Fun with Flicks

[cowpaste]

As I had nothing better to do, I wrote I simple computer program that determines the amount of time a foil tip is depressed. I ended up using an old printer cable and an old body cord to interface a foil with my computer through the parallel port. Then I proceeded to flick random things in my room 100 times. This was to try and simulate various flicks in a real bout situation. I tried my best to keep my flicking strength constant, but I did not attempt to perform "good" flicks all the time. By a "good" flick, I mean a flick where the point is almost 90 degrees from the surface of the target. Refering to the file "tips.jpg" attached below, case a.) is what I consider a "good" flick while case b.) is what I consider a "bad" flick. "Good" and "bad" are probably not the best ways to describe the final tip positions, but I do not give a damn.

The file "graph.jpg" shows a graph of the results. The average tip depression time was 13.35 [ms]. Assuming that 15 [ms] is the depression time required for the flick to register, 32% of the flicks would have registered.

Of all the flicks I performed (much more than 100), I noticed some happy, fun trends:
- "Good" flicks had a very low chance of registering (under 15[ms]), while "bad" flicks depressed the tip for over 20[ms] most of the time.
- I did not notice a significant difference between performing a "good" flick lightly and a "good" flick forcefully. The forceful ones would just be quicker and harder to parry.
- Flicking "badly" and forcefully was advantageous however, primarly because I had trouble bending the tip in the "bad" way with such little force.

Conclusion? If you want to flick, you can stilll do it by wacking your opponent really hard with a bad flick.

[ReverseLunge]

I'm going to look for your *** in the pools and vote for you right now! And a rep point.

[cowpaste]

Woop! Thanks.

[mrbiggs]

Originally Posted by ReverseLunge

I'm going to look for your *** in the pools and vote for you right now! And a rep point.

w00t!

Way to go, cowpaste! Well done!

Next thing you need to do is straight attacks, and determine the timing necessary to ensure that they register. Then email the results to the FIE.

(I gave you rep too recently. Remind me to give one if I forget)

[cowpaste]

Actually, I already have some information on thrusts. In general, thrusts are harder to test repeatedly in the same manner. There are just many more factors than with flicks: arm retraction speed (hard to keep constant), deepness of the thrust, "bouncing" off of the target, blade bending too fast, grounding of the blade from tip screws (cannot test with this setup).

If the target was directly facing me, the quickest thrust I could perform ended up being around 35[ms]. These would always register of course (assuming no grounding issues occur). What bothered me the most were the glancing blows. I tested this by turning a padded chair sideways (to simulate an opponent's chest) and thrusted in such a manner that the tip would click and slide off. These types of tip depression tended to last 7 or 8 [ms], ie. they would all register on the old timings; however, NONE of them were more than 15 [ms] and usually ended up being shorter than flicks!!! (Well, that is not totally true. You can fence ultra slow and make ultra slow "glancing" shots that hardly glance) Therefore, a glancing blow will pretty much NEVER register. This is what makes manplates so powerful, as they greatly increase the chances of glancing blows. Also, flicking harder surfaces tends to make the tip depress for shorter amounts of time. For example, flicks to my table depressed the tip for much shorter amounts of time than flicks to my bed, so wear a manplate in lefty-righty bouts to help prevent flicks to the chest (duh :P ). To give ya'll some numbers, I draped a plastron over a table and flicked it a couple of times. I was never able to get a tip depression longer than 15[ms] with a "good" flick. With a strong bad flick, the longest I was able to get was 15[ms] exactly. I added more layers of cloth on top of the table, which made it easier and easier to get flicks over 15[ms] with "bad" flicks.

Everyone knows this, but I'll restate: wear a manplate and you r teh winner!

[LUDICROUS]

Cowpaste Is Satan!!! 666 Posts!!! Yeah! Iron Maiden!

[oso97]

Holy crap.

Seriously, you need to set up a controlled situation with a proper set of targets, etc. Maybe one of the senior armourers, someone like Dan Duchuane would be interested.

[mackillian]

That's amazing.

[noahz]

Originally Posted by cowpaste

Conclusion? If you want to flick, you can stilll do it by wacking your opponent really hard with a bad flick.

With all seriousness, you should consider submitting this to American Fencing or perhaps the FIE fencing magazine...

[LUDICROUS]

All hail Satan (cowpaste), lord of the fencing.net underworld!!!

but NICE show on the flicking business. Send it to a magazine, people would love to hear about that.

[Alain]

Okay, I admit that some people have toooo much time on their hands... but THAT was pretty cool!

I think you should get it published, too! Good show, rep points awarded!

[Larrison]

Now this is cool!!!!!! I'd love to see a better description of the experiment and the code used for capturing the data...

Finally, some real data to argue from!

Did you have a chance to plot the data from the flicks to a plastron? or the "straight" lunges?

Rep points for sure!

[Mergs]

Excellent work! Have you done any analysis on your data to find out a mean and standard deviation and correlate it to 'good' and 'bad' flicks?

Interesting information about the glancing shots vice solid thrusts.

[cowpaste]

There are a few major problems with submitting this information to anybody important. First of all, there are many, many ways to depress the point. There are numerous types of flicks and numerous types of thrusts. I would have to categorize each type of point depression and be able to simulate them consistently. Even with a simple flick, there are so many factors involved ranging from angle of wrist, finger strength, arm strength, wrist strength..."technique" in other words. It would be hard to simulate this stuff with some mechanical arm.

Basically I would like to try and see if a scoring box could ever tell the difference between and flick and a thrust based on tip depression time. So far, results indicate no. This is the type of information that would actually be useful. Someone could use the results to try and convince the FIE that they should try to eliminate the flick in a different manner.

There is also a slight problem with my computer program. Namely, I do not know how accurate it is. The resolution of the timer I use in the program is 1 [ms]. This is actually precise enough for my purposes, but I still know nothing about the accuracy of the timer. If I have background programs running or if there is an ultra minor power surge, does the timer "hiccup" and slow down for a bit? Also, as the foil tip is a mechanical switch, I had some minor problems with switch bouncing. It would be nice to know if or how real scoring machines combat this.

For those of you who are interested, I wrote the program in Visual Basic. I think I would like to transfer it to Matlab, as that is a much more technically inclined language/tool. However, I would need to learn how to read off the parallel port in real time with Matlab.

[jBirch]

If you're using VB, there are a couple of high resolution timers you can use that go sub 1 micro second. QueryPerformanceCounter will get you less then 1 micro second and you can read RDTSC right off the chip if you are comfortable using assembly. The timeslice on Windows is 15ms to 30ms so you are right to be concerned that you may be being preempted. However, you're trapping hardware interupts (I assume) and just timing the start and end of the signal. It's highly unlikely that the GetTickCount call you're using is going to be preempted between its invocation and its return. If it were, that would probably account for only a very small portion of the variance you saw in your test. The way to compensate for that is to run the test many times and take the average, as you've done. That oughtta fix your resolution problem.

Lot's of stuff you can do with this data. It's mere existance is useful. While you're right, you can't submit it as the be-all and end-all, one thing you should realise is that the gross results show interesting things. Basically, the 15ms time is necessary to wipe out "good" flicks but at the same time wipes out angulated direct attacks. What would be interesting to capture as well would be the effects of different tip strength and traversal distances on the same flick. Find out what correlates most strongly to the flick from those three parameters and then see if you can control for it without any other ill effects.

At any rate, AWESOME work. You should definitely submit to ./ and see if you can bring F.NET down for a bit.

[Gav]

This is good as a preliminary set of tests however I think that the method is flawed. It's too artificial (as you note).

Here are a couple of suggestions:

Set up a couple of boxes (old and new timings) to compare.
Hook your test equipment into the signal line and take similair readings in bout situations.
Get your self a good video camera and film [u]everything[/i]. The more footage from more angles the better
Make sure that all equipment that you use has been properly vetted for faults.

I'm sure your bigwigs at the USFA would be interested in such a set of tests.

If you are studying for a sports science degree I am even sure that some credits could be obtained by such rigorous testing.

[shlepzig]

Very Cool work indeed.

This agrees with my deconstruction of why I was not landing straight attacks under the new times (in the Italian Criticism thread, link to post )

I personally doubt that tip bouncing has much to do with anything, and is just a preliminary theory. I suspect the controlling factor is going to be in the surface properties of what you are testing against.

If you simulated the Lame covering over your chair test for straight thrusts to recreate the mechanical properties of a fencers surface. Or get someone to put on all the kit and stand in as a test dummy for your testing, to properly simulate the surface of a human fencer with a lame.

Your work capturing actual depression times provides an excellent tool to determining whether the contact time change is the right tack on the problem of flicks, whether it needs revision or rejection.

Shlep.

[Mr Epee]

Always a fan of life by the numbers...

Sounds like you are getting a pretty good handle on the tech side of this experiment.

Have you tried with other fencers to see if there are any distinct variances that are dependent on the individual fencer's technique?

[Robert (2)]

Originally Posted by cowpaste

First of all, there are many, many ways to depress the point.

Leaving aside the considerable problems with the data, could you perform an additional test. Hit a direct thrust into a solid flat target (such as a wall) and drive through (so that the blade becomes flat against the target and the point pops back out again), in other words a crude simulation of a hit from too close against a chest plate.

It would be useful to see the spread of times you get for this, both for simple arm extensions, and also for lunges.

All very intrigueing so far.

Robert

[Larrison]

I was thinking about this experiment too..

I think one of the issues with the recent FIE ruling about male chest protectors and needing them to be worn directly against the skin (in comparision to wearing them over a t-shirt) -- is there is no real data to argue from regarding this ruling. What's being argued are "impressions" and "feelings" about what is happening under the new timings, and NOT from real data.

Someone could take this experiment apparatus and test for tip "bounce" on a hard surface impact -- first against a flat smooth surface (such as painted, plastered wall, or metal sheet) and look to see what the tip impact time was. Do this for a flick and a straight attack and look for a difference -- I bet you see one.

Then do this with a lame over a protector over a jacket against the smooth flat surface, to see what is done for "bounce" on a hard underlying surface.

Then do the same, without the protector.

If possible, it would be good to do this experiment with someone wearing the setup -- but a hard flat surface would probably be more repeatable.

I bet you could gather enough data in an afternoon to really make a case for or against the recent FIE ruling...