Quick Pre-Practice Post

I haven't written in my swordblog in over two weeks. SADNESS!

This post is gonna jump around a bit, because I've been having several different swordthoughts and haven't had as much time to try them out as usual.

*****

So, the current thing I'm working is related to the previous post. Last post, I talked about high guards which defend the high line, and low guards which defend the low line. Then, I went into excruciating detail about why it is important to flip back and forth between high and low guards. The summary is that doing this "automatically catches" your opponent's attempts to disengage. This is a vast over-simplification, but the complex version is in the damn post.

So, over the past few weeks, I've been working on something that I wanted to be the "next step" in that process. You see, when you have less range, and you're fighting against someone with more range, you generally need to close 3 or 4 steps, rather than one step of covering and a second step of attack. As such, I tried to develop a third type of atajo.

First, we have the "weak" atajos, in which we put our tip near their hilt. These are "near the tip of our blade".

Then, we have the "strong" atajos, in which we cross the middle of our blade over the middle of their blade. These are "near the middle of our blade".

I then wanted to create something that was "near the tip of our opponent's blade", as a third sort of atajo. Sadly, every attempt I have made at this has failed miserably. It's possible that I am just not doing it correctly, but it seems mostly like covering your opponent's tip just isn't a viable tactic.

In theory, you would need to have a wider "angle" - weak atajos have almost no angle, and strong atajos have a 45º angle, so following that progression means that we would have something like a 90º angle with our opponent's sword, to cover more space.

When I was trying this out last Monday, I was not doing the 90º angle, so perhaps that is wrong.

In terms of historical sources, Romagnan's Rada treatise talks about "virtual" atajos, in which you place an atajo on a sword which is far away from you. It seems like this might be similar to the concept of "covering their tip", just because of how far away you are. But I'm unsure.

Oh well. Further work is required.

*****

Another possible solution here would be to just go for the strong atajo, then switch to a weak atajo and then the opposing weak atajo. Switching between strong atajos takes far too long, but switching between weak atajos is a tiny action. This worked pretty well against Thomas's longsword, in as much as it allowed me to take that third step while closing lines, but I'm not sure how it would work against other opponents.

This solution vexes me because it forces me into a weak atajo. If someone is using case, I need to be able to switch between strong atajos in order to take advantage of the relatively narrow opportunities that they allow me.

*****

A possible solution to the "switching strong atajos takes too long" problem is a possible alternate way to do strong atajos. In the traditional Narrowing/Line in Cross, we block the high line with Line in Cross and the low line with Narrowing. In theory, if we lower or raise our sword more, we could block the low line with Line in Cross, or the high line with Narrowing.

The idea behind this is the "sword-punch" thing that Sorcha used to do far more often. Basically, she applies Line in Cross while stepping in a circle, then pushes hard downwards with her hilt in order to switch from blocking the high line to blocking the low line. This, exacerbated by the truly brutal quillons that her old sword has, prevented the disengage that would normally be the correct response to the high-line-defending Line in Cross that she, in her Italian-esque way, performed.

The drawback to this is that you are no longer actually protecting the high line - by the principles of opposition, your opponent should be able to push through your sword to get at your high line, if they anticipate your action. However, it does still provide some small protection for the high line if your opponent is not trained in the art of Opposition, and it also creates the illusion that you are, in fact, protected.

Similarly, you could take Narrowing and "punch upwards" with it to cover your high line.

The "punch downwards" is something I reflexively do already as part of my Line in Cross, if I feel my opponent gaining a tempo-based advantage. I lean over and punch downwards to catch their disengage. If I integrated a switch from that atajo to a weak atajo, I could possibly continue forwards?

*****

Hmm hmm hmm. Vexing questions.

Next time, I'll probably try to write about "Directions of Stepping And The Attacks You Should Apply While Stepping In Them". Until then, fence well!

Principles of Attacking and Excruciatingly-in-Depth Tempo

Sorcha's home with recently-removed wisdom teeth, I'm stressed about the election, and I wasn't able to go to fencing practice yesterday. So what do I do? I write up a blog post.

Today's blog post is about attacking. An ideal attack has two properties:

• Presents a legitimate threat to your opponent.
• Creates a strong defense against your opponent.

Those are both very squishy definitions, though. What do we mean by "legitimate threat"? What do we mean by "strong defense"? How do we prioritize these things? All very difficult questions to answer. I would propose the following modifications to those definitions, to make them more grounded and specific:

• Presents a legitimate threat to your opponent.
• Will stab your opponent if they don't act within one tempo.
• Creates a strong defense against your opponent.
• Creates a defense such that attacking you is a parryable action.

I concentrate on "one tempo" as the minimum measurement of action and reaction. This, as I have detailed before, is because human reaction speed is about 200 milliseconds at best. This minimum measurement of time gives us boundaries to what can and cannot be done. If one person is fencing at 1/4 speed and the other person is fencing at full-speed, there are all kinds of things that the full-speed person can do safely. However, when both people are fencing at full speed, this forces us to recognize that there is only so fast that each player in this game can act and react, due to the speed of electrical impulses in nerves.

Offensive Principles

So, let's tackle these definitions individually. Again, here's the first:

• Will stab your opponent if they don't act within one tempo.

That is simple enough. Leaning thrusts are one-tempo actions. But we can extrapolate that out a bit. If our opponent is badly positioned, it isn't a stretch to imagine that they wouldn't be able to bring their blade to bear in time. Let's, in a completely arbitrary and anecdotally-supported way, call that a 1.5-tempi action. So, the above statement expands into the following:

• Either
• You have a clear leaning-thrust you can perform, without being immediately blocked.
• You can continue blocking your opponent's sword into a disadvantageous position while you lunge.
• You can push your opponent's sword into a disadvantageous position while you lunge.

This still leaves a bit of vagueness - what is a "disadvantageous position", for example? To me, tt's a position where they can't get their sword to a place where they can defend themselves from your attack in 1.5 tempi. This is different for different attacks - for example, attacking the arm is much faster than attacking the flank, which is a bit faster than attacking the face.

Defensive Principles and Excruciatingly-in-Depth Tempo

So, with that vagueness as out-of-the-way as we're going to get it, let's continue on:

• Creates a defense such that attacking you is a parryable action.

I'm having some trouble sorting out my sources (this one being the closest I can find to what I wanted), but the processing time for an action relying on visual stimuli is as follows:

• Eye sees the image, transmits it to brain (20-40 milliseconds)
• Brain processes the image, decides on an action (50-80 milliseconds)
• Nerves convey impulse to arm, arm moves (80 milliseconds)
• This last one was pulled from my memory. Not sure if it's correct. If you find actually-correct numbers, please let me know.

This yields about 200 milliseconds plus some amount of actual movement time for a full reaction, as above. But what this means is that by the time your brain has finished processing what has happened, your opponent has moved from that point, whether they are continuing their current action or starting a new action. Looking above, the time it takes to process data is about 100 milliseconds - half a tempo.

So, this means that all of our defensive actions must assume that our opponent has half a tempo more of action than they have performed yet. This means that all of our parries and protective actions must be, to a certain degree, proactive.

It takes a certain amount of time for a blade to accelerate from nothing to moving. I would wager it is about 100 milliseconds. (I don't have real data on this bit, so I could easily be wrong here.) 

Having your opponent's blade blocked to a place where they cannot thrust straight-in means that they cannot be that half-tempo through your blade, because it's there and blocking them. However, this means that you must assume that they are already a half-tempo through performing the disengage they need to hit you in your openings.

Henceforth, I will be measuring things in terms of tempo. 1T is one tempo, which is 200 milliseconds. 0.5T is half a tempo, which is 100ms.

Let's pretend, for the moment, that you have them blocked onto your inside line. Their only options are to attack from the high inside line and low inside line, from how your stance is structured. The options you must consider collapse into the following, if you wish to defend against all attacks:

1. Opponent attacks into the high inside line
• 0T: They cannot thrust because your sword is in the way.
• 0.5T: Your sword moves and they happen to have already chosen that moment to thrust. Their thrust proceeds.
• 1T: Their thrust is moving, your sword is moving.
• 1.5T: Their thrust strikes home.
• 2T: If you attempted to parry as soon as you saw them move, your parry would complete here.
2. Opponent attacks into the low inside line.
• 0T: They begin thrusting.
• 0.5T: Your sword begins to move. They are still thrusting.
• 1T: Their thrust strikes home.
• 1.5T: If you were not attacking and you attempted to parry as soon as your brain understood that they were moving at 0.5T, your parry would complete here.

So, the ideal attack would defend against the low line attack primarily, while still creating enough of an obstruction on the high line that your opponent must perform at least 0.5T of disengage.

We can collapse our definition of what a well-defended attack is further:

• If your opponent attacks you through a place that your stance currently defends and will after you attack, their attack will fail, then your attack is well-defended.
• If your opponent attacks you through a place that your stance currently does not defend but will after you attack, it will take them at least 1T to strike you, then your attack is well-defended.
• If your opponent attacks you through a place that your stance currently defends but will not after you attack, it will take them at least 1.5T to strike you, then your attack is well-defended.
• If your opponent attacks you through a place that your stance currently does not defend and will not after you attack, it will take them at least 2T to strike you, then your attack is well-defended.

There are two more items on that chart than in the above timing chart. The very first item should be obvious - if you attack in a way that defends you, then you should remain defended. The last item is a bit more complex - the math-esque thinking there is as follows:

• Location that your stance does not defend and will not after you attack.
• 0T: They begin thrusting.
• 0.5T: Your sword begins moving. You realize that they are thrusting. 
• 1T: Their thrust lands. The signal from your brain to your arm that it should halt its attack and defend you reaches your arm.
• 1.5T: Your arm succeeds at stopping your attack and begins to defend.
• 2T: Your parry finishes enough that, had it been 1T faster, you would not be stabbed.

As such, these four principles should be enough to cover all cases, when thinking of swords in terms of lines and openings.

In theory, any system that adheres to these principles will be a Good and Correct system. The rest of this post goes in-depth about how I, personally, implement these principles in my own fencing. Some amount of this is conjecture, because I have not fenced enough recently.

A Brief Digression

The reason I collapse this into high/low is because the Spanish tradition has four lines:

• High Inside
• Low Inside
• High Outside
• Low Outside

According to the Spanish tradition (per Romagnan's Rada manual), you keep people's swords to either the left or right of you - also called the "left defensive plane" and "right defensive plane". These are strongly defensive positions. The quickest openings that you can be attacked on are openings in which your opponent switches from high to low - as such, you only have to consider the transition from high line to low line when defending yourself, so long as your stance remains Right and Proper.

As such, when attacking, switching from the high line to the low line is the ideal action. This is expressed in Viedma's "High General" and "Low General". The "Low General" is a switch from a strong, high-line parry (Line in Cross) to a weak, low-line parry (Weak Under Strong). Similarly, the "High General" is a switch from a strong, low-line parry (Narrowing) to a weak, high-line parry (Weak Over Strong). In my own fencing, I generalize that to a principle that switching from high-line to low-line is inherently the safest way to attack.

So, the Spanish tradition (as I understand it) is structured to create one "defended" line in exchange for creating exactly one "open" line, when in-stance. The "open lines" of Italian traditions are a bit more abstract - there's "inside" and "outside", but aside from techniques that clearly make use of high and low lines, the Italian masters don't really consider things like that. In fact, I would go so far as to say that Capoferro and Giganti don't really consider the low line at all, preferring to keep the idea of "opening" more abstract and "here's a set of examples, now go to town". While there is certainly clear value in this, it doesn't jibe with the way I personally work.

Conclusion

In summary, we have the following principles:

• Presents a legitimate threat to your opponent.
• Will stab your opponent if they don't act within one tempo.
• You have a clear 1T attack such as leaning-thrust you can perform, without being immediately blocked.
• You can continue blocking your opponent's sword into a disadvantageous position while you perform a 1.5T attack such as a lunge.
• You can push your opponent's sword into a disadvantageous position while you perform a 1.5T attack such as a lunge.
• Creates a strong defense against your opponent.
• Creates a defense such that attacking you is a parryable action, meaning that your attack is well-defended. The circumstances which produce well-defended attacks are as follows:
• If your opponent attacks you through a place that your stance currently defends and will after you attack, their attack will fail.
• If your opponent attacks you through a place that your stance currently does not defend but will after you attack, it will take them at least 1T to strike you.
• If your opponent attacks you through a place that your stance currently defends but will not after you attack, it will take them at least 1.5T to strike you.
• If your opponent attacks you through a place that your stance currently does not defend and will not after you attack, it will take them at least 2T to strike you.

If we fight in a way that abides by these principles, attacking will not present an opening. So, doing these things allows us to be As Safe As Possible™. A future, less exciting blog post will cover stances and how to remain safe in-stance. I might have already blogged about this, but my ideas on the subject have most likely changed since last I wrote on it, so it bears repeating. Next post will most likely be an in-detail look at how I implement these principles in my personal mix of Italian and Spanish fencing.

(Edited for comprehension on 2018-03-26)

The Pokétron and the Overtron

I am pleased with myself. Before you continue - this post has very little if anything to do with fencing.

Last night, I re-implemented a tool for taking an arbitrary "team" and giving advice about the best choices to minimize the weaknesses of the team. This is because I have been playing a lot of Overwatch. One problem in Overwatch is that sometimes, the character you select is just bad against the other team's composition. However - it's hard to tell the difference between performing badly and being counter-picked. So, we solve the problem with technology.

Originally, the tool was created for fighting the Pokémons online. I naively scraped "weak against" data from Smogon University, and then used that data to create teams for whom there were no un-compensated-for weaknesses.

An example. So, I have added Swampert to my team. Swampert is weak to Celebi and Mamoswine. Celebi is weak to Skarmory and Venusaur, while Mamoswine is weak to Skarmory and Bronzong. We're going to pretend that there are no other weaknesses, for the moment.

So, the program would advise that I use Skarmory, since Skarmory counters both and doesn't add any new weaknesses. Yay!

If the options listed did add weaknesses, then I would check to see if they were weak to anyone currently on the team. If they were, that would be fine. If they weren't, then they would be a less-good option according to the program.

In reality, this relies on having a good, simple data-set. It's a very useful program for Pokémon, for instance, because the data-set is so large and there is a website sitting there with ready, scrapeable data. It's all simple sets though, and didn't take *that* long to implement.

The reason I've done this recently, is because Overwatch has a similar extended-rock-paper-scissors element to it. So, I've implemented the Overtron. What I do is I input my team, then it tells me what the best choices are, in ascending order of awesomeness.

There's some weirdness to the data. In the Pokétron, there are a few Pokémon who, according to Smogon University's dataset, are so dominating of the meta-game that they were almost always the best choice. Skarmory and Scizor were the two particular ones, with Skarmory frequently being between 400% better than the next-best option, and then the next several options being only about 20% different from each other.

It's hard to tell if this is because the data is generated by humans, or if Skarmory was actually that dominating of the meta-game.

Similarly, the data-set I'm using for Overwatch friggin' loves Reaper and Winston. Loves 'em. The brief testing I've undertaken has allowed me to see that yes, they are good options very much of the time. But it's still hilarious to me how this algorithm has such dominant "favorites". It's probably a sign that I either need to modify the algorithm, or I need to edit the data-set.

Part of me believes that this sort of set manipulation could be used to determine effective maneuvers. Because fencing does have a subtle sort of rock-paper-scissors to it, in various maneuvers. Sadly, the data-set is probably too squishy for me to ever create the Fencingtron.