Dominion Strategy Forum

Dominion => Simulation => Topic started by: dedicateddan on December 05, 2017, 02:47:25 am

Title: Testing Big Money Ultimate
Post by: dedicateddan on December 05, 2017, 02:47:25 am
Bots are often benchmarked against Big Money Ultimate (BMU), and I'm interested in how BMU performs in ladder matches.

In an initial set of 10 games, the strategy went 1.5-8.5. Attacks hurt a lot, but around half of the games were close.

To get reasonable statistics, I'd like to play around 100 games total. If you'd like to pitch in a few games, shoot me a message me on discord.

For reference, the BMU buy rules are:
- PROVINCE
- DUCHY when provinces left <= 5
- ESTATE when provinces left <= 2
- GOLD
- SILVER
Title: Re: Testing Big Money Ultimate
Post by: JW on December 05, 2017, 10:47:46 am
What is a ladder match?
Title: Re: Testing Big Money Ultimate
Post by: Awaclus on December 05, 2017, 11:46:13 am
What is a ladder match?

A rated game.
Title: Re: Testing Big Money Ultimate
Post by: JW on December 05, 2017, 03:12:45 pm
So I take it that you want to play a bunch of rated games against (random) human opponents using BMU to see what rating a bot with this strategy would attain?

Presumably you don't have a subscription for the test account. So given the low expected ranking of the BMU strategy, you'll see mainly base-only games. But if you have a Gold subscription for this account, you'll presumably be playing against lots of players for whom many of the cards are new. So then the results may not reflect those player's skill but rather their unfamiliarity with the cards (the first game I played with Doctor, I tried to use the on-play to trigger Tunnels...). I'd expect that against players who know all of the cards, BMU gets quite a bit weaker with the Gold subscription in part because ignoring events (Donate!), landmarks (Bandit Fort!), and alternate sources of VP (Colony!) often nearly assures defeat.
Title: Re: Testing Big Money Ultimate
Post by: dedicateddan on December 05, 2017, 05:31:52 pm
Yes.

The idea is to see what ranking BMU corresponds to on the shuffle IT leaderboard. The "Big Money Ultimate" account doesn't have a subscription, so it will be mostly base games.

This can serve as an approximate anchor between bot-on-bot matches and the shuffle IT leaderboard. Any new bots that we create will play a number of games against the BMU bot to initialize their expected ranking.
Title: Re: Testing Big Money Ultimate
Post by: 2.71828..... on December 06, 2017, 12:48:15 am
Tell me when you guys are doing this, I volunteer to play against the BMU player.
Title: Re: Testing Big Money Ultimate
Post by: dedicateddan on December 06, 2017, 12:49:59 am
Tell me when you guys are doing this, I volunteer to play against the BMU player.

Markus was suggesting that I test BMU with him ... on my main account :)
Title: Re: Testing Big Money Ultimate
Post by: enfynet on December 06, 2017, 01:31:03 am
You might get accused of sandbagging
Title: Re: Testing Big Money Ultimate
Post by: Chris is me on December 06, 2017, 08:14:24 am
A few things:

1. If you play people who know you are playing BMU from before the opening buy, that’s a significant amount of information that informs their strategy. Granted, a lot of the time the Silver Silver opening just gives that away and adjustments are quickly made, but still.

2. If you seek out opponents from this forum, your sample is going to be influenced by the relative strength of a forum-goer vs other players.

In any case, I’m still interested in seeing this data.

I’m also interested in seeing how a more adaptive money player does on the ladder, using terminal draw or other enablers, but that’s harder to automate and even hard to do manually. (It’s SO easy to just absentmindedly move toward an engine...)
Title: Re: Testing Big Money Ultimate
Post by: Limetime on December 06, 2017, 10:11:44 am
A few things:

1. If you play people who know you are playing BMU from before the opening buy, that’s a significant amount of information that informs their strategy. Granted, a lot of the time the Silver Silver opening just gives that away and adjustments are quickly made, but still.

2. If you seek out opponents from this forum, your sample is going to be influenced by the relative strength of a forum-goer vs other players.
This is countered by the fact that Big Money Ultimate matches against under 45 players the vast majority of the time.
Title: Re: Testing Big Money Ultimate
Post by: dedicateddan on December 06, 2017, 10:53:49 pm
So I think my original post might not be the most clear and the request is a little bit odd.

I'm looking for volunteers to play the Big Money Ultimate strategy in ranked games against random, similarly leveled (~35ish) opposition.

There have been some pretty interesting games so far (21 games in)!
Title: Re: Testing Big Money Ultimate
Post by: Cave-o-sapien on December 07, 2017, 03:01:26 pm
So I think my original post might not be the most clear and the request is a little bit odd.

I'm looking for volunteers to play the Big Money Ultimate strategy in ranked games against random, similarly leveled (~35ish) opposition.

There have been some pretty interesting games so far (21 games in)!

So we PM you, you give us credentials for the BMU account and we play some games with it?
Title: Re: Testing Big Money Ultimate
Post by: dedicateddan on December 07, 2017, 03:26:58 pm
So we PM you, you give us credentials for the BMU account and we play some games with it?

That's exactly how this works :)
Title: Re: Testing Big Money Ultimate
Post by: JW on December 07, 2017, 04:24:31 pm
So we PM you, you give us credentials for the BMU account and we play some games with it?

That's exactly how this works :)

And he'll be reloading the ends of the games too, so no cheating by taking Gold over Duchy with <5 Provinces left just because Palace is on the board!
Title: Re: Testing Big Money Ultimate
Post by: Accatitippi on December 07, 2017, 10:54:20 pm
I had one opponent loading silvers on a Bandit fort board.
This explains everything.