Did fight fans recently experience “The Small Cage of Destruction” at UFC Fight Night: Boetsch vs. Henderson? Ten fights ended in a knockout or submission, tying it for second-most in UFC history. Were these finishes caused by the 25′ Octagon or were they destined to happen even in a 30′ cage? This is the key question to the cage size debate. What would have happened if the fight we just watched had taken place in a different-sized cage? It’s a world we never get to see, but can try to estimate. We can analyze these situations with treatment effect models, where the causal effect of a treatment (large or small cage) is calculated for various outcomes (fight performance metrics).
I first started thinking seriously about this issue two years ago while listening to a presentation on how changes in the technology of playing professional basketball, specifically the advent of the 3-point line, affected the productivity of different position players (hint: think center and forward spacing and rebounding). I listened to the speaker but kept drifting off to cage size in MMA. I knew Fightnomics had made one of his charts-of-averages but had no idea where his data came from, along with my usual gripe about his charts.
The largest hurdle to a more rigorous analysis was the fact that official cage-size data didn’t exist. Rami Genauer, the CEO of FightMetric, just changed all that. He’s always been extremely supportive of anything advancing the cause of metrics in fighting (pun intended) and was recently able to procure official cage-size data for every Zuffa UFC and WEC event. I got the data a few weeks ago and it was time to get down to business.
What follows is a summary of the preliminary draft of my academic research on cage size in MMA, which will be presented tomorrow at the WEAI conference in front of professional critics and then submitted to a scientific journal and peer reviewed. Today’s piece will contain minimal charts and tables – and no charts of substance – as I get nervous that a journal may try to claim the work has already been published. It’s mostly a verbal summary of preliminary findings.
The Study
This study is the first-ever comprehensive analysis of cage size using official data from the UFC and FightMetric. I analyzed every Zuffa UFC/WEC event and men’s bout from Feb. 23, 2001 through the end of my current dataset, Apr. 25, 2015. It included 312 events and 3,156 bouts, 79% of which were fought in a large, 30′ cage and 21% of which were fought in a small, 25′ cage. The only exclusion was women since they comprise the newest Zuffa divisions, have the least amount of data and may be structurally different from men in a way which could be statistically controlled, but likely would require more data than we currently have available.
Raw Data
The table below shows the raw breakdown of finish rates by weight class and cage size.
There generally seems to be more finishes in a smaller cage, but in reality finish rates are only statistically different in the middleweight through lightweight divisions. Perhaps we should separate out knockouts and submissions? The next two tables show these more detailed raw breakdowns.
The knockout chart doesn’t tell us much of anything about cage size and the submission chart only has statistical differences for welterweights and lightweights (and not even for flyweights).
Even with statistical differences, charting raw data runs into the problem that fight stats don’t come from a laboratory. They come from the real world where many other things can interfere and mess with the numbers. So that’s where more advanced modeling comes in to play.
Primary Question
What’s the observable effect on performance metrics of going from the traditional 30′ cage to a smaller 25′ cage?
My strongest findings are that we get faster fights, a larger proportion of fight time spent in the clinch (mostly with control against the cage), a lower success rate of takedown shots from distance and a higher rate of choke attempts per five minutes. Slightly weaker findings reveal that we should see more fight finishes, fewer decisions and more upsets. Other metrics are much weaker individually, yet may come together to form what is sometimes referred to as “suggestive evidence” towards a story that helps explain the stronger findings.
Faster Fights, More Finishes, Less Predictability
In a 25′ Octagon, total fight time decreases by an average of 1:04 per bout – a 10.3% reduction from the typical 10:19 fight if all cages were large. Fight time can decline for two possible reasons: (1) fights that were going to be finished anyway now end faster and (2) we get more knockouts and submissions as fewer fights go the distance. It appears both reasons help explain shorter fight times. Finished fights end on average 45 seconds faster and we see 4.3 – 5.4 percent more knockouts and submissions, roughly evenly split between the two.
If we take a middle ground estimate of the increase in finishes (4.8 percent), that’s roughly one additional finish every 21 bouts, or one additional finish every 1 2/3 events. This may not seem like much but over the course of a 500-bout year it’s 24 more finishes than if all cages were large.
Might more finishes mean more unpredictability? While not an enormous change, the rate of upsets increases 2.7 percent in a 25′ Octagon. In other words, fans should expect to see one additional upset every 3-4 events or 13-14 more upsets every 500-bout year. Almost half of these are big upsets with +300 or greater true odd underdogs prevailing over the favorite.
Fence Effects
As with all good things in life, nothing’s ever free. The knockout/submission tradeoff is the clinch, or more precisely cage control. While true fight fans appreciate the nuances and strategy of a clinch fight against the fence, the average fan likely finds it boring. Statistical evidence supporting this claim can be seen in how the UFC punishes fighters much more for idle time in the clinch than idle time at distance or on the ground when handing out Fight of the Night awards.
In a 25′ Octagon, the proportion of fight time spent in the clinch increases 2.3 percent, pulling from time spent at distance and on the ground. Virtually all of this is control time with one fighter pressing the other against the fence. This shift in fight time represents about 13-14 seconds on average for each bout, or 2 ½ – 3 minutes in a 12-bout event card.
How are fighters getting to the clinch? The most obvious candidate is that distance takedown attempts have a 5.7 percent lower success rate. There’s less space to travel in a smaller cage before being pushed into the fence for support. You can eyeball test this by examining fighter positioning in large- and small-cage environments (use the most recent Fight for the Troops event on Nov. 6, 2013 for the small cage and any major pay-per-view for the large cage). In the 30′ Octagon, when one fighter has the center, the other is usually a step or so in front of the line. Move to the 25′ Octagon and the off-center fighter is usually standing right on the line.
It appears that fights (marginally) shift towards clinch control on the fence in smaller cages. It also seems reasonable – but can’t yet be tested – that time spent at distance and on the ground would shift closer to the fence. The question then is how does this lead to more stoppages? Positioning could be key.
On the submission front, I find that the rate of choke attempts per minute increases while joint locks remain the same or decline in a small cage. The choke attempts in question are primarily guillotines and rear naked chokes (RNCs). If there’s more clinch control, we know from other analyses that this goes hand-in-hand with guillotine attempts – think of one fighter sticking his neck where he shouldn’t. If more ground time takes place at or near the fence, RNC attempts should increase as one fighter tries to get his back to the fence and standup. Meanwhile, none of these positions is good for locks. The clinch is like kryptonite to joint locks and being on the ground near the fence gives less room to maneuver for armbars, kimuras, americanas, etc.
In the brain-rattling department, interestingly enough, the total number of knockdowns doesn’t statistically increase. At the same time, knockdowns appear to be more effective when they happen, and proximity to the fence may be the reason. Knocked down fighters may have less room to roll and escape follow-up blows, especially when close to the cage.
Cage Size and Striking Effectiveness?
The information above represents my stronger statistical findings. There are also weaker, statistically insignificant findings that collectively frame a personal theory as to what else might be going on.
While clinch knockdown effectiveness improves in a small cage, there’s some evidence that overall knockdown and damage effectiveness also improve. Put another way, there’s some evidence that each strike landed does more damage in a small cage. Remember, this is on average, not that every single strike does more damage.
If fight positioning shifts to the fence or near it, a key standing defensive option is removed – the ability to step back out of range. Does this make fenced-in fighters more vulnerable? Around 75 percent of the variation in knockouts is explained solely by looking at who had more knockdowns. While we don’t see a statistical increase in the number of knockdowns, perhaps we’re getting fewer flash knockdowns in the center of the cage and better knockdown shots or finishing along the fence.
When knockdowns don’t lead to a knockout, there’s a strong, positive correlation with choke attempts and no effect on joint locks – perfectly consistent with what we’re seeing more of in a small cage. Are fighters who aren’t knocked out but are a little more disoriented getting rear naked choked on the ground? Are more damaged fighters shooting takedowns into guillotine chokes? If that’s the case, the success rate of these chokes should improve, and that’s exactly what we see.
This section was a personal theory as opposed to solid statistical evidence. Do you have any other ideas or see a different story in any of the numbers you’ve seen today? Let your voice be heard and share them in the comments below.
What Did You Expect?
Is this what you thought would happen in a small cage? Did you expect more action-oriented fights with more strikes thrown and landed? We actually see fewer strikes in total because fight time is shorter, but interestingly enough, striking rates are statistically unchanged. In fact, many metrics are statistically unchanged: strikes attempted and landed per minute (significant or otherwise), knockdowns per minute, having advanced positions while on the ground, and the overall closeness of the fights (measured objectively with RoboJudge).
The main effect of a small cage seems to be to shift fights towards the fence, creating submission opportunities and strike defense vulnerabilities. The two could possibly be related as more effective striking leads to more effective chokes, but this is just a personal theory at the moment.
Remember, these are all preliminary findings and are subject to change. The analysis also doesn’t and can’t account for any training adjustments fighters and coaches might make if the UFC were ever to permanently switch to a 25′ Octagon.
Does the small cage create crazy action, knockouts and stoppages galore? I wouldn’t go that far, but it does appear to be associated with a statistical story. The story is Fence and Finishes.
Paul is Bloody Elbow’s analytics writer. All mistakes are his own and they’ve been known to happen sometimes. Follow him at @MMAanalytics. Fight data provided by FightMetric. Special thanks to Rami Genauer of FightMetric for championing the cause of acquiring official cage-size data.