Some pros and cons of various B match-making approaches. Another long blog, another current hot-topic, feel free not to read it. But for those who care I hope it's a decent summary of the various options and opinions about them.

This doesn't touch on the issue of people not wanting their teams to get beat up - I see that as partly a side-effect of the other issues, and partly what B is supposed to be about in the first place anyway. It's mostly trying to deal with "min-maxing", without introducing new problems.

Incidentally, to me, "min-maxing" is what people are supposed to be doing. But, when people start throwing the term around though it's a sign that what people "should" be doing (i.e. playing optimally) is at odds with what people want to do (i.e. have fun). That is a problem, I just see it as a problem with game design rather than min-maxers themselves. Since the match-making itself is a game design parameter that Christer is at least theoretically willing to change it's worth talking about what could be done.

So, here's some of the possible inputs to matchmaking, and the pros and cons of using them.

TV
+ Reduces the number of inducements given out (they work ok but you can get sick of having 1M worth of them every game)
+ Better matchups (inducements reduce the imbalance but aren't even supposed to remove it). In particular avoids massive TV skilled-out teams totally outclassing rookies (which could be WMD's style bashing, or just any rock-solid high TV team)
- "min-maxing". The core of the problem is that normally (e.g. in leagues) each bit of TV only needs to be worth its weight in inducements, and so in general most TV increases are worth having. With TV-matching, each bit of TV needs to be worth its weight in opponent TV. It might seem like a subtle difference, but in a game theory sense it creates a downward spiral in TV, because in general the least-useful piece of TV that you have is worth getting rid of because on average your opponent will lose something equally valuable. The spiral only stops at new teams, who are constrained in how they spend their TV, and at "sweet spots" for particular races/builds where to go lower they would be giving up something more valuable than their opponents of a different race. Rookie teams getting beat-up and losing in a frustrating manner is just a side-effect of the underlying system.

Games played
+ Avoids new teams being preyed upon by experienced teams
- Doesn't help old but beaten-up teams, or indeed even teams that are simply slow to develop
- Allows cheesy play by repeatedly starting new Amazon/Undead teams to pick on new teams of races that are weak early on.

Team W/D/L record
+ Min-maxers can play against other min-maxers, everyone's happy.
+ Depending on the formula, it naturally accounts for games played as new teams have similar W/D/L records. Though it also pulls in the negatives of "games played" in the process...
- Need some way to balance recent record against long-term record, that accounts for both random variation and a team's changing line-up. In particular it's unavoidably going to be slow to account for teams getting beat up (or, in theory, bloated) and no longer being as competitive as they once were.

CR/BWR
- Generally inferior to W/D/L record. Some coaches have a mix of joke and competitive teams. Even the NAF breaks up coach records by race.

Race - by which I mean adjusting any of the above factors to account for weak/strong races or racial matchups
+ Mitigates some of the problems of other methods
- Doesn't adjust for different "builds", e.g. rookie chaos team vs min-maxed chaos team.
- High increase in complexity for marginal gains.

Fan Factor
+ An existing value that already provides a formula with many of the desirable characteristics of a W/D/L-based system.
- Heavily punishes teams for buying FF. Not a big deal as such, but a silly property of the system. Also makes teams hate rolling ff+, which is an odd thing to do.
- Low signal to noise ratio. Basically there's about 4 fan-factor tiers: new (0-4), poor/average (5-7), good (8-10), on-a-roll (11+). Given that your fan factor is always +-2 or so just from dice rolls, FF is not strongly differentiating.
- Impossible to fine-tune its behaviour.


My personal recommendation:
To me, W/D/L is the winner. I would create "TR", a Team Rating based off the W/D/L record. I would use something like the ELO-based system used for CR, except with a high adjustment factor (i.e. make it change rapidly, to account for a team's changing fortunes relatively quickly). At zero games, I would actually initialise the value of "TR" with the coach's own "BWR". This is partly to counter the strategy of (for instance) repeatedly creating new Amazon teams then retiring them when their "TR" gets too high, and partly just a reasonable thing to do anyway.

I would then match-make based on a combination of TR and TV, though primarily TR. Including TV though will avoid having huge TV differences too often, which I think is worthwhile as good as inducements are. The other is to more rapidly adjust to teams getting beat up - TR will still take time to adjust while TV drops immediately.

I would also allow any and all games to be played in preference to no game. If there's only two teams in the box, then (unless arranged games at quiet times starts to become a problem or something) let them play, whoever they are. TV and TR would be used to choose who plays, but not to limit it.

Personally I think this combination captures the best elements of all of the above ideas. But I welcome any informed and sensible opinions to the contrary.