Lower Weight Classes Delivering More Upsets

by GregS123 on Oct 17, 2009 2:29 PM EDT Comment 20 comments

Over the past two years 30% of MMA bouts ended with an underdog pulling off an upset over the oddsmaker’s favorite. Tapology broke these bouts down by weight class and found a strong correlation between weight division and upset rates. The lighter the fighters are, the more likely there is to be an upset, with a full 7% more bantamweight fights than heavyweight fights being won by an underdog. While it can’t tell you which fights to wager on, it does show that MMA oddsmakers are less capable of predicting what will happen in the smaller divisions.

32% of 135 pound bantamweight and 145 pound featherweight fights end in an upset by the underdog. The upset rate declines with each step up in weight, all the way up to 265 pound heavyweights where upsets only occur 25% of the time. The one notable exception is the 205 pound light heavyweight division, where, strangely, over the past two years there have been more upsets than in any other weight class. Whether this is due to the fighting styles at light heavyweight, the parity of the division or simple randomness is difficult to say.

Despite the 205 pound anomaly, the overall trend is clear: the smaller the fighters, the greater the chance of an upset. Take this into account the next time you are on the fence about going with a bantamweight underdog versus a heavyweight underdog.

Weight Class	Fight Count	Favorite Won	Underdog Won
Bantamweight – 135 lbs	34	68%	32%
Featherweight – 145 lbs	74	68%	32%
Lightweight – 155 lbs	187	69%	31%
Welterweight – 170 lbs	127	72%	28%
Middleweight – 185 lbs	146	73%	27%
Light Heavyweight – 205 lbs	85	67%	33%
Heavyweight – 265 lbs	76	75%	25%

Source: Tapology.com

The FanPosts are solely the subjective opinions of Bloody Elbow readers and do not necessarily reflect the views of Bloody Elbow editors or staff.

4 recs | Comment 20 comments

Story-email Email Printer Print

Comments

Your methodology is off. You can’t just take “favorite” and “underdog” and find out what differences, if any, there are between weight classes. You have to

1) Remove the juice from the line.
2) Convert the line to an expected win percentage.
3) Compare the expected win percentage to the actual win percentage.

Twitter: @Mike_Fagan_13
http://www.sackmikegoldberg.com

by Mike Fagan on Oct 17, 2009 3:11 PM EDT reply actions 0 recs

Also, I have no data to back this up, but I’d expect that there are more mismatches at the heavier weights.

Twitter: @Mike_Fagan_13
http://www.sackmikegoldberg.com

by Mike Fagan on Oct 17, 2009 3:17 PM EDT up reply actions 0 recs

It could also be the case that lighter weight classes are followed less than the heavier weight classes. That kind of popularity imbalance could skew the betting lines.

by George Lucas on Oct 18, 2009 2:50 AM EDT up reply actions 0 recs

hmm… but i worry that would focus so much on analyzing oddsmaker efficiency (and get complicated and confusing fast) rather than just show people in a simple fashion “upset rates” by weight class.

www.tapology.com | twitter.com/tapology

by GregS123 on Oct 17, 2009 3:20 PM EDT up reply actions 0 recs

But a +110 “underdog” winning isn’t the same thing as a +800 superdog winning. And, like I postulated, I have a feeling the lines for the lighter weights are much closer than the big boys.

And you are analyzing oddsmaker efficiency. You’re just doing it poorly.

Twitter: @Mike_Fagan_13
http://www.sackmikegoldberg.com

by Mike Fagan on Oct 17, 2009 4:05 PM EDT up reply actions 0 recs

And, like I postulated, I have a feeling the lines for the lighter weights are much closer than the big boys.

I took a look, it’s kind of interesting. Split it by WEC vs UFC because that was the easiest thing to start with, and is somewhat like light vs heavy.

-100 to -199……46% of WEC fights……39% of UFC fights
-200 to -299……15% of WEC fights……30% of UFC fights
-300 to -399……21% of WEC fights……17% of UFC fights
-400 to -499……8% of WEC fights…….7% of UFC fights
-500 or more……10% of WEC fights……7% of UFC fights

What do you think? I could give you an average instead of a distribution, but this seemed better.

www.tapology.com | twitter.com/tapology

by GregS123 on Oct 17, 2009 4:29 PM EDT up reply actions 0 recs

The lines are of equal width but not of equal percentage widths, the percentage widths is what really matters. Just to illustrate:

-100 to -199 is a range of 50% to 66% – a 16% difference
-200 to -299 is a range of 66% to 75% – a 9% difference
-300 to -399 is a range of 75% to 80% – a 5% difference
-400 to -499 is a roange of 80% to 83% – a 3% difference

So those ranges you posted don’t have consistent or equal winning percentage ranges. If you want equal ranges they are:

50% to 55% is -100 to -122
55% to 60% is -122 to -150
60% to 65% is -150 to -185
65% to 70% is -185 to -233
70% to 75% is -233 to -300
75% to 80% is -300 to -400
80% to 85% is -400 to -566

When you have small equal ranges you can then compare the winning percentage to the money line percentages, which you can’t do for the first few odds in your post as the interval is just too wide.

by Donk696 on Oct 17, 2009 5:29 PM EDT up reply actions 0 recs

cool. i could do the breakdown for you with these ranges, but it would take a little effort. if you want it, say the word, but i’ll hold off until then.

www.tapology.com | twitter.com/tapology

by GregS123 on Oct 17, 2009 5:36 PM EDT up reply actions 0 recs

Mate that would be awesome if you could. I’ve looked through alot of statistics myself for UFC events, but never ones by weight class and never across promotions. I really need any edge i can get over the bookies, i’ve put in hundreds of hours of research but mostly compiling databases, but i havn’t come up with as compelling statistics as you have.

What i did find out from my own research is big discrepencies in winning percentages vs money line percentages that favour the favourite in main card fights, but not the case for favourites in the prelims. But looking only in UFC events from UFC 80 – UFC 103. Some more research here would be important, especially across weight classes.

I also found a slight anomally in the prelims where the underdog wins slightly more than he should. Giving you a small positive bankroll on average. I think this could be to do with up and coming talent mixing in with established talent in the prelims. I wanted to do some research into UFC Fight Night events as this is more where the up and comers first make their appearances and mix in with the established talent pool, but i havn’t had the time recently to do a study. Breaking it down by weight class here would be really usefull as well.

If you could make a breakdown by weight class and by main card and prelim this would be a big, big help for me. All of this across UFC, WEC and UFC Fight Nights would be the ultimate guide. I think looking at recent events say mid 2007 onwards would be the most relevant as there are some big scenarios before then that skew the findings like when Pride mixed with the UFC and so on.

Also is the money line you used in the above example the best money line right before the event? Or is this the best money line for a particular bookmaker over the course of when they accepted bets? Or an average across bookmakers before the event?

by Donk696 on Oct 17, 2009 6:03 PM EDT up reply actions 0 recs

Dude, as long as you have a more than casual interest in MMA and can hit early lines/lineshop, you’ve already got a huge edge on the bookies.

Twitter: @Mike_Fagan_13
http://www.sackmikegoldberg.com

by Mike Fagan on Oct 17, 2009 6:13 PM EDT up reply actions 0 recs

I hope so…

Acurately predicting percentages is the hardest part for me though, but by knowing if a line is going to get less favourable is the best way to make a bet I find. Then line shopping just maximises the process.

Well, in theory…

by Donk696 on Oct 17, 2009 6:25 PM EDT up reply actions 0 recs

it’s the average of the best bet and the average bet, taken off of bestfightodds.com because they seem to do a solid job in keeping it all updated. it’s the final odds right before the fights, which i know are not the same as the odds when the lines first come out.

btw i don’t personally bet on fights, and you guys clearly know a lot about it. but playing with the data is interesting to me, and so we’re posting articles about it on our site.

www.tapology.com | twitter.com/tapology

by GregS123 on Oct 17, 2009 6:14 PM EDT up reply actions 0 recs

I use bestfight odds a lot as well, but do you input the data in manually by hand or write your own software to read them out?

I did all the UFC fights from 83 to 103 by hand together with all the outcomes and what portion of the card they were placed and it took me forever. I went mad near the end.

by Donk696 on Oct 17, 2009 6:30 PM EDT up reply actions 0 recs

ummm yeah doing this by hand would drive you insane. it needs to be set up in a relational database so that you can sort and filter and change it easily.

www.tapology.com | twitter.com/tapology

by GregS123 on Oct 17, 2009 6:42 PM EDT up reply actions 0 recs

also...

uh… happy to help you, but i’ll need you to ask for something a little easier.

breakdown by:
- 7 weight classes
- 2 or more billing types (maincard, undercard, title, etc…)
- 3 or more promotion (wec, ufc, fight nights, etc…)
- 7 or more moneyline ranges (100 to 122, etc…)

that’s like 7*2*3*7 = nearly 300 buckets. it would take a long time and plus there is only a sample size of like 800 fights in the past 2 years to work with.

www.tapology.com | twitter.com/tapology

by GregS123 on Oct 17, 2009 6:35 PM EDT up reply actions 0 recs

Yeah that’s a bit of a killer…

I think you’re gonna need at least 30 fighters per bucket to have anything meaningful.

Perhaps you can group the weight classes together that show a similar performance, i would imagine Bantam and Feather could be grouped, as could Light and Welter weight, Middleweight and Light Heavy could be grouped and leave Heavy weight on its own. That’s 4 groups. Then say have 10% gaps on the money line ranges instead and use -400 plus as the last range, that would give you 4 ranges. But that’s still 96 buckets.

Take out the billing and that’s 48 buckets, and 800 fights that’s about 20 per bucket, still not great but better than nothing. This would be a good study.

I bet on such a range of money lines that to me it’s not a major issue, so you could do a seperate one with putting in the billing and taking out the money line ranges and that would take it down to 24 buckets or about 40 per bucket which is a good sample size. For me the billing is an interesting topic as it has shown some good results, so a seperate one with that in would be fantastic.

Good job Greg, thanks for the effort.

by Donk696 on Oct 17, 2009 7:05 PM EDT up reply actions 0 recs

In my own time i did a study of UFC events from UFC 80 – UFC 103 and looked at the average winning percentages of favourites and underdogs on the main cards and prelims vs their money line winning percentages.

On the main cards i found the favourites win around 75% of the time on average vs. the best money line average at the time of the fight of around 70%. In other words the favourite on the main card wins 5% more than he should on average. The bookie loses money here.

For the underdog the news isn’t as good as he wins only around 24% on average vs the best money line odds at the time of the fight of around 34%. Here the bookie is overestimating the odds by about 10% on average for the underdog. The bookie makes alot of money here.

For the prelims it’s bad news for the favourite this time of 64% vs 70% and for the underdog 35% vs 35% (just under 35%). The bookie is doing a good job here.

From the data, it’s the favourites on the main card that show the most promise, with the bookies underestimating their chance’s by about 5% on average at the time of the fight. Considering the odds are usually at their least favourable point at the time of the fight, the favourites on the main card is where the best money is to be made.

MMA betting blog has their own studies from UFC 67 – UFC 77 and they have some nice statistics that show that in the last two fights of UFC cards (the main events of the card), the underdog wins just below 50% of the time. Which is a severe difference from the average money line percentage. This trend continues to around UFC 82 i think, but then evapourates afterwards. The reasons are that this was the time when 3 separate pools of fighters merged together, the UFC and pride mix together with a mix of up and comers (e.g. Shogun vs Forrest, Liddell vs Rashad). But this percentage falls rapidly after UFC 82 as you don’t have such mix of seperate talent pools.

by Donk696 on Oct 17, 2009 5:05 PM EDT reply actions 0 recs