accuracy/reliability of odds

[thesloppy]

This issue is kinda creeping into another suggestion of mine, and I think it's a valid issue with the potential for lots of discussion, so thought I'd give it it's own space. The issue with the odds is kind of a complex one, but for me personally it breaks down like this:

Currently, the engine does what I think is a really good job of generating upsets, so long as the betting-line of each fighter is somewhere near the -300, +300 range. Outside of that range, upsets seem particularly rare, which makes the betting-line far too accurate in a lot of cases....compounding the problem is the fact that it's much easier and more common to set-up mismatches than it is even fights (whether on purpose or not). The upshot is that it's all too easy to spoil a fight's results by clicking on the tale-of-the-tape and seeing surprisingly lop-sided odds that clearly point out the victor.

Any possible solution seems tricky, because as noted I think that right now the engine handles close matchups well, and generates a proper amount of upsets, so I think we'd all like the logic in those cases to remain the same. What would be nice to have addressed is those cases when a fighter is a -300 to -12000 favorite, which seems to result in nearly guaranteed victory. Likewise, it's hard to even nail down what we would want from a solution? More/occasional upsets for the long odd fights (and if so, to what degree)? Maybe just less variance between odds (ie fighters would rarely get to -600 rather than -1200) I don't pay attention to real-life fight lines, so I have no clue if that is less realistic, but 12-1 odds in a sport as unpredictable as MMA seems unlikely to me.

Any thoughts at to what you do/don't like about the current odds/line logic? Too accurate for you, or no bother? Do you ignore it? Any ideas to clean it up (or abstract it further)?

[thesloppy]

I don't pay attention to real-life fight lines, so I have no clue if that is less realistic, but 12-1 odds in a sport as unpredictable as MMA seems unlikely to me.

To that end, I went and did some quick and dirty research on a few sites. This place has lots of consolidated odds/lines that match up with the other sites:

http://www.ufcdogbet.com/

It looks pretty rare to get a fighter favored beyond 4-1 odds (although I did see one guy at -900 on another site), whereas that happens pretty consistently in WMMA, and the fights between +/-200 are the rarity. I'd love to think this could be a quick-fix done by simply compressing the scale of the odds from something like a max of -1200 to a max of -600, but again the threat there seems to be similarly compressing the range of what actually are the even matchups. In other words if currently fights between +/-300 allow for a liberal upsets, while fights between +/-300to1200 point to guaranteed victories, I'd hate to see compressing the overall odds range from -1200 down to around -600 also compress the range of likely upsets to occur only in fights between +/-150, which wouldn't be much of a fix at all and just increase the feeling that the odds were too accurate.

I think the magic spot would be to reduce the outlying range down from -1200 to the more realistic +/- 600 level, while keeping the logic the same for fights somewhere around the +/- 300 level. I also don't think it would hurt to ever-so slightly increase the possibility of those +300 long-shot wins. But all of those things sound a lot easier said then done, depending how intertwined the betting-line logic is with the match engine.

[Adam Ryland]

depending how intertwined the betting-line logic is with the match engine.

To recap from the other thread on the same subject, they have absolutely nothing to do with each other. Betting lines are done on reputation and skill, nothing more. Any talk of "increasing the chances of upsets", etc, is not viable as it simply doesn't make any sense within the context of the game.

[thesloppy]

Good to know, thanks Adam. I missed the other thread, is it in the suggestions forum?

As an out-of-the box thought, what about something like taking a snapshot of the fighter's skills immediately after each of his fights, and using that to generate the odds, rather than their match-day skill levels (assuming that's what you use)? That slight difference would allow for each fighter's skills to possibly decay or improve in the interim between a fight (or have a bad camp or whatever), which would throw some slight variance into the odds and perhaps allow for a few more upsets, at least to the user's eyes. It would also somewhat mimic reality, since Vegas and the betting public are largely working off their last memory of each fighter, while a matchmaker would be privy to more timely info.

Does the scouting level have any effect on how the skills effect the odds? Maybe that could provide another level of abstraction, if it's not already figured in. I'd imagine the less fights between two fighters (and less scouting info), the closer the betting lines are...but again, I don't particularly know what I'm talking about, and am just making assumptions (my specialty!)...both in the case of Vegas and WMMA.