Matchpoints and IMPs, in Bridge and Politics
Duplicate bridge is the world's greatest card game, in part because it eliminates much of the luck from the game, thus allowing better comparisons of skill. When you play a hand of duplicate bridge, you and your partner are not directly competing against your two opponents at your table; instead, you are competing against dozens of other teams that are playing the exact same hands as you and your partner at other tables, and you are measured based on the relative amount of points you get on the hand compared to them.1 Even if you lose points to your opponents at your table, you will still get at top score if you lose fewer points than all the other teams holding your cards at other tables.2
There are two main ways to score duplicate bridge, matchpoints and IMPs. In Matchpoints scoring, you get 1 point for each team sitting your direction that you beat, and half a point for each team you tie. For example, if you scored 420 on a hand and other N-S teams score 450, 450, 420, 400, 170, 170, 170, -100, then you would get 5.5 matchpoints for that hand (you beat 5 teams and tied 1). In effect, you get the full value of beating another team no matter how little you beat them by, and you get no extra credit for beating them by a lot. Scoring 20 more than another team
In IMPs, the amount you beat another team by matters. Winning or losing a hand by 20 points---which means everything in matchpoints---means almost nothing at IMPs, you only gain 1 IMP. But if you beat a team by 520 points, you get 11 IMPs, a huge gain, whereas in matchpoints you get the same 1 point as if you beat them by just 20.
This has huge implications in bridge strategy, both for bidding and play. Face with two possible slam-bonus contracts you might want to bid (say, 6 spades or 6 notrump), at IMPs your goal is to choose the safest contract. Which one will you successfully make more often? The fact that 6 spades scores 980 and 6 notrump score 990 makes literally no difference; at IMPs, a 10 point difference in score is worth zero IMPs. What can crush you is not making the contract---getting a score of zero while everyone else gets 980. If you are 85% to make the 6 spades and 75% to make the 6 notrump, the only play is to go for 6 spades.
At matchpoints, the 10 point difference is worth everything, and losing by 980 is no different than losing by 10. So there, you usually want to bid the 6 notrump contract, because you will get the full value of beating everyone who plays 6 spades when both contracts make, and you will only lose 1 point to each team the small amount of time 6 spades makes and 6 notrump gets beat. So you give up on safety and stretch to make a small number of extra points.
Likewise, this safety vs. stretching affects how you play the hands. At IMPs, you would basically never make a play that jeopardizes your contract in order to try to gain an overtrick. That is, if your contract is for 9 tricks, you never make a play to try for a 10th trick if it has any chance of costing you from making 9. There's no point in risking a sure-thing 420 point contract just to get an extra 30 and finish with 450.
Conversely, at matchpoints you often do risk the contract---it's usually3 mathematically correct to go for it if you have a 50% chance or greater---to make an overtrick, because getting to 450 might mean fully beating every other team, and the downside loss of going for it and failing is capped at a single matchpoint, rather than a pile of IMPs.
There are a lot of applications to this in politics.
Elections are essentially played at matchpoints. The goal of any candidate is to not to maximize their vote share, but to maximize the probability that they get more than 50% of the vote. There's no prize for losing by a little, so you go for broke to get to 51%. 48% and 28% get you the same score.
Conversely, once a candidate has built an electoral coalition that guarantees them, say, 57% of the vote, they should never do anything to grow that vote if it risks blowing up the coalition and dropping them below 50%. Even if a strategy has a 95% chance of getting you from 57% to 67%, if it has a 5% risk of dumping you below 50%, it's garbage. You have already gotten the full value of the victory.
On the other hand, policy negotiations in Congress or the executive branch look more like IMPs. You can "win" in ways big and small, and the size of the victory or loss matters. This cuts both ways, of course. If your policy push completely blows up on you and you get nothing, that's a huge loss compared to just taking the safe road and passing something modest. On the other hand, if you settle for something modest when you could have shot the moon, you have left a ton on the table. The magnitude of alternative outcomes matters, and the win/loss is not binary.
Similarly, your opponents at your table are playing against all the pairs of people holding their cards at other tables. That is, if you and your partner are sitting N-S and your opponents E-W, you are actually competing against all the people sitting N-S, and none of the people sitting E-W.↩
Duplicate bridge does not remove all the luck from the game. Some opponents you play at the table might be better or worse. To minimize this issue at a duplicate bridge match, you play different E-W opponents for a few hands each, so that everybody sitting N-S plays about the same amount with everyone sitting E-W. This doesn't fully solve things; you can get lucky and play a really good opponent for a bunch of hands that are very easy and have little skill to them, negating their edge. Or vice-versa.↩
Duplicate bridge is pretty wild because part of the estimation for these decisions is based on your assessment of the field you are playing against. For example, if you knew for a fact that every other table had bid 3NT, but you had managed to get to 6NT for the potential slam bonus, now you would never risk the contract for an overtrick, even at matchpoints, because you know just safely making 6NT is going to allow you beat every other team. No one said bridge was easy.↩