Data Mining – SoCal Swordfight Longsword

This article is a cursory summary of the overall results of the SoCal Swordfight Longsword tournaments over the last three years. One must be careful when drawing too many conclusions from this data — likely there is enough random variation to make all but the most significant trends meaningless. In many cases, what is most interesting is not what changed, but what did not change.

I don’t pretend that this constitutes a complete understanding of the situation. Consider it more as ‘food for thought’, a starting point to determine what questions should be asked of ruleset changes, and how to determine if they have been effective.

The data is a lumped combination of all the different longsword tournaments held at SoCal Swordfight events. It will also (hopefully) be an interesting experience for you to compare and contrast with the CombatCon Longsword tournaments, which originally had the same rules but have evolved in a slightly different way. Data Mining – CombatCon Longsword.

Additional noise factors

More or less I didn’t address the following in my analysis:

Judge Error: The data used is what was recorded by the score table. The assumption is that the judge error was random rather than systematic in nature, and on average does not affect the overall results.

Point Averaging: When faced with judges calling for a 3 point and a 1 point target, the director will average the two scores into 2 points. It is believed that this occurs seldom enough, and is evenly distributed enough, that it does not constitute a significant distortion of the results.

Attendance: The attendance of the three tournaments is only somewhat homogeneous. While the main clubs in attendance remained the same, they all experienced significant turn-over in the individuals who attended. Additionally, while the results were dominated by a few clubs, there was a large base of fighters from other schools which was fairly transient in nature.

Skill Development: In addition to the change in targeting brought about by point values, the skill level of HEMA practitioners is rapidly improving year to year. This means that the techniques and tactics available to practitioners would be greatly expanded between 2015 and 2018.

What about SoCal 2016?

When SoCal moved from a November date to a February date it skipped a year. There was no SoCal 2016.

Significant Score Values

SoCal 2015 / Baseline

The following ruleset was used at multiple events in 2015, including SoCal. The results of all three events are combined to give the baseline statistics.

• Head: 4 points
• Torso: 3 points
• Leg: 2 points
• Arm: 1 point
• Direct targeting of hands forbidden, points given to hand strikes if defender moved them towards the path of an oncoming blade
• Deductive afterblow: An afterblow reduces the point value of the initial strike by 1 point.

(SoCal 2015, CombatCon 2015, and PNW 2015)

SoCal 2017

• Addition of controlled attack Head/Torso for 5 points.
• ‘Head’ reclassified as Upper Opening. Includes strikes landing on the top of the shoulder.
• Torso: Point value reduced from 3 –> 2.
• Legs: Point value reduced from 2 –> 1.
• Hands are a valid target

SoCal 2018

• Cut to Torso: Point value reduced from 3 –> 1.
• Controlled attack now a +1 point bonus to any target

Target Areas

This data represents the number of scoring exchanges, not the total number of exchanges. Non-scoring exchanges, such as double hits, are not represented.

SoCal 2015 ->  SoCal 2017

Arm/Leg Strikes

The data from SoCal 2017 is a little more difficult to interpret, as both the arm strike and the leg strike were worth 1 point and lumped together (in the plot they are both in the ‘Arms’ group). However, the data shows that at SoCal 2017 there were more points awarded to this region than both Arms and Legs combined in 2015.

Given that that Legs were worth less points, and the prohibition against direct Hand targeting was removed, it is likely that the legs were hit as much or less than the baseline, and the arms were hit significantly more.

Torso Strikes

As is expected, the reduction of the point values for the torso decreased the number of attacks striking the torso. It is worth noting that the expansion of ‘Head’ to ‘Upper Opening’, including the tops of the shoulders, would also have decreased the number of strikes being recorded as torso strikes.

Head Strikes

The 2017 rules had the head as a relatively higher value target in comparison to other targets. In addition the target area of the head was increased.

This did not increase the number of attacks landing on the head, and considering the circumstances could actually be interpreted as a decrease in the total number of attacks striking the head.

Possible explanations:

• WIth the torso worth less, and the ability to strike hands, fighters were less likely to engage and attack deeply.
• Fighters were more concerned about being hit in the head, and defended it more seriously.
• As the skill level of the community increases, fighters tend to not drop their hands as often, which leaves the head less vulnerable and the fighter more likely to be hit in the hands.

SoCal 2017 -> SoCal 2018

Arm/Leg Strikes

This data allows us to separate the arm and leg strikes. The total of these two attacks combined are similar to the levels in 2017. The extremely small percentage of leg strikes supports the hypothesis that in 2017 the 1 point targets were primarily attacks to the arms.

This also shows that when the legs are not given a point advantage over targets like the arms, they are very seldomly targeted — a conclusion which is in line with what source materials advocate in technique.

Head Strikes

Despite the organizers’ best efforts to incentivise Head targeting, there is still not an appreciable increase over the baseline. Based on the 2018 changes and the results produced, it’s not possible to rule out any of the explanations given in the 2017 section.

Control Points

The number of control points achieved against the head and torso increased from 9% to 13%. This could possibly reflect an increase in fighter skill, or could simply be a random fluctuation.

New Control Point targets

In 2018, fighters were able to be awarded a control point on top of any scoring actions. This was not widely used, with only 1% of attacks to previously non-controllable targets receiving control points. Because of this I conclude that the effect of this change is minimal.

It is also a possibility that fighters are not in the habit of training/attempting techniques like Hand Pressing in tournaments, and did not possess the skills necessary to exploit the rule change. This is a variable that should continue to be monitored.

Exchange Types

Over the three tournaments, the BpE* remained fairly stable between 19 and 21%. Currently there is not a wide-spread use of BpE, and it is difficult to compare to events around the world. Based on my experience (running different events/grabbing score sheets where I can/tallying points on live streams), these numbers are as clean as, or cleaner than many other longsword tournaments.

In 2017 the number of No Exchange calls increased sharply, a trend which was observed in many West Coast events at the time. It was speculated that this was more likely the result of increased judge skill in detecting insufficient cuts, rather than a reduction in the judges’ ability to determine the exchange. As a result a new call, No Quality, was added to the score system in an attempt to quantify the two.

In 2018, the No Quality calls more than make up for the increase in No Exchange calls between 2015 and 2017.

*Bilaterals per Exchange = (Doubles + Afterblows) / All Scoring Exchanges

Conclusion

As was mentioned in the introduction, it is not possible to draw many hard conclusions from this data. Here are a few items that stand out, but there are still questions as to why.

• Compared to 2015, judges are more willing/able to deny a fighter points based on the quality of their attack.
• The overall BpE has been largely unaffected by changing the point values of targets.
• Efforts to increase the attractiveness of the head as a target have not been successful, for reasons that are not completely understood.

Appendix A: Full Data

Targeting – Pre 2018

Targeting – SoCal 2018

Exchange Type