mitzVah - the "worst" pangrams part 2

Apr 03, 2024

update apr 2: I accidentally deleted this article and now have re-posted it

in part 1, I introduced the spelling bee and the concept of the "worst" pangram, which I defined as the one which produced the fewest possible words. Then I wrote and described a program that found them, and labelled equivoke the Worst Pangramâ„¢.

I thought I was done getting nerdsniped after that, but a few people helpfully provided bits of information I was missing the first time, and I also thought a bit more about the problem to change its definition a bit.

more better information

People shared two bits of information with me that I didn't have when I started:

I took Brad's word list (thanks!) and added calculation of the word score and each pangram's total score. I didn't mention how words are scored in the first article:

This led to a funny result, where fuckwit was the new "worst" pangram:

back to the beginning

The initial prompt for this exploration was a question about whether there were any spelling bee puzzles where just the pangram would get you to the "genius" level.

I had changed the question to suit the knowledge I had at the start, but now I realized I had enough information to try and answer the original question.

I looked at that list and realized that if you set the puzzle as jacquard, and used the q as a required letter, you'd have only the pangram and three additional words: aqua, qajaq (an alternative spelling of kayak, apparently!) and quad. That would give you:

So jacQuard would result in a puzzle where the pangram was responsible for 15/22 points, or just a tiny hair below genius level. Could we do better?

I took the same program I used in part 1 and added a search through each letter of the pangram, scoring the result of making that letter required.

Here's a table of every pangram my program finds that would reach genius level all by itself:

score pangram words
14 mitzvah
14 princox
15 vagotomy
15 viburnum
15 conflux flux
15 jukebox jeux
15 cazique quiz
15 gazpacho
16 bovinity viny
16 checkbox exec
16 foxhound doxx
16 quixote exit,text
17 bikeway bike,kiwi,wiki
17 judicial jail,juju
17 exiguity exit,text
17 tubifex exit,ibex,text
17 activize zeta,ziti
18 jacquard card,crud,curd
18 highjack gaga,high,jagg
18 bullwhip whip,whup,will
18 oxazepam apex,exam,expo
19 gazpacho agog,gaga,goop,pogo
19 puffbird burp,drip,puff,purr
19 exiguity eggy,itty,yegg,yeti
19 buzzword bozo,buzz,orzo,ouzo
19 zucchetto chez,ooze,ouzo
20 mitzvoth vomit
20 vanguard guava
20 duckwalk claw,wack,walk,wall,waul
20 kiwifruit kiwi,twit,wiki,writ
20 quixotic toxic
21 bijective beet,bite,jibb,jibe,vibe
21 cheapjack hajj,jack,jake,jape,jeep
21 equivoke evoke,kook
21 jacquard aqua,aura,crud,curd,juju,quad
21 vibraharp brava
21 boxthorn hotbox
21 pyxidium mixup,pixy
22 tubificid biff,buff,cuff,duff,tiff,tuft
22 exemptive peeve,veep
22 ivorybill ivory,viol
22 phototoxic toxic
23 toxicologic toxic
24 epexegetic epigeic
25 aquacultural aqua,quart


Following the same start of the program I used to find the word count for every pangram in part 1, I added a couple more loops:

The implementation in python:

def wscore(w):
    return (1 if len(w) == 4 else len(w)) + (7 if len(allwordsets[w]) == 7 else 0)

# find all matches for each pangram, then the score for each possible required
# letter
scores_with_req_letters = []
pangram_matches = {}
for points, _, pangram in pangrams:
    ps = allwordsets[pangram]
    matches = [w for w in allwords if allwordsets[w].issubset(ps) and w != pangram]
    for l in ps:
        lmatches = [m for m in matches if l in m]
        pangram_matches[(pangram, l)] = lmatches
        score = len(pangram) + 7 + sum(wscore(m) for m in lmatches)
        scores_with_req_letters.append((score, l, pangram))

def hl(w: str, l: str) -> str:
    return f"{red}{w.replace(l, f'{yellow}{l}{red}')}"

red = "\N{esc}[0;31m"
green = "\N{esc}[0;32m"
yellow = "\N{esc}[0;33m"
blue = "\N{esc}[0;34m"
reset = "\N{esc}[0m"
for points, l, pangram in scores_with_req_letters:
    matches = pangram_matches[(pangram, l)]
    panscore = wscore(pangram)
    total = panscore + sum(wscore(m) for m in matches)
    if (panscore / total) > 0.7:
        matchstr = ",".join(matches)
        if len(matchstr) > 60:
            matchstr = matchstr[:59] + "..."
            f"{blue}{points: <8}{hl(pangram, l)}{' ' * (16-len(pangram))}{reset}{matchstr}"

I didn't dive much into performance, but here's a few notes:

With that, I feel like I can finally wash my hands of this problem! (Maybe? we'll see!)

↑ up