Let’s look at the data from the perspective of someone that would like more exciting fights (ie. promoter, matchmaker, or fan). I am interested in the proportion of fights that end in finishes vs. decisions. Everyone loves a finish, but mismatches (Overeem vs. TBA) are not as exciting as competitive fights (Condit vs. any UFC welterweight).

Only 53.5% of fights between fighters with identical reach end in a finish. This compares with 63.0% for fights between fighters of reasonably similar, but unequal reach (<1 SD (standard deviation from the mean), but not equal; effectively 0.5"-2.0"). That is a percentage increase of 17.8% from equal reach. 

This sounds great, but I want to make sure this doesn’t lead to lopsided match-ups. So let’s check the win percentage of the taller guys in this 0.5"-2.0" reach difference range. Out of this set, the longer fighter won only 48.0% of the time. Clearly there is not a reach advantage here. As others have already shown, a significant reach advantage does appear when the reach differential is greater. At >1 SD (>2.1") the longer man won 55.2% of fights, and at >2 SD (>4.3") the longer man won 58.4%.

In conclusion, if I were Joe Silva, I’d try to book fights with 0.5"-2.0" reach differentials and avoid fights between fighters of equal reach.

Caveat: Just because a fight goes to decision does not mean it is boring. See Sanchez vs. Kampmann for example. To truly see how avoiding fights between fighters of equal reach will affect entertainment value, we need entertainment value data. For all we know, more decisions between fighters of equal reach earn Fight of the Night vs. decisions between fighters with a reach differential. I suggest FightMetric collect survey data on entertainment value of past/future fights so we can play with that in future Research Challenges. 

More Stats after the jump...

Stats Overtime: To determine if the finishing percentages of 63.0% vs. 53.5% for similar reach vs. equal reach is significant I used a z-test for 2 proportions from a handy calculator site. The sample sizes of 633 for similar reach and 157 for equal reach resulted in a confidence level of 96.4% that these values are different for some reason other than merely chance. 95% is often a good cut off, so as your gut likely told you, this difference is significant.

The difference between the win % of fighters with >1SD reach advantage and the win % of fighters with <1SD reach advantage was very significant (98.9% confidence). However the finishing rate of fighters with a big reach advantage was not significantly greater than fighters with a small reach advantage (at least with this sample size). This is likely due to the surprisingly large change in finishing rate noted above when there is any reach difference. Though the sample sizes at each individual reach difference (whether divided by 0.5" or 1") are too small for statistical significance, it is worth noting that the values of each segment above 0" are greater than 60% with no trend toward the 0" outlier.