#vasher does programming

[vash3r]

reminder that i am streaming Advent of Code every day this december (when it releases at midnight EST / 5am UTC). On day 1 I placed on the global leaderboard (currently 40th place with 114 global score :)

[cosmerelists]

Cosmere Protagonists Who Would Support the Villain...of Other Cosmere Planets

As requested by @asteroidfieldgame :)

Sure, in their own stories the villains are not well-liked by the majority of the cast. But if we took those villains and had them meet protagonists of other Cosmere novels...well, then it might be different, mightn't it?

[Here there be spoilers! I would skip any entry if you're not caught up on all the books for both characters involved!]

1. Moash: Supported by Kelsier

Kelsier: [holding Moash protectively ala that one meme with Kevin Hart] Kelsier: My boy Moash has LITERALLY never done anything wrong ever! Kelsier: Ooooh, did he kill his oppressors? Is he mean to poor helpless nobility? Did someone Rich experience a Consequence? Moash: (mumbling) I did try to drive my friend to suicide. Kelsier: Shh...baby boy, it's okay! Kelsier: I made my disciple advocate for old people to kill themselves with spikes to preserve their power! Kelsier: You're literally fine.

2. Odium: Supported by Harmony

Harmony: Listen, I'm not saying I agree with everything Odium does. Harmony: But making it so that you have a planet of mighty, battle-hardened residents to prepare for the coming Cosmere-wide conflict? Harmony: Sometimes I wonder if I should have done something more like that. Harmony: Instead of giving everyone an easy life and making giraffes because giraffes are neat. Odium: Have you considered...battle giraffes? Harmony: Please don't patronize me.

3. Hrathen: Supported by Ellista (that one Ardent who was really into romance novels)

Ellista: I am not immune to a handsome man in bloodred armor with a troubled past whose hard, mean exterior is slowly worn away through love. Ellista: A man who chooses love--unrequited love!--over his own god! Ellista: Problematic for an Ardent like me, but so, so compelling. Ellista: (sighs dreamily) Hrathen: ... Hrathen: Could I have someone else's support please?

4. Riina [from Tress]: Supported by Wayne

Wayne (counting off reasons on his fingers): One, if you're gonna be the sort to be giving out curses, giving out breakable curses makes you less of an unforgivable villain and more of a garden-variety dick. Wayne: Two, rats are cute. Makin' someone a rat is better 'n' making them a grub or something. Wayne: Three, and most importantly, her penis spaceship is hilarious. Riina: IT WAS NOT A PENIS SPACESHIP

5. The Lord Ruler: Supported by Dalinar

Dalinar: I suppose I am simply more aware than most that being a king is hard and can involve less-than-ideal-choices. Dalinar: And didn't he essentially become immortal while trying to cage an evil god and save his planet? I am literally setting myself up to maybe experience that as we speak. Dalinar: I know he took an entire population and made them slaves, but I did tell Jasnah not to free our slaves so, like... Dalinar: ... Dalinar: Maybe Wit was right. Maybe I AM a tyrant. Wit: You've never done an enforced breeding program with humans, though. That I know of. The Lord Ruler did that! Like, a lot! Dalinar: ... Dalinar: Wait I want to mitigate my support a little.

6. Taravangian: Supported by Shai

Shai: That Diagram of his...well, it's pure art, isn't it? Shai: I'm not immune to the appeal of incredibly complex and well-researched plans turned into an artistic representation so beautiful that it attracts worshippers. Taravangian: It's really more science than it is art, I'd say. Shai: Hilarious that you think those things are different.

7. Nale: Supported by Marasi

Marasi: He knows the law and follows the law. Marasi: Don't think it would ever occur to him that someone could be quote unquote "above" the law. Marasi: That makes him better than, like, quite a lot of my coworkers.

8. The Machine [from Yumi and the Nightmare Painter]: Supported by Vasher

Vasher: Obviously a soul-eating machine that nearly destroys a planet and turns people into nightmares is not ideal. Vasher: But it's not the machine's fault it was given an ill-considered Command. Vasher: Hey! Bald guy! Wanna gentle-parent the machine too? Szeth: Uh

9. Raboniel: Supported by Elend

Elend: Not for nothing am I chair of the Support Women's Wrongs fanclub. Elend: Nor can I fail to support a Woman in Science. Basic feminism. Elend: And although I know my bar is on floor, a parent who will do anything to save their child from eternal suffering gets me right here. (points to his heart)

10. Straff: Supported by...no one

Straff: ... Straff: ... Straff: OH COME ON

[junker-town]

Texas Tech’s offense will be great, but the defense will probably waste it. Aaaaagain.

This is the 2017 edition of the Texas Tech preview.

Air raid guys get a bad rap when it comes to defense. Their propensity for tempo and passing mean that their opponents get lots of possessions and lots of snaps. That leads to allowing quite a few yards and points even if they have pretty good defenses.

Adjusting for tempo and opponent, plenty of guys on the Hal Mumme tree have produced decent defensive performances.

Mike Leach’s last Texas Tech defense ranked 27th in Def. S&P+. His Washington State defenses have improved for a couple of years in a row, growing from 101st in 2014 to 63rd in 2016.

Dana Holgorsen’s West Virginia defenses have been top-50 for three straight years and top-40 for the last two.

Art Briles’ 2013 Baylor defense ranked 26th.

Sonny Dykes’ 2011 Louisiana Tech defense ranked 25th.

Mark Mangino had Kansas defenses at 21st in 2005 and 35th in 2007.

In Seth Littrell’s first year as North Texas head coach in 2016, the Mean Green improved from 120th to 93rd.

These are anecdotes and not statistically significant proof. But it shows that having an up-tempo, potentially pass-heavy attack and a solid defense is possible. Adopting the air raid does not commit you to awful defense.

Someone needs to tell that to Kingsbury.

Four years ago, Texas Tech’s hire was one of the most natural you’ll ever see. A former Red Raider quarterback under first Spike Dykes, then Mike Leach, Kingsbury enjoyed some time in the pros before returning to college football and succeeding as an offensive coach. Houston ranked seventh in Off. S&P+ in 2011 with Kingsbury as co-coordinator, and when he followed Sumlin to Texas A&M in 2012, his first Aggie offense, led by Johnny Manziel, ranked second.

Kingsbury was only 33 when Tech named him head coach in December 2012. It hadn’t even been 10 years since he threw his last pass — against Clemson in a dominating Tangerine Bowl win — in Red Raider red and black. Still, this felt right. After Leach’s controversial firing in 2009, Tommy Tuberville had taken over, and while he engineered two eight-win seasons and plenty of yards and points, the fit just wasn’t there.

Thus far as head man, Kingsbury has proved himself a great offensive coach. After a first-year reset, the Red Raiders have ranked 19th, first, and sixth in Off. S&P+. And in that three-year span, they have lost four times while scoring at least 50 points and 12 times while scoring at least 30.

Tech’s Def. S&P+ ranking the last three years: 114th, 124th, and 125th. The Red Raiders are the exception that proves the rule — yes, you can play decent defense with this offensive system. But Tech can’t. Or at least, Tech hasn’t yet.

After going 8-5 in his first year (with his worst offense to date, no less), Kingsbury has since gone just 16-21. The 2016 season was his second bowl-free campaign in three years. His status as a Tech legend has earned him a little more rope than another coach might get, but he’s stretched that rope out pretty far. His offense will remain prolific despite the loss of first-round quarterback Pat Mahomes II, but he and defensive coordinator David Gibbs have to rebuild their defensive line and hope injury was the primary cause for Tech’s dreadful pass defense.

There were a lot of injuries to deal with, for what it’s worth. Twelve different defensive backs made at least 6.5 tackles, and three of them played in all 12 games. The shuffling up front was nearly as bad. Even Nick Saban needs some semblance of continuity; Tech’s 2016 defense never had a chance.

Still, for most of three years the Tech defense has been the unfair air raid stereotype. And as magical as Kingsbury’s offensive touch is — again, I’m simply assuming another top-15 offense despite the Red Raiders losing an incredible quarterback, plus a nearly 1,200-yard receiver — he might be looking at a coordinator gig in 2018 if Tech can’t make a few stops per game.

2016 in review

2016 Texas Tech statistical profile.

If you’re looking for hope, maybe there’s this: Tech’s defense had three of its five best performances in the first five weeks, before attrition had set in.

First 5 games (3-2): Avg. percentile performance: 72% (~top 35) — 86% offense, 41% defense | Avg. score: Tech 55, Opp 39 (plus-16)

Last 7 games (2-5): Avg. percentile performance: 38% (~top 80) — 49% offense, 24% defense | Avg. yards per play: Opp 47, Tech 35 (minus-12)

The first five games did include frustrating shootout losses to Arizona State and Kansas State, but the KSU loss was fluky — Tech outgained the Wildcats by 257 yards and was only minus-1 in turnovers, but KSU scored two return touchdowns and won all four fourth downs (Tech was 0-for-3, KSU 1-for-1). Still, six weeks into the season, Tech was still a healthy 36th in S&P+. Six games later, the Red Raiders were 82nd following a 56-point loss to Iowa State.

Their final ranking bounced back to 66th thanks to a blowout of collapsing Baylor, but the last seven games saw five miserable defensive performances and a couple of offensive duds as well. Injuries played a role, but this was a disturbing look at how low the team’s floor is. Will good health lead to a longer look at the ceiling?

Offense

Full advanced stats glossary.

Even adjusting for tempo (and nagging injuries to Mahomes), Tech’s offense was ridiculous. Only four teams produced an adjusted scoring average (i.e. an Off. S&P+ rating) of 40-plus points per game in both 2015 and 2016: Cal, Clemson, Oklahoma, and Texas Tech. That’s two air raid programs (Cal was led by Sonny Dykes both years) and two that combined to go 50-6.

That Tech’s offense maintained most of its 2015 form was impressive, considering Mahomes had shoulder and wrist issues for much of the year. Mahomes was also asked to carry more of a load because of a complete restart in the run game.

The 2015 Tech attack featured 1,500-yard rusher DeAndre Washington behind a line that included three three-year starters; in 2016, freshman Da’Leon Ward led Tech backs with just 428 yards (4.2 per carry) behind a line that only had a couple of experienced pieces and had to start eight different guys at least once. Five of the eight were either freshmen or sophomores.

And Tech fell all the way from first to sixth in Off. S&P+.

Kingsbury’s track record makes it impossible to worry about points. In a limited sample size, Mahomes’ backup Nic Shimonek completed 66 percent of his passes and nearly matched Mahomes’ per-attempt averages. He wasn’t nearly as much of a mobility threat, but if he wins the starting job in 2017, he’ll probably be fine.

Of course, if junior Payne Sullins, JUCO transfer McLane Carter, or freshman Xavier Martin wins the job instead, he’ll probably be fine, too. But it’ll probably be Shimonek.

Michael C. Johnson-USA TODAY Sports

Nic Shimonek (16) and Demarcus Felton (27)

Leading receiver Jonathan Giles elected to transfer, which would be a source of concern for most offenses, but Giles was one of four Red Raiders targeted at least 80 times last year, and the other three are back.

Giles was impressive, mind you — he combined a 62 percent success rate with 16.8 yards per catch, a 97th-percentile performance overall. Still, the experience level at receiver is solid. Inside receivers Keke Coutee and Cameron Batson combined for 116 catches at 9.5 yards per target, Dylan Cantrell is a physical possession guy on the outside, and between senior Derrick Willies, sophomores Quan Shorts and Antoine Wesley, and redshirt freshman T.J. Vasher, odds are good that at least one more high-proficiency wideout will emerge.

Tech will be able to pass, and that should result in at least 30 points in most 2017 games. But to reach 2015 levels, the ground attack will be key. The line is in a far steadier place, especially if it finds itself less reliant on freshmen. But a back still needs to emerge.

It was easy to assume that guy would be Justin Stockton a year ago. Through his first two seasons, Stockton had averaged seven yards per carry with seven receiving touchdowns. But aside from a couple of long early receptions, he was woefully unproductive in 2016, and he spent the spring dealing with issues related to head injuries.

Da’Leon Ward, meanwhile, was efficient for a freshman but offered no explosiveness whatsoever; per carry, Demarcus Felton was the most productive of the backs, but 237 of his 354 yards came in a three-game span early in the season. Kingsbury has said that he’s looking for a potential graduate transfer to fill in the two-deep here. It’s not hard to see why, especially if Stockton remains a question mark.

Michael C. Johnson-USA TODAY Sports

Keke Coutee

Defense

New quarterback? Iffy running back? No matter. Tech will score points. Bowl hopes will again hinge on whether the defense can be at least semi-competent.

The Red Raiders were bad in every way a defense can be bad in 2016.

128th in plays of 10-plus yards (241, or 20 per game!)

128th in passing downs success rate (40.9 percent)

126th in success rate (49.4 percent)

126th in Adj. Sack Rate (32.9), 126th in passing downs sack rate (2.4 percent)

124th in plays of 30-plus yards (44, nearly four per game!)

123rd in rushing success rate (51.8 percent)

119th in standard downs success rate (52.5 percent)

118th in passing success rate (46.9 percent)

105th in points per scoring opportunity (4.92)

101st in stuff rate (16.7 percent)

Or, in chart form:

Tech had by far the worst success rate and IsoPPP average in the Big 12, a conference not known for defensive prowess.

Coordinator Gibbs is, on paper, the perfect air raid defensive coordinator. He found success as D.C. at both Minnesota and Auburn at times, and he calibrates his entire defense around forcing turnovers.

Well, that’s what he wants to do, anyway. He couldn’t do much of anything with Tech’s young personnel in 2016. Tech managed a paltry 13 takeaways (112th in FBS) and forced just nine fumbles (78th).

This went beyond injury; after all, the Red Raiders boasted two former blue-chip defensive tackles in Breiden Fehoko and Ondre Pipkins, they missed one game between then, and they still couldn’t stop the run to save their lives. Still:

Leading returning defensive end Gary Moore missed five games.

Leading returning linebacker Dakota Allen was kicked off the team in May (and then rejoined in December).

Second leading returning linebacker D’Vonta Hinton missed seven games.

Second leading returning safety Tevin Madison missed 10.

Safety Payton Hendrix missed six. Safety Keenon Ward missed three.

Cornerback Nigel Bethel II transferred before the season, and corners Paul Banks III and Desmon Smith missed a combined seven games.

And on and on. It was like coaching a brand new defense every week. And when you’re in a league full of high-tempo offenses adept at exploiting any crack, that’s deadly.

Photo by John Weast/Getty Images

Jah'Shawn Johnson

We don’t know if Kingsbury puts the right amount of emphasis on defense, and we don’t know if Gibbs’ turnovers-über-alles approach has merit. But continuity would help us find out.

It would allow us to figure out if a secondary that allowed a 151.6 passer rating (with 28 touchdowns to just five picks) was bad because of talent or inexperience. Six DBs with at least 13 tackles return, and five of them were either freshmen or sophomores last year. Safeties Jah’Shawn Johnson and Douglas Coleman III combined for 15 passes defensed and four tackles for loss, which might be a good sign of potential play-making ability, and at nearly four stars, junior Payton Hendrix was the closest thing to a star recruit the secondary has.

The cornerback situation is less clear, but between senior D.J. Polite-Bray, sophomore Dsmon Smith, and JUCO transfers Octavious Morgan and Jaylon Lane, there are options, I guess.

Allen’s return gives Tech a couple of potential play-makers at linebacker. He made six tackles for loss with three passes defensed as a redshirt freshman in 2015, and Jordyn Brooks had five and four, respectively, as a freshman last year. But they could be having to shed a lot of blockers if the line can’t do its job. The line was iffy last year with blue-chippers Pipkins and Fehoko; now both are gone. So is Kris Williams, the only player on the team to record more than one sack. Yikes.

I’m not particularly worried about the tackles. A foursome of 2016 backups — seniors Zach Barnes and Mychealon Thomas, sophomores Broderick Washington and Joseph Wallace — combined for 7.5 tackles for loss among 36 tackles, and Wallace had three in only eight games. The ends, though? Kolin Hill is the closest thing to a proven play-maker, and he had four TFLs and three breakups. Gibbs doesn’t mind a bend-don’t-break approach at times, but you need at least some attacking potential.

Photo by John Weast/Getty Images

Jordyn Brooks

Special Teams

Because Tech was good at finishing drives in the end zone, kicker Clayton Hatfield was only asked to take 14 field goals last year. But five of them were longer than 40 yards, and he missed only one. That’s an encouraging sign considering he might be asked to salvage a few more stalling drives this time around. (His five missed PATs are a red flag, however.)

Hatfield’s great per-kick average salvaged a No. 67 Special Teams S&P+ ranking for a unit that was iffy at punting (not that they punted much) and had no oomph in kick returns. There’s a good chance he’s relied on even more in 2017.

2017 outlook

2017 Schedule & Projection Factors

Date Opponent Proj. S&P+ Rk Proj. Margin Win Probability 2-Sep Eastern Washington NR 19.3 87% 16-Sep Arizona State 58 1.4 53% 23-Sep at Houston 49 -4.6 40% 30-Sep Oklahoma State 22 -8.0 32% 7-Oct at Kansas 107 10.9 74% 14-Oct at West Virginia 69 -0.9 48% 21-Oct Iowa State 57 1.3 53% 28-Oct at Oklahoma 5 -23.1 9% 4-Nov Kansas State 35 -2.7 44% 11-Nov at Baylor 28 -9.3 29% 18-Nov TCU 21 -8.2 32% 25-Nov at Texas 16 -15.0 19%

Projected S&P+ Rk 66 Proj. Off. / Def. Rk 4 / 124 Projected wins 5.2 Five-Year S&P+ Rk 3.6 (51) 2- and 5-Year Recruiting Rk 45 / 45 2016 TO Margin / Adj. TO Margin* -4 / -1.0 2016 TO Luck/Game -1.2 Returning Production (Off. / Def.) 61% (62%, 59%) 2016 Second-order wins (difference) 5.3 (-0.3)

S&P+ projects Tech fourth on offense and 124th on defense. The offense will be its typical awesome self (though maybe not quite as awesome — Giles’ transfer wasn’t part of the initial S&P+ projections), and the defense faces more burden of proof than any unit in the country.

I assume the secondary will come through, but it’s hard to paint an optimistic face on the defensive front. That means a lot more shootouts. We expect nothing less in Lubbock.

So can the Red Raiders win some more shootouts? S&P+ projects five relative tossups — Arizona State, at Houston, at WVU, Iowa State, and Kansas State are all between 40 and 53 percent win probability — and Tech will probably have to win at least three of those to bowl.

At some point, bowling might not be enough, but it would be a step forward. Kingsbury needs a few of those, but you can’t get a few without the first one.

Team preview stats

All preview data to date.

[vash3r]

@twocubes i kind of want to try make a mod for Balatro (poker roguelike) that implements your deck (extra suits, number cards, tarot, etc). I think there's enough info in #nouveau tarot nouveau that i'd mostly just be trying to figure out if it's possible in code. do you have any requests/suggestions?

it might be a bit of a hassle to figure out hand types for the added cards....

[vash3r]

yippee new program https://windhawk.net/mods/chrome-ui-tweaks

[vash3r]

i should have guessed that the effect i wanted to accomplish is fundamentally unsuitable for a fragment/pixel shader, because they are 'read-oriented' instead of 'write-oriented' (to avoid data races with the high parallelism)

... but maybe i can hack around it by just looping over all the potential sources... a for loop inside a shader...

[vash3r]

today i went down a rabbithole of starting to completely rewrite a mdns implementation to understand how it works. (probably should have actually read the whole spec before i did that huh)

[vash3r]

the wungler is gonna be something like 20 lines of code total btw. insane how easy this shit is. you can wungle a post in one line of javascript, the rest is css and making a usable UI

test wungle 󠁹󠁯󠁵󠀠󠁪󠁵󠁳󠁴󠀠󠁧󠁯󠁴󠀠󠁷󠁵󠁮󠁧󠁬󠁥󠁤󠀡󠀠󠁲󠁥󠁢󠁬󠁯󠁧󠀠󠁴󠁯󠀠󠁴󠁯󠁴󠁡󠁬󠁬󠁹󠀠󠁷󠁵󠁮󠁧󠁬󠁥󠀠󠁡󠁬󠁬󠀠󠁹󠁯󠁵󠁲󠀠󠁦󠁲󠁩󠁥󠁮󠁤󠁳󠀠󠀺󠀩post

this one has wungle content encoded in unicode Tag characters (between the words 'wungle' and 'post'), which are part of the main post content (and can be copy-pasted) but don't display glyphs. they can be inserted in the html editor with escape codes, but the plan is for the extension to encode/decode wungle text from the clipboard or something ala qaz.wtf. it only supports ascii though and right now i'm not up to implementing some kind of base64 for this

[vash3r]

custom puzzle I made for Shenzhen I/O, also known as "tis-100 in shenzhen io REAL?"

[vash3r]

breakthrough in current coding project, expect some more info soonish 😎

[vash3r]

@unittest.skip class DeadlockExample(AsyncTestCase): """ Example of how deadlock can occur. Make sure your test dependencies don't contain cycles! ;) """ @UnittestAsyncTest async def test1(self): print("starting test1 async") await self.test2 print("done test1 async") @UnittestAsyncTest async def test2(self): print("starting test2 async") await self.test1 print("done test2 async")

It Just Works(tm)

(The solution was to forget about inheritance and just write the magic methods myself while avoiding __get__ at all costs)

[vash3r]

i managed to boot into NOR but it now needs a password just to connect? wtf

[vash3r]

kernel reverse engineering is haaaard :(

i now understand exactly one (1) function in the windows kernel

[vash3r]

ugh, tried building openwrt/lede 17.01.2 and couldn’t compile `e2fsprogs-1.43.3/debugfs/create_inode.o` because of a conflicting function declaration for `copy_file_range` in `unistd.h`

time to clone a different openwrt branch and try again

(there are 4 different git repos with various states of openwrt and I have no idea if any of them will work)

[vash3r]

why is *formatted html* inside the api

it’s not even anywhere near consistent (you can tell it was made by several different people, many of which had wysiwyg editors): some entries have a bunch of <strong> and <em> tags while some other ones are just messes of 30+  ’s in a row, some of them have tables, some have multiple lines in one entry with <br> tags (instead of splitting across multiple entries).

in general it’s not really generally parsable because of the high variation.

[vash3r]

Windows Homestuck Color Emoji code uploaded

I’ve uploaded the Python code I recently used to generate my most recent Homestuck Color Emoji for Windows 10. (font file upload coming soon.) You can find it here!

Now you won’t have to wait for a homestuck-ified font to become available: Whenever Windows 10 updates their emoji files, you can easily modify them yourself! Not only can you keep your Homestuck-colored emoji, you also get the latest emoji instead of being stuck seeing blank boxes with an older version!

Note that this code requires Python 2.7 and fontTools. (Python 3 with 2to3 *may* work.) Also, it only works with Segoe UI Emoji, the default emoji font for Windows 10. There isn’t really a good general solution for emoji fonts; only case-by-case.

As usual, you can check out the Homestuck Color Emoji page on my blog. (It’ll be a bit out of date though, i haven’t updated it yet.)

(thanks to the people who interacted with my earlier post for motivation!)