Doubling Cube applied to Go

[yoyoma]

I was thinking about the topic of people complaining that their opponent doesn't resign, or *gasp*, resigns during score phase. Personally I have a pretty high tolerance for opponents not resigning, but I thought of a fun idea: Use a doubling cube. This is from Backgammon where when one side gets an advantage, he proposes to double the stakes the game is played for. The opponent then has to choose: Pass (aka resign for Go), or accept the game be played for double stakes.

"Double stakes" applied to Go could be: The same/similar to backgammon, if the games are played for money or if they are played for some sort of online fun/honor betting points. Or it could just be increasing the amount of rating points at stake (we know from some other threads that many players are very protective of their rating points! ).

The most obvious problem is that the correct doubling point is 75% winning chance, so if you double with ~76% winning chance, the opponent should resign. I feel for a game like Go this is too low, correct strategy would lead to doubling out the opponent too early. So the doubling cube could be modified to instead be offering to give 6:2 odds or something (person offers to pay 6X if he loses, or take 2X if he wins -- this raises the drop point to 87.5%). Or if you wanted to use this device purely as a way to get a stubborn opponent to resign a game that is clearly lost, you could make the odds even more steep.

Well I don't foresee this idea being taken up by many places, certainly not KGS, but it just seemed like a fun idea to think of.

[Chew Terr]

Or, so you do not 'double out' your opponent too early-- add a rule that you can't double again until you've captured 1+ stone. So unless there's a huge ko fight (in which time it's a properly exciting time to double), you can't just double every turn to raise the stakes too early.

I haven't played a ton of backgammon, but I've always thought the 'meta-gamey' aspect of the doubling cube was interesting.

[yoyoma]

Or, so you do not 'double out' your opponent too early-- add a rule that you can't double again until you've captured 1+ stone. So unless there's a huge ko fight (in which time it's a properly exciting time to double), you can't just double every turn to raise the stakes too early.

I haven't played a ton of backgammon, but I've always thought the 'meta-gamey' aspect of the doubling cube was interesting.

Standard backgammon cube rules apply: After the initial double, only the player who took the cube can redouble. So only if the underdog makes a turnaround.

[Chew Terr]

Or, so you do not 'double out' your opponent too early-- add a rule that you can't double again until you've captured 1+ stone. So unless there's a huge ko fight (in which time it's a properly exciting time to double), you can't just double every turn to raise the stakes too early.

I haven't played a ton of backgammon, but I've always thought the 'meta-gamey' aspect of the doubling cube was interesting.

Standard backgammon cube rules apply: After the initial double, only the player who took the cube can redouble. So only if the underdog makes a turnaround.

Oh, that makes sense. I forgot how exactly it worked, just remembered the basics.

[TMark]

This was done at Francis Roads' 60th birthday party, some years ago. The extra feature was that one player at the party wasn't told the rules (me) and had to figure out what was going on. I think that I ended up winning every game played.

Best wishes.

[yoyoma]

This was done at Francis Roads' 60th birthday party, some years ago. The extra feature was that one player at the party wasn't told the rules (me) and had to figure out what was going on.

Since you didn't give details, I am imagining the rules included how much money was at stake.

I think that I ended up winning every game played.

Best wishes.

And now I'm imagining the conspirators being forced to pay up after TMark learned the rules.

[Mef]

The most obvious problem is that the correct doubling point is 75% winning chance, so if you double with ~76% winning chance, the opponent should resign.

Actually....I think the most obvious problems are:

-Backgammon is a game where you can reasonably calculate specific probabilities for winning at pivotal moments (ex: If you will hit and have a winning position if you roll a 5 on either die, odds of getting a 5 are 11/36)

-Backgammon is a game where you can score extra points for a stronger victory (it can be too good to double if you have a strong possibility of winning 2 or more points, and give your opponent the opportunity to resign for just 1)

-Backgammon has high volatility even with perfect play because of the randomness involved.


Now....There are ways you could adapt go to make the cube more interesting....The easiest way would be to incorporate any system where margin of victory is valued. A couple other modifications: You need to set a baseline value for a cube resignation (say 10 points?). An opponent with an 80% chance of a 10 pt win and a 20% chance of a 20pt loss (+4pts net), may want to double and have the opponent accept. (position is now worse for them as it is +8 points for you, but not as bad as a dropping the cube and losing 10 points)...but if you had a 80% chance of winning with 20 points, they may drop the cube, because they could accept a 10 point loss...

[jts]

I think this idea makes perfect sense if you're betting substantial sums of money on Go. If you want it to interact with the rating system, on the other hand, it will just give unscrupulous players another way to monkey around with ranks while making the game less fun for everyone else.

[Bazoo]

How about playing a round robin or similar, where the point difference in each game is your score: ie territory scoring 94-86 gives the players a +8 & -8 respectively. If a resignation counts as, say, -100, then the 'too good to double' works too.

Also to note, the correct doubling point in backgammon isn't 75%. It changes, depending on various factors such as volatility considerations etc. 25% is the theoretical take point though this too is usually shaved a little due to ownership of the cube. This is probably explained very badly - where is stonedaze when you need him?

Bazoo

[shapenaji]

The other way to do this is tournament style, so you can get more than just 1 point for a win.

Instead of doubling though, perhaps the better way to do it is just an incremental gain, make it worth 1.5 games instead of just 1, then 2 games, then 2.5 etc...

[tapir]

There are millions of people who play Backgammon without Doubling Cube.

[hyperpape]

There are also many people who play go without handicaps, and millions throughout history who played it without komi.

[Joaz Banbeck]

Instead of using the cube to affect money, use it to affect ratings. When the doubling cube is used, the drop in the loser's rating is doubled, and the increase in the winner's rating is doubled.

[speedchase]

Joaz the main issue with this idea in my opinion is that the purpose of the rating system is to find accurate matches, not award people who do better. honestly this system will only hurt the accuracy of the system if it was used for rating. In addition there is also the issue of people using it to mess with the rating system as jts already articulated.

[tapir]

There are also many people who play go without handicaps, and millions throughout history who played it without komi.

I object to the idea that the doubling cube is part of "standard backgammon", it is only in a myopic view that takes New York as the center of the world and doesn't look far beyond. I am pretty sure that the majority of players doesn't use it.

The introduction of a doubling cube to Backgammon has a far bigger impact to the game than komi had on Go. But the rationale behind making an assessment of the winning percentage so important is clearly the role of randomness in the game. In late endgame in Go you could safely double even when you are ahead only a little, and even earlier the small lead can easily translate into very high winning percentages (especially for better players). No randomness -> no doubling cube.