmbmbmbmb
I think the official solo variant is great, but there is one thing that multi-player does better: the race to complete a mission before your opponent does. It seems to me that the optimal solo strategy is to play very conservatively -- spend as much time as possible testing technologies and time things so that you have guaranteed-to-succeed missions on the final year. This makes it into a great puzzle, but is missing the excitement and tension that comes from racing and trying to outguess your opponent.

With that in mind, I propose the following variant which simulates "automa-like" AIs (see https://boardgamegeek.com/blogpost/37830/automa-approach-i-m... for what I mean by "automa")

1. Set up just as you would for a multiplayer game with any desired number of players, except some players can be AIs. Organize missions in a line and AIs in a column, like so:


| mission 1 | mission 2 | mission 3 | mission 4 | ...
----------------------------------------------------------
AI 1 | | | | | ...
AI 2 | | | | | ...
... | | | | | ...


The basic idea is that each location in this matrix will contain money that each AI spends towards progress on that mission, and can score that mission when a completion criterion is triggered. AIs are treated as normal players in all respects except as noted below.

2. AI turn: if an AI has money when their turn starts, that AI will assign all their money towards a random mission (out of the ones still remaining and possible) and end their turn. Roll a die or dice to randomly select a mission and place that AI's money in the appropriate location in the above matrix. If an AI has no money at start of their turn, they pass.

3. AI mission scoring: an AI scores a mission if at any time

$AI / N >= mission_VP,

where $AI is the total money assigned by an AI towards a mission, N represents the AI difficulty (essentially, the average millions of $ that AI spends per VP), and mission_VP is the victory point goal for that mission. A mission is scored immediately unless it is supposed to be scored at start of year, in which case it is scored at start of year per normal rules. Any other missions that are strictly a subset of the mission being scored are scored as well (e.g. sounding rocket is also scored whenever artificial satellite is scored). $10M per completed mission is awarded normally to all other players, including AIs.

The parameter N is something that needs to be tuned through playtesting, but I am guessing N ~ 10 (i.e. $10M / VP) would be a good place to start for "normal-difficulty" AI. Increase N to make the AI easier and decrease to make it more difficult. Joe, if you have average values of VP per player per year from your playtests at different difficulty settings, that would be enormously useful in tuning N.

Alternative to #2: instead of assigning all $25M (or more) that the AI has at start of their turn to a single random mission, assign each banknote separately to possibly different random missions. In this case, AIs always get their money in denominations of $10 and $5, so that an AI with $25M at start of their turn will assign their money to up to 3 missions. This should reduce the variance in the game and make the AIs more predictable.

If you find this variant interesting, please post play reports here, and how you think N should be adjusted!
5 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Colby Brown
United States
Katy
Texas
flag msg tools
Avatar
mbmbmbmbmb
I just tried this variant today. Works pretty well!

I played the alternative where each nation divides up their bills each year; so they put $10, $10, and $5 to a mission at the start of a year, plus another $10 if someone else completes a mission. I will have to play the other way to know for sure, but I did not feel like this made the AI's "more predictable" in any way. Their money was very spread out, and they took a long time to finish even the low-VP cards. Having each AI put all $25 million towards one mission each year might help this, but I don't think it'll make that big of a difference (since they'll be contributing to less missions each year, it'll all average out in the end).

It did add a bit of a "race" feel to the solo game, because Russia did seem to like putting money towards the mission that I really wanted. However, the AI's put money towards missions sporadically, so they took a long time to build up the right amount of money. If the AI's were changed to prioritize missions (possibly randomly) and "focus" their efforts, they would be much more difficult... and add a lot more of that "race" feel you were going for. I found it was easy to beat the AI's to the low point missions (the 5 - 10 points, not the 1 - 4; they could earn those in one or two rounds), and the high point cards took too long for them to build up money for.

One element that I really enjoyed was that there was a meaningful choice to make when surveying / landing for revealing the location or not. Looking at a card and then leaving it "unrevealed" allowed the AI's to waste time dumping more money into missions that were impossible or that I was sure I could still beat them to, but also made it possible for them to steal the mission from underneath me. It was also nice that China finished the "Moon Survey" mission before me and revealed that location as a dusty wasteland before I sent Buzz Aldrin on an impossible and deadly mission.

Hopefully I'll have time to play again soon and maybe tweak the rules a little bit more! Also looking forward to seeing if anyone else gives this variant a try.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rus
United States
Mountain View
California
flag msg tools
Avatar
mbmbmbmbmb
Awesome, thanks for the report! If you found that it was easy to beat the AI, the easiest thing to do is to reduce N. Maybe N = 5?

Your report also suggests the following possible rules tweaks:

1. Forget the idea of randomly assigning individual bills, it sounds like it's better to just put all of AI's money towards one mission per turn.

In order to have the AI focus on missions, perhaps one of the following would work:

2a. When selecting missions, roll the D8. If 1-4, then the money is assigned to the AI's current highest funded mission. Otherwise, money is assigned randomly. The 1-4 may need to be tweaked.

2b. The AI can have at most 4 missions on which it is working. This can be implemented in either of 2 ways:
2b,i. The AI random mission selection is always limited to the bottom 4 missions (this also makes it easier to randomize with the D8)
2b,ii. If the AI has < 4 missions funded, assign AI's money to missions randomly. If it has 4 missions funded, assign randomly between those 4 missions.

The fact that the hard missions are easier to race than lower point missions is more serious, and suggests that the formula

$AI / N >= mission_VP

needs to be changed. One possibility is:

3. Assign a $ value for each mission to be met, i.e. $AI >= mission_$, where mission_$ is probably something like sqrt(mission_VP)*10.

Alternatively, something like:

4. Instead of assigning money, place that AI's spacecraft token on the mission. Easy missions require 1 token to complete, medium missions require 2, and hard missions require 3 (might need tuning and depends on AI difficulty). This needs to be combined with 2b above since there are only 4 tokens. If a mission already has a spacecraft token, place a time token on top of the spacecraft token.

I actually think #4 is my favorite because it saves table space -- you no longer need the matrix of missions and AIs, you just place spacecraft tokens right on the missions. It also avoids running out of money pieces.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Gonzalo Herreros
Ireland
Dublin
flag msg tools
badge
Avatar
mbmbmbmbmb
Haven't tried it but here are some comments:
As you have noticed, dividing the money randomly makes the AI weak and it's also not very thematic.
To make the AI competitive with multiple missions, the N value would have to be low, which might result in the AI having very different performance on each game. When that happens is hard to know if you are making improvements in your own game.

Another concern I have there is that the player could always just focus on thwarting the AI medium/long missions.
In a multiplayer game you don't have that precise knowledge of what the other player plan and progress until they launch.
Having the AI indirectly choosing the missions for me is a bit annoying, that's why I think a mechanism to discourage that a bit would help. I came up with the idea of the AI moving progress in most cases either partially or full.

The other approach you could take is to just accumulate money and each turn try to use it to complete any mission. Maybe add an outcome card to complete so the AI is not too powerful





1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.