IMP Pairs

[awm]

Most actions can be classified by their expected score and their variance. In general we want to avoid actions with very negative expected scores -- they're just bad on average and while you might get lucky occasionally and get a good score, this is greatly outweighed by the frequency and magnitude of the bad scores. And in general actions with very positive expected scores are good. The more interesting question is deciding between actions of roughly equal expectation, but very different variance.

 

In a long match, variance tends to be damped out. If you do something that wins ten imps half the time and loses ten imps half the time, that's potentially high variance, but after twenty or thirty trials the total variance is reduced. In other words, with a lot of boards left to play you're normally better off just taking the action with the best expectation.

 

In a short match (or late in a long match where you know your approximate standing) it can become a good idea to seek high variance tactics. For example, say you have five boards to play and you need to generate 30 imps to win. You're more likely to manage this by taking "swingy" actions than taking "down the middle" actions. Again, you want to avoid things with very negative expected scores, but given the choice between two actions of roughly comparable expected score, you want the one with higher variance (or in a team match, the one not taken at the other table). On the other hand, in a short match where you are leading and need to "not lose 30 imps" in order to win, it's a good idea to avoid high variance tactics (again holding expectation equal).

 

My suggestion is that 12-board IMP pairs tourneys with a large field (as held frequently on BBO) suggest high variance tactics, since you basically need to generate 4 imps per board in order to win. In a 12-board team match with roughly equal teams, there's no reason to favor high variance tactics (unless you're already down) since you just need to generate more positive imps than the other team.

[mike777]

Just based on what I have seen online in these imp pairs-12 bds:

1) Just play the very best you can in the play, defense and bidding, normal the first ten boards.

2) Last two, if you think you are doing very well, change nothing. If not, then think of doing something a little swingy in the play, example play for 3-3 as opposed to a finesse or make an offbeat preempt or opening psyche in third seat.

 

This is sort of the opposite of what adam suggests. Again the first ten boards just play your normal, best imp game.

[barmar]

I think we need to go back to basics. The basic strategy in IMPs consists of 3 things:

 

1. During the auction, be a bit aggressive (but not reckless) in bidding games, especially when vulnerable. Be less aggressive in sacrificing, especially against part-scores. Double for penalties only when you're sure you're setting them, especially if you're doubling a part-score into game.

 

2. When declaring, play safely to make your contract. Don't worry much about overtricks unless they're close to risk-free. If the contract is almost hopeless, play for whatever lay of the cards would allow it to make, even if this risks an extra overtrick or two; the exception is if you're doubled (i.e. don't go for a number).

 

3. When defending, setting the contract is the primary goal. Unless you doubled them, don't worry too much about how many undertricks you get.

 

These are the basic principles that distinguish IMP from Matchpoint play, and they're just as valid for IMP Pairs and IMP Teams.

 

About the only time you really need to take high-variance action is when you're pretty sure you're way behind, because you just made a stupid mistake that gave away double-digit IMPs. In a team game, you might just pray that your teammates did something brilliant to make up for it. But in IMP Pairs, your teammates are "the field", and brilliancies tend to get averaged away, so you're on your own. In a short match, you probably need to get at least two good boards to recover from that one disaster if you want to make it back into the money.

[mycroft]

in a 12-board BBO IMP pair game, it's interesting - if I think I'm doing well, I'm in a bind. If I think I'm doing not well, it doesn't matter what I do, "not well" means I'm going to need 20+ IMPS on the last two boards, and still may not come close.

 

If I think I'm in the hunt, it's a question of "do I want to win or to place?" If second place is the first loser, then play like you're down 10. If taking a good game back to zero or -5 is bad, then play a good IMP game and see if I can crack second place.

 

Remember, I've seen winning scores in a 12-board BBO tourney that rival winning scores in a 27-board IMP pair session.

 

Michael.

[Rossoneri]

Hmm thanks for all the replies.

 

I played my first IMP pairs tournament today, 2 sessions, with a friend of mine (but first time we have actually sat down as partners) and we got....-112 IMPs. Lol.

 

Interestingly, the pair that won only obtained an average of slightly more than 1 IMP per board. (~56 IMPs over 54 boards)

[ArtK78]

ArtK78, don't you see how you are contradicting yourself?

I find that you can create swings out of thin air by very aggressive preempting and overcalling at IMP pairs.  If you are good at it, the vast majority of these swings will be in your favor.

There is nothing worse than having one pair of a team trying for a "big score" and accomplishing exactly the opposite (which is what usually happens)

Leaving 12 IMP pairs tourneys on BBO aside, I really don't see a difference between a long IMP pairs event and a long KO team match.

No, I am not contradicting myself. You are taking my quotes out of context.

 

If I am playing in a team match, with teammates of my choosing, I do not try to create swings. I assume that my teammates will play well, and if my partner and I play well, we will win. This is the context for my second quote.

 

In IMP pairs, I know that it may be necessary to create swings to do well. This is the context of my first quote.

 

Team events and pair events are very different animals.