[gwnn]

Suppose we have a super-intelligent computer from the year 2200 that can calculate EV* from all actions.

Suppose we miss a game and lose 10 imps. I made a decision that had an EV of -1 imps, and my partner one decision that had an EV of +0.5. ATB.

This is an odd example, but one that I think is interesting.

*EV compared to our opponents in the team match, who are completely deterministic and the computer can predict all decisions made by them

[helene_t]

Obviously it is 10% you, -5% partner and 95% bad luck

Assume you made the decision first. Your decision had EV -1 imps. Does this mean that you had an alternative that would be one imp better? If so, partner would have difficulty assessing the EV of his alternatives since he would assume you would make the best decision so he would envision different cards in your hand. Was his +1 EV based on the assumption that you would do what you did, or was it based on the assumption of an optimal gwnn?

As for "what does blame mean": The loss of 10 imps might not be caused by your (or partner's) decisions. Maybe you would have lost 15 imps if you both had optimized EV. So the word "blame" is not really meaningful in this context imho. A better question is how good your decision was, and how good partner's decision was.

The word "blame" is more meaningful in a situation where your combined assets were inadequate for the contract you reached. So either your methods were ineffective for the given situation, or one of you misbid. Whether this led to a loss of imps is not so interesting, although of course the word "blame" would not be used if the misbidding led to a good result.

[bluecalm]

Quote

Suppose we have a super-intelligent computer from the year 2200 that can calculate EV* from all actions.

Suppose we miss a game and lose 10 imps. I made a decision that had an EV of -1 imps, and my partner one decision that had an EV of +0.5. ATB.

If the computer is that super intelligent no action has +EV comapred to actions made by it.
Assuming you just made a mistake of 1 imp and your partner bid optimally you of course get 100% of blame.
Now if we are to quantify how big your mistake was there are 2 ways of doing this:
a)you just get "real" cost which is -10imps
b)you get EV of your mistake which is -1imp

Both are equivalent, a) just has more variance as you don't "pay" on hands where your mistake didn't matter but pay a lot of if they does matter. My favorite methods of assessing my performance is a) as it's easier to calculate. Sometimes I will get more blame than I should and sometimes less but it will even out..

What you can't do (but some people try to) is to count EV of mistake on hands when mistake mattered but don't count mistakes on hands when they didn't or (much more rare) count "real" blame for mistakes which mattered and add some for boards when they didn't.

[gwnn]

bluecalm, did you notice the asterisk next to EV?

The computer could simulate one million billion billion hands and it could simulate the other 7 participants' behaviour of the team match accurately. It would then say, for example "if you x, you will win on average 1 imp/board".

[han]

bluecalm, on 2011-May-04, 07:16, said:

My favorite methods of assessing my performance is a) as it's easier to calculate. Sometimes I will get more blame than I should and sometimes less but it will even out..

In other words, you are a result merchant.

[bluecalm]

Quote

In other words, you are a result merchant.

Not really as I try to objectively (by analysing, asking better players etc.) assess if I made a mistake. Only if I think I made a mistake I "score it" by "real cost". This gives me the same expected sum of mistakes as analyzing every error separately and trying to guess what EV it had (which imo is impossible).
Result merchant would assume that every play which isn't successful is a mistake which I don't do.

Quote

The computer could simulate one million billion billion hands and it could simulate the other 7 participants' behaviour of the team match accurately. It would then say, for example "if you x, you will win on average 1 imp/board".

This is only matter of naming stuff. In you example we could as well say "if you do this (the best action) your EV is 0, while if you do other thing you would lose 1imps compared to your best action).
I think assuming that there are only mistakes and 0 plays makes it easier to think about problems.

[gnasher]

I'd blame the computer.

[gwnn]

gnasher, on 2011-May-04, 07:41, said:

I'd blame the computer.

Yes but what do you mean by blame?

[matmat]

gnasher, on 2011-May-04, 07:41, said:

I'd blame the computer.

nah. gotta blame the programmer.

[awm]

I think the idea is that the "blame" goes to the person who reduced our expected value on the board.

It's quite possible for there to be a bad result where there is no blame (i.e. only positive expected value actions were taken, but the luck was not on our side). It is also possible for there to be a good result where there is blame (i.e. someone took a negative expected value action, but we were lucky or opponents made a mistake so it worked out).

This is why these types of things are usually written like "assess the blame for missing the good slam" or "assess the blame for overbidding to the no-play game" and not "assess the blame for losing ten imps on this board."

[nigel_k]

awm, on 2011-May-04, 12:23, said:

I think the idea is that the "blame" goes to the person who reduced our expected value on the board.

It's quite possible for there to be a bad result where there is no blame (i.e. only positive expected value actions were taken, but the luck was not on our side). It is also possible for there to be a good result where there is blame (i.e. someone took a negative expected value action, but we were lucky or opponents made a mistake so it worked out).

This is why these types of things are usually written like "assess the blame for missing the good slam" or "assess the blame for overbidding to the no-play game" and not "assess the blame for losing ten imps on this board."

There can also be a bad result where -EV actions were taken but there is no blame because those actions didn't contribute to the bad result. Blame is complicated.