AG Open 2023 Stats – Point Values vs Targets Hit

I haven’t posted my usual Friday article for the past 2 weeks. Two Fridays ago it was because I was at the AG Open event in Plymouth Michigan to compete, ref, and lecture. Last week I just forgot. But now we’re back, and I’m going to post something related to the event.

Something that I have theorized over the past year or so is that awarding more points for a target leads to that target getting hit less overall. I have a huge article about this which I will post at some point, and link here when I do. There are a couple of reasons that this could be. Mathematically, if you have a point cap, the more high value targets you hit, the faster you will reach that point cap, and therefore you will have fewer opportunities to hit high value targets, and people who tend to get to the point cap by hitting low value targets will be over represented. There could also be psychological issues at play, like if my deep targets are more valuable, I might be more inclined to try to defend them, and therefore more hesitant to attack, and more likely to open lower value targets, since getting hit in a high value target could be catastrophic for me. Regardless of the reason, I have done an experiment and also observed a few different cases, and I think it generally holds up. 

This year’s AG Open hosted by Ars Gladii in Plymouth Michigan presented a good opportunity to test this theory. The rules were almost entirely the same as last year, except deep targets were worth 2 instead of 3. Because of this, it should be simple enough to go to the tournament stats and observe that the ratio of deep hits to shallow hits was higher this year than last year. Unfortunately for me this was not the case, AGO 2022 saw across the board 61% 1 point hits, 39% 3 point hits for LS, while AGO 2023 saw 62% 1 point for tier C, 59% for tier B, 60% for tier A, 61% for women’s+ tier B, and 63% for women’s+ tier A. So overall, the ratios ended up being about the same. Does this mean my theory is wrong? It doesn’t look good, but I’m not ready to give up on it just yet. 

Although it was the same tournament at the same place, it did not have all the same people. This year saw 113 fencers across 3 tiers of open longsword, while last year had 93 across 2 tiers. Although tier A was about the same size, the fencers in it were completely different, and I didn’t fence anyone this year that I fenced last year. Luckily there is an event that kept the same rules as last year, which a large sample of each year’s competitors participated in: the team longsword relay. In both 2022 and 2023, the longsword relay valued deep targets at 2 and shallow at 1, and awards both sides their score in the case of a double. If we compare the deep to shallow ratios from 2022 to 2023, we can see that in 2022 there were 53% shallow targets hit during the team relay, and in 2023 there were 58% shallow targets hit. This increase of 5% suggests to me that the general population of the tournament preferred hitting shallow targets this year, which would account for the open longsword tournaments being about the same between the two years despite the point value differences. 

In addition, there were a couple of other quirks in the stats when comparing this year to last year. This year the sense on the floor was that matches were tending to go longer, and this culminated in some delays. This can definitely be partially attributed to the point value changes, as it takes longer to reach 7 if you don’t have the option of a 3 point deep target. However, my anecdotal experience was that matches seemed to drag when they included a lot of “shallow hit with afterblow” exchanges, which halted the match and resulted in no net score. Looking at the stats, there is an increase in the amount of bilaterals (doubles and afterblows) per exchange between this year and last year. This year non-relay longsword tournaments had an average of 27.4% bilaterals per exchange, compared to 24.3% last year. Relay is not included in that calculation because since both teams gain points on a double, it does not slow down the match (it might even make it faster). Additionally, since deep targets were 3 last year, a deep hit with afterblow still gave 2, which gave them a lot more mileage in terms of getting to the point cap. 

Finally, we can look at the single handed tournaments. Last year it was a “one handed medieval sword” tournament and this year it was a messer tournament, but for all intents and purposes I am going to consider them to be similar enough to be comparable, and I will call them both messer for brevity. Last year, the messer tournament had the same ruleset as the longsword tournaments, 3 points deep 1 point shallow. This year all targets were worth 1, but there was a mechanic called the “held attack”, which subtracted 2 points from the opponent in addition to giving the hitter the point, and was awarded when the messer was held on the target that it hit throughout the full afterblow window without the attacker being hit. Although a held attack could be awarded for any target, it implicitly afforded deep attacks, because it is much easier to score a held attack on a deep target. We are also given a special treat with this year’s messer tournament, it is a tournament in which all targets are worth 1, but target area is still recorded, which means we can dig into how often deep and shallow targets are hit even when everything is worth 1. We usually don’t get this, because tournaments in which all targets have equal value are uncommon, and when they do happen they tend not to record target area because there’s no reason to.

The result was in AG 2022 messer, there were 62% shallow targets, and in AG 2023 messer there were 59% shallow targets (tier A and tier B were about the same). While the drop is not huge, I think it shows at the very least that awarding equal points for all targets does not devolve the game into a hand snipe fest. The held attack did exist, which was an implicit incentive to go for deep targets; if you think this led to more deep target hits, then I would say at the very least that proves that you can have a working incentive structure other than simply awarding more points for a deep target. However, anecdotally, I’m not sure how much effect the held attack had on individuals’ fencing. When we practiced the ruleset I found it hard to force it to work, and while judging I observed competitors fencing in a relatively normal way. If anything I think it affords more grappling, which may have occurred, but there are no stats on that. 

Overall, I still think my ideas about weighted targets still hold water. The point that it’s trying to make is about incentive structure. I’m trying to challenge the logical assumption that awarding more points for a deeper hit will lead to more of them happening because they provide an incentive for doing so. In order for this to be hard disproven, we would need to see an increase in shallow targets as the value of deep targets decrease, given an equal or similar population. That is not what we observed, as the amount of shallow targets was about the same, and the population tended to prefer shallow targets more than last year given the same ruleset in longsword relay.